[RS6000] Bootstrap failure compiling xcoffout.c
[official-gcc.git] / gcc / ChangeLog
blob33739adf674edd849d39b8c0a9bf1fa952d6091d
1 2018-11-28  Alan Modra  <amodra@gmail.com>
3         * xcoffout.c (do_block): Signed/unsigned warning fix.
5 2018-11-28  Richard Biener  <rguenther@suse.de>
7         PR tree-optimization/79351
8         * tree-ssa-sccvn.c (vn_reference_lookup_3): For assignments from
9         empty CONSTRUCTORs ensure the store is at a constant position.
11 2018-11-28  Richard Biener  <rguenther@suse.de>
13         PR tree-optimization/88229
14         * tree-ssa.c (non_rewritable_mem_ref_base): Check DECL_SIZE_UNIT
15         is an INTEGER_CST before accessing it so.
17 2018-11-28  Sam Tebbs  <sam.tebbs@arm.com>
19         * doc/invoke.texi (-mtune=): Remove obsolete CPU names.
21 2018-11-28  Jakub Jelinek  <jakub@redhat.com>
23         PR target/88189
24         * config/i386/i386.c (ix86_expand_sse_movcc): Handle DFmode and
25         SFmode using sse4_1_blendvs[sd] with TARGET_SSE4_1.  Formatting fixes.
26         * config/i386/sse.md (sse4_1_blendv<ssemodesuffix>): New pattern.
28 2018-11-27  Martin Liska  <mliska@suse.cz>
30         * gcov.c (generate_results): Append current_working_directory
31         only when exists.
33 2018-11-27  Mike Gulick  <mgulick@mathworks.com>
35         PR preprocessor/83173
36         * input.c (dump_location_info): Dump reason and included_from
37         fields from line_map_ordinary struct.  Fix indentation when
38         location > 5 digits.
39         * diagnostic-show-locus.c (num_digits, num_digits): Move to
40         diagnostic.c to allow it to be utilized by input.c.
41         * diagnostic.c (num_digits, selftest::test_num_digits): Moved
42         here.
43         (selftest::diagnostic_c_tests): Run selftest::test_num_digits.
44         * diagnostic.h (num_digits): Add extern definition.
46 2018-11-27  Fredrik Noring  <noring@nocrew.org>
48         * config/mips/mips.c (mips_reorg_process_insns)
49         (mips_option_override): Handle `-mfix-r5900'.
50         * config/mips/mips.h (ASM_SPEC): Add `mfix-r5900' and
51         `mno-fix-r5900'.
52         * config/mips/mips.opt (mfix-r5900): New option.
53         * doc/invoke.texi: Document the `r5900' processor name, and
54         `-mfix-r5900' and `-mno-fix-r5900' options.
56 2018-11-27  Jakub Jelinek  <jakub@redhat.com>
58         PR target/88188
59         * config/rs6000/rs6000.c (print_operand) <case 'D'>: Use
60         output_operand_lossage instead of gcc_assert.
61         <case 't'>: Likewise.
62         <case 'z'>: Likewise.
63         <case 'V'>: Use output_operand_lossage instead of gcc_unreachable.
65 2018-11-27  Jeff Law  <law@redhat.com>
67         * config/riscv/riscv (riscv_block_mvoe_straight): Use RETURN_BEGIN
68         in call to move_by_pieces.
70         * config/sh/sh-mem.c (expand_block_move): Use RETURN_BEGIN in call
71         to move_by_pieces.
73         * config/lm32/lm32.c (lm32_block_move_inline): Use RETURN_BEGIN in
74         call to move_by_pieces.
76         * config/mips/mips.c (mips_block_move_straight): Use RETURN_BEGIN
77         in call to move_by_pieces.
79         * config/microblaze/microblaze.c (microblaze_block_move_straight): Use
80         RETURN_BEGIN in call to move_by_pieces.
81         (microblaze_expand_block_move): Likewise.
83 2018-11-27  Tamar Christina  <tamar.christina@arm.com>
85         * doc/md.texi (xorsign): Document it.
87 2018-11-27  Martin Liska  <mliska@suse.cz>
89         * config/gnu-user.h (TARGET_F951_OPTIONS): New.
90         * gcc.c (find_fortran_preinclude_file): New function
91         to handle Fortran pre-include.
93 2018-11-27  Martin Liska  <mliska@suse.cz>
95         * asan.c (asan_emit_stack_protection): Use new enum values
96         instead of int constants.
97         * builtins.c (expand_builtin_memory_copy_args): Replace int
98         type with memop_ret enum type.
99         (expand_builtin_mempcpy_args): Likewise.
100         (expand_builtin_memcpy): Use new enum values
101         instead of int constants. Likewise.
102         (expand_builtin_mempcpy): Likewise.
103         (expand_movstr): Likewise.
104         (expand_builtin_strcpy_args): Likewise.
105         (expand_builtin_stpcpy_1): Likewise.
106         (expand_builtin_strncpy): Likewise.
107         (expand_builtin_memset_args): Likewise.
108         * expr.c (move_by_pieces_d::finish_endp): Rename to ...
109         (move_by_pieces_d::finish_retmode): ... this.
110         (move_by_pieces): Change last argument type to memop_ret.
111         (store_by_pieces): Use new enum values
112         instead of int constants.
113         (emit_block_move_hints): Likewise.
114         (emit_push_insn): Likewise.
115         (store_expr): Likewise.
116         * expr.h (store_by_pieces): Change int to newly added enum
117         type.
118         * rtl.h (enum memop_ret): Define.
119         (move_by_pieces): Use the enum type.
121 2018-11-27  Alan Modra  <amodra@gmail.com>
123         * config/rs6000/aix71.h (ASM_SPEC): Don't select default -maix64
124         cpu here.
125         (ASM_CPU_SPEC): Do so here.  Rewrite using if .. else if .. specs
126         form.  Error on missing -mcpu case.
127         * config/rs6000/driver-rs6000.c (asm_names <_AIX>): Update NULL case.
128         (asm_names <!_AIX>): Add missing cpus.  Update NULL case.  Apply
129         PR63177 fix for -mcpu=power8 and -mcpu=powerpc64le.
130         * config/rs6000/rs6000.h (ASM_CPU_SPEC): Rewrite using if ..
131         else if .. specs form.  Error on missing -mcpu case.  Don't output
132         duplicate -maltivec.  Apply PR63177 fix for -mcpu=powerpc64le.
134 2018-11-26  David Malcolm  <dmalcolm@redhat.com>
136         * dump-context.h (dump_context::dump_loc): Convert 1st param from
137         dump_flags_t to const dump_metadata_t &.  Convert 2nd param from
138         const dump_location_t & to const dump_user_location_t &.
139         (dump_context::dump_loc_immediate): Convert 2nd param from
140         const dump_location_t & to const dump_user_location_t &.
141         (dump_context::dump_gimple_stmt): Convert 1st param from
142         dump_flags_t to const dump_metadata_t &.
143         (dump_context::void dump_gimple_stmt_loc): Likewise; convert
144         2nd param from const dump_location_t & to
145         const dump_user_location_t &.
146         (dump_context::dump_gimple_expr): Convert 1st param from
147         dump_flags_t to const dump_metadata_t &.
148         (dump_context::dump_gimple_expr_loc): Likewise; convert
149         2nd param from const dump_location_t & to
150         const dump_user_location_t &.
151         (dump_context::dump_generic_expr): Convert 1st param from
152         dump_flags_t to const dump_metadata_t &.
153         (dump_context::dump_generic_expr_loc): Likewise; convert
154         2nd param from const dump_location_t & to
155         const dump_user_location_t &.
156         (dump_context::dump_printf_va): Convert 1st param from
157         dump_flags_t to const dump_metadata_t &.
158         (dump_context::dump_printf_loc_va): Likewise; convert
159         2nd param from const dump_location_t & to
160         const dump_user_location_t &.
161         (dump_context::dump_dec): Convert 1st param from
162         dump_flags_t to const dump_metadata_t &.
163         (dump_context::dump_symtab_node): Likewise.
164         (dump_context::begin_scope): Split out 2nd param into
165         user and impl locations.
166         (dump_context::ensure_pending_optinfo): Add metadata param.
167         (dump_context::begin_next_optinfo): Replace dump_location_t param
168         with metadata and user location.
169         * dumpfile.c (dump_context::dump_loc): Convert 1st param from
170         dump_flags_t to const dump_metadata_t &.  Convert 2nd param from
171         const dump_location_t & to const dump_user_location_t &.
172         (dump_context::dump_loc_immediate): Convert 2nd param from
173         const dump_location_t & to const dump_user_location_t &.
174         (dump_context::dump_gimple_stmt): Convert 1st param from
175         dump_flags_t to const dump_metadata_t &.
176         (dump_context::void dump_gimple_stmt_loc): Likewise; convert
177         2nd param from const dump_location_t & to
178         const dump_user_location_t &.
179         (dump_context::dump_gimple_expr): Convert 1st param from
180         dump_flags_t to const dump_metadata_t &.
181         (dump_context::dump_gimple_expr_loc): Likewise; convert
182         2nd param from const dump_location_t & to
183         const dump_user_location_t &.
184         (dump_context::dump_generic_expr): Convert 1st param from
185         dump_flags_t to const dump_metadata_t &.
186         (dump_context::dump_generic_expr_loc): Likewise; convert
187         2nd param from const dump_location_t & to
188         const dump_user_location_t &.
189         (dump_context::dump_printf_va): Convert 1st param from
190         dump_flags_t to const dump_metadata_t &.
191         (dump_context::dump_printf_loc_va): Likewise; convert
192         2nd param from const dump_location_t & to
193         const dump_user_location_t &.
194         (dump_context::dump_dec): Convert 1st param from
195         dump_flags_t to const dump_metadata_t &.
196         (dump_context::dump_symtab_node): Likewise.
197         (dump_context::begin_scope): Split out 2nd param into
198         user and impl locations.
199         (dump_context::ensure_pending_optinfo): Add metadata param.
200         (dump_context::begin_next_optinfo): Replace dump_location_t param
201         with metadata and user location.
202         (dump_gimple_stmt): Convert 1st param from dump_flags_t to
203         const dump_metadata_t &.
204         (dump_gimple_stmt_loc): Likewise; convert 2nd param from
205         const dump_location_t & to const dump_user_location_t &.
206         (dump_gimple_expr): Convert 1st param from dump_flags_t to
207         const dump_metadata_t &.
208         (dump_gimple_expr_loc): Likewise; convert 2nd param from
209         const dump_location_t & to const dump_user_location_t &.
210         (dump_generic_expr): Convert 1st param from dump_flags_t to
211         const dump_metadata_t &.
212         (dump_generic_expr_loc): Likewise; convert 2nd param from
213         const dump_location_t & to const dump_user_location_t &.
214         (dump_printf): Convert 1st param from dump_flags_t to
215         const dump_metadata_t &.
216         (dump_printf_loc): Likewise; convert 2nd param from
217         const dump_location_t & to const dump_user_location_t &.
218         (dump_dec): Convert 1st param from dump_flags_t to
219         const dump_metadata_t &.
220         (dump_symtab_node): Likewise.
221         (dump_begin_scope): Split out 2nd param into user and impl
222         locations.
223         (selftest::assert_impl_location_eq): New function.
224         (ASSERT_IMPL_LOCATION_EQ): New macro.
225         (selftest::test_impl_location): Update to use
226         ASSERT_IMPL_LOCATION_EQ.
227         (selftest::test_capture_of_dump_calls): Convert "loc" to
228         dump_user_location_t.  Add ASSERT_IMPL_LOCATION_EQ throughout,
229         verifying line numbers of dump emissions.
230         * dumpfile.h (class dump_metadata_t): New class.
231         (dump_printf): Convert 1st param from dump_flags_t to
232         const dump_metadata_t &.
233         (dump_printf_loc): Likewise; convert 2nd param from
234         const dump_location_t & to const dump_user_location_t &.
235         (dump_generic_expr_loc): Likewise.
236         (dump_generic_expr): Convert 1st param from dump_flags_t to
237         const dump_metadata_t &.
238         (dump_gimple_stmt_loc): Likewise; convert 2nd param from
239         const dump_location_t & to const dump_user_location_t &.
240         (dump_gimple_stmt): Convert 1st param from dump_flags_t to
241         const dump_metadata_t &.
242         (dump_gimple_expr_loc): Likewise; convert 2nd param from
243         const dump_location_t & to const dump_user_location_t &.
244         (dump_gimple_expr): Convert 1st param from dump_flags_t to
245         const dump_metadata_t &.
246         (dump_symtab_node): Likewise.
247         (dump_dec): Likewise.
248         (dump_begin_scope): Split out 2nd param into user and impl
249         locations.
250         (auto_dump_scope::auto_dump_scope): Split "loc" param into a user
251         location and impl_location, and capture the impl_location.
252         (AUTO_DUMP_SCOPE): Rename param from LOC to USER_LOC.
253         * loop-unroll.c (report_unroll): Update for changes to
254         dump_printf_loc and dump_printf.
255         * opt-problem.cc (opt_problem::opt_problem): Update for change to
256         dump_loc.
257         * optinfo-emit-json.cc
258         (selftest::test_building_json_from_dump_calls): Convert "loc" from
259         dump_location_t to dump_user_location_t.
260         * optinfo.cc (optinfo::emit_for_opt_problem): Update for change in
261         dump_loc_immediate.
262         * profile.c (compute_branch_probabilities): Update for change to
263         dump_printf_loc.
264         * selftest.h (ASSERT_STR_CONTAINS_AT): New macro.
265         * tree-vect-slp.c (vect_print_slp_tree): Update for change to
266         dump_printf_loc.
268 2018-11-27  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
270         * doc/sourcebuild.texi: Document check_effective_target_int_eq_float
271         and check_effective_target_ptr_eq_long.
273 2018-11-27  Alan Modra  <amodra@gmail.com>
275         * config.gcc (powerpc*-*-freebsd*, powerpc-*-netbsd*),
276         (powerpc-*-eabisimaltivec*, powerpc-*-eabisim*, powerpc-*-elf*),
277         (powerpc-*-eabialtivec*, powerpc-*-eabi*, powerpc-*-rtems*),
278         (powerpc-wrs-vxworks*, powerpc-*-lynxos*, powerpcle-*-elf*),
279         (powerpcle-*-eabisim*, powerpcle-*-eabi*): Add gnu-user.h to tm_file.
280         * config/rs6000/freebsd.h (CPLUSPLUS_CPP_SPEC),
281         (LINK_GCC_C_SEQUENCE_SPEC): Undef.
282         (ASM_APP_ON, ASM_APP_OFF): Don't define.
283         * config/rs6000/freebsd64.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
284         * config/rs6000/lynx.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
285         * config/rs6000/linux64.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
286         * config/rs6000/netbsd.h (CPLUSPLUS_CPP_SPEC),
287         (LINK_GCC_C_SEQUENCE_SPEC): Undef.
288         * config/rs6000/rtems.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
289         * config/rs6000/sysv4.h (GNU_USER_TARGET_CRTI): Redefine.
290         (GNU_USER_TARGET_CRTN): Redefine.
291         (CC1_SPEC): Use GNU_USER_TARGET_CC1_SPEC.
292         (LIB_LINUX_SPEC): Use GNU_USER_TARGET_LIB_SPEC.
293         (CRTOFFLOADBEGIN, CRTOFFLOADEND): Don't define.
294         (STARTFILE_LINUX_SPEC): Define as GNU_USER_TARGET_STARTFILE_SPEC.
295         (ENDFILE_LINUX_SPEC): Define as GNU_USER_TARGET_ENDFILE_SPEC.
296         (UCLIBC_DYNAMIC_LINKER, CHOOSE_DYNAMIC_LINKER): Don't define.
297         (LINK_EH_SPEC): Don't define.
299 2018-11-26  Martin Sebor  <msebor@redhat.com>
301         PR 87756
302         * expr.c (string_constant): Handle top-level decls of all character
303         types and subobjects of narrow character type.
305 2018-11-27  Alan Modra  <amodra@gmail.com>
307         * config.gcc (powerpc*-*-linux*): Add linux.h to tm_file.
308         * config/rs6000/linux.h (TARGET_OS_CPP_BUILTINS): Use
309         GNU_USER_TARGET_OS_CPP_BUILTINS.
310         (RS6000_ABI_NAME): Define.
311         * config/rs6000/linux64.h (TARGET_OS_CPP_BUILTINS): Use
312         GNU_USER_TARGET_OS_CPP_BUILTINS.
313         (MUSL_DYNAMIC_LINKER32): Undef before defining.
314         (UCLIBC_DYNAMIC_LINKER32, UCLIBC_DYNAMIC_LINKER64): Don't define.
315         (CHOOSE_DYNAMIC_LINKER): Don't define.
316         (GNU_USER_DYNAMIC_LINKER32, GNU_USER_DYNAMIC_LINKER64): Don't define.
317         * config/rs6000/sysv4.h (MUSL_DYNAMIC_LINKER): Undef before defining.
318         (CHOOSE_DYNAMIC_LINKER, GNU_USER_DYNAMIC_LINKER): Only define when
319         not already defined.
320         (CPP_OS_LINUX_SPEC): Remove defines and asserts handled by
321         TARGET_OS_CPP_BUILTINS.
323 2018-11-26  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
325         * doc/sourcebuild.texi: Document check_effective_target_newlib_nano_io.
327 2018-11-26  Jakub Jelinek  <jakub@redhat.com>
329         PR target/88195
330         * config/i386/i386.c (def_builtin2): If tcode == VOID_FTYPE_UINT64
331         and !TARGET_64BIT, return NULL_TREE.
333         PR c++/86900
334         * dwarf2out.c (secname_for_decl): For functions with
335         DECL_SECTION_NAME if in_cold_section_p, try to return
336         current_function_section's name if it is a named section.
338 2018-11-26  Maya Rashish  <coypu@sdf.org>
340         PR target/58397
341         * config.host (*-*-netbsd*): Use x-netbsd and host-netbsd.o.
342         * config/x-netbsd: New file.
343         * config/host-netbsd: New file.
345 2018-11-26  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
347         * doc/sourcebuild.texi: Document check_effective_target_size20plus.
348         Clarify documentation for check_effective_target_size32plus.
350 2018-11-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
352         * config/rs6000/rs6000-string.c (expand_cmp_vec_sequence): Rename
353         and modify expand_strncmp_vec_sequence.
354         (emit_final_compare_vec): Rename and modify emit_final_str_compare_vec.
355         (generate_6432_conversion): New function.
356         (expand_block_compare): Add support for vsx.
357         (expand_block_compare_gpr): New function.
358         * config/rs6000/rs6000.opt (rs6000_block_compare_inline_limit): Increase
359         default limit to 63 because of more compact vsx code.
361 2018-11-26  Uros Bizjak  <ubizjak@gmail.com>
363         PR target/88178
364         * config/i386/i386.c (dbx_register_map): Use IGNORED_DWARF_REGNUM
365         for registers for which no debug information can be generated.
366         Use INVALID_REGNUM for non-existent registers.
367         (dbx64_register_map): Ditto.
368         (svr4_dbx_register_map): Ditto.
370 2018-11-26  Sandra Loosemore  <sandra@codesourcery.com>
372         PR c/57182
373         * doc/invoke.texi (Optimize Options): Document that
374         -freorder-functions also uses "hot" and "cold" attributes.
376 2018-11-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
378         * config/aarch64/aarch64.c (aarch64_mangle_type): Fix typo in comment.
380 2018-11-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
382         * config/arm/types.md (mul64): Rename to...
383         (widen_mul64): ... This.
384         * config/arm/arm-generic.md: Rename mul64 to widen_mul64.
385         * config/arm/cortex-a15.md: Likewise.
386         * config/arm/cortex-a5.md: Likewise.
387         * config/arm/cortex-a53.md: Likewise.
388         * config/arm/cortex-a57.md: Likewise.
389         * config/arm/cortex-a7.md: Likewise.
390         * config/arm/cortex-m4.md: Likewise.
391         * config/arm/exynos-m1.md: Likewise.
392         * config/arm/marvell-pj4.md: Likewise.
393         * config/arm/xgene1.md: Likewise.
395 2018-11-26  Richard Biener  <rguenther@suse.de>
397         PR tree-optimization/88182
398         * tree-vect-loop.c (vectorizable_reduction): Pick up single
399         correct reduc_def_info.
400         * tree-vect-slp.c (vect_analyze_slp_instance): Set
401         STMT_VINFO_REDUC_DEF of the first stmt.
403 2018-11-26  Andreas Krebbel  <krebbel@linux.ibm.com>
405         * doc/invoke.texi: Document z14/arch12 -march option.
407 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
409         PR c/57166
410         * doc/invoke.texi (Option Summary): Add -Wmissing-noreturn.
411         (Warning Options): Likewise.
413 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
415         PR web/79738
416         * doc/extend.texi (Common Function Attributes): Clarify that
417         functions with "const" attribute can read const global variables.
419 2018-11-26  Alan Modra  <amodra@gmail.com>
421         * config/rs6000/rs6000.h (OBJECT_PEF, TARGET_MACOS): Delete.
423 2018-11-25  Sandra Loosemore  <sandra@codesourcery.com>
425         PR other/54265
426         * doc/extend.texi (Common Variable Attributes): Use preferred
427         placement of type attributes in examples, plus whitespace fixes.
428         (Type Attributes): Clarify why placement of attributes
429         immediately after struct/union/enum keyword is preferred.
430         (Common Type Attributes): Use preferred placement of type
431         attributes in examples, plus more whitespace fixes.
433 2018-11-25  Paul Koning  <ni1d@arrl.net>
435         * config/pdp11/pdp11.h (TARGET_HAS_NO_HW_DIVIDE): Define.
437 2018-11-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
439         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
440         attributes): Document d_runtime.
442 2018-11-25  Vladimir Makarov  <vmakarov@redhat.com>
444         PR bootstrap/88157
445         * ira-costs.c (record_operand_costs): Initiate register move cost
446         for mode.
448 2018-11-23  Jeff Law  <law@redhat.com>
450         PR rtl-optimization/87468
451         * tree-ssa-threadupdate.c (create_block_for_threading): Clear
452         EDGE_IGNORE on all outgoing edges of the duplicate block.
454 2018-11-23  Vladimir Makarov  <vmakarov@redhat.com>
456         PR bootstrap/88157
457         * ira-costs.c (record_operand_costs): Use bigger hard reg class if
458         its mode does not fit to the original class.
460 2018-11-23  Martin Sebor  <msebor@redhat.com>
462         PR tree-optimization/87756
463         * expr.c (string_constant): Synthesize a string literal from
464         the address of a constant character.
465         * tree.c (build_string_literal): Add an argument.
466         * tree.h (build_string_literal): Same.
468 2018-11-23  Christoph Muellner  <christoph.muellner@theobroma-systems.com>
470         * config/aarch64/aarch64-cores.def: Define emag.
471         * config/aarch64/aarch64-tune.md: Regenerated with emag.
472         * config/aarch64/aarch64.c (emag_tunings): New struct.
473         * doc/invoke.texi: Document mtune value.
475 2018-11-23  Christoph Muellner  <christoph.muellner@theobroma-systems.com>
476             Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
478         * config/aarch64/aarch64.c (xgene1_tunings): Optimize Xgene1 tunings
479         for GCC 9.
480         * config/aarch64/aarch64.c (xgene1_tunings): Add Xgene1 specific
481         prefetch tunings.
482         * config/aarch64/aarch64.c (xgene1_addrcost_table): Correct the post-modify
483         costs.
484         * config/arm/aarch-cost-tables.h (xgene1_extra_costs): Update the cost table
485         for XGene1.
487 2018-11-23  Richard Biener  <rguenther@suse.de>
489         PR tree-optimization/88149
490         * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
491         the case where there are two different def types for the
492         same operand at different operand position in the same stmt.
494 2018-23-11  Mihail Ionescu  <mihail.ionescu@arm.com>
496         * config/arm/arm.c (arm_expand_compare_and_swap): Simplify and call
497         gen_atomic_compare_swap_1.
498         (arm_evpc_neon_vuzp): Likewise gen_neon_vuzp_internal.
499         (arm_evpc_neon_vtrn): Likewise gen_neon_vtrn_internal.
500         (arm_evpc_neon_vext): Likewise gen_neon_vext_internal.
501         (arm_evpc_neon_vzip): Likewise gen_neon_vzip_internal.
502         (arm_evpc_neon_vrev): Replace the function pointer and simplify the mode
503         checks.
504         * config/arm/arm.md (neon_vext<mode>),
505         (neon_vrev64<mode>, neon_vrev32<mode>),
506         (neon_vrev16<mode>, neon_vtrn<mode>_internal),
507         (neon_vzip<mode>_internal, neon_vuzp<mode>_internal): Add an '@'character
508         before the pattern name.
509         * config/arm/sync.md:
510         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1),
511         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
513 2018-11-23  Jakub Jelinek  <jakub@redhat.com>
515         PR tree-optimization/86614
516         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return early
517         if TREE_NO_WARNING is set on ref.ref.
519 2018-11-22  Sandra Loosemore  <sandra@codesourcery.com>
520             Alan Coopersmith  <alan.coopersmith@oracle.com>
522         PR c/53608
523         * doc/extend.texi (Designated Inits): Clarify handling of multiple
524         initializers for unions.
526 2018-11-22  Jan Hubicka  <jh@suse.cz>
528         PR lto/88142
529         * ipa-devirt.c (type_variants_equivalent_p): Drop warn and warned
530         parameters; do not warn here.
531         (odr_subtypes_equivalent_p): Likewise.
532         (warn_odr): Fix typo.
533         (warn_types_mismatch): Do not output confused warnings on integer types.
534         (odr_types_equivalent_p): Update.
536 2018-11-22  Jakub Jelinek  <jakub@redhat.com>
538         * config/i386/i386.c (ix86_option_override_internal,
539         ix86_can_inline_p, classify_argument, construct_container,
540         ix86_expand_prologue, ix86_expand_split_stack_prologue,
541         ix86_expand_carry_flag_compare, expand_set_or_movmem_via_loop,
542         expand_setmem_epilogue_via_loop, promote_duplicated_reg,
543         ix86_expand_set_or_movmem, ix86_init_builtins_va_builtins_abi):
544         Formatting fixes.
546         * config/i386/i386.c (ix86_option_override_internal): For
547         stack_protector_guard related options, use opts_set->x_ instead
548         of global_options_set. and prefix options with opts->x_ .  Move
549         defaults for offset and reg into else block.
551 2018-11-22  Uros Bizjak  <ubizjak@gmail.com>
553         * config/i386/i386.c (ix86_check_avx_upper_register):
554         Return true for all SSE registers with mode bitsize > 128.
556 2018-11-22  Vladimir Makarov  <vmakarov@redhat.com>
558         PR rtl-optimization/87718
559         * ira-costs.c: Remove trailing white-spaces.
560         (record_operand_costs): Add a special treatment for moves
561         involving a hard register.
563 2018-11-22  Uros Bizjak  <ubizjak@gmail.com>
565         * config/i386/i386.c (ix86_avx_emit_vzeroupper): Remove.
566         (ix86_emit_mode_set) <case AVX_U128>: Emit vzeroupper here.
568 2018-11-22  Martin Liska  <mliska@suse.cz>
570         * common/config/i386/i386-common.c (processor_names): Add
571         static assert and add missing "znver2".
572         (ix86_get_valid_option_values): Add checking assert for null
573         values and add "native" value if feasible.
574         * config/i386/i386.h: Do not declare size of processor_names.
575         * common/config/i386/i386-common.c:
576         * config/i386/i386.c: Add static assert for size
577         of processor_cost_table.
579 2018-11-22  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
581         * target-insns.def (stack_protect_combined_set): Define new standard
582         pattern name.
583         (stack_protect_combined_test): Likewise.
584         * cfgexpand.c (stack_protect_prologue): Try new
585         stack_protect_combined_set pattern first.
586         * function.c (stack_protect_epilogue): Try new
587         stack_protect_combined_test pattern first.
588         * config/arm/arm.c (require_pic_register): Add pic_reg and compute_now
589         parameters to control which register to use as PIC register and force
590         reloading PIC register respectively.  Insert in the stream of insns if
591         possible.
592         (legitimize_pic_address): Expose above new parameters in prototype and
593         adapt recursive calls accordingly.  Use pic_reg if non null instead of
594         cached one.
595         (arm_load_pic_register): Add pic_reg parameter and use it if non null.
596         (arm_legitimize_address): Adapt to new legitimize_pic_address
597         prototype.
598         (thumb_legitimize_address): Likewise.
599         (arm_emit_call_insn): Adapt to require_pic_register prototype change.
600         (arm_expand_prologue): Adapt to arm_load_pic_register prototype change.
601         (thumb1_expand_prologue): Likewise.
602         * config/arm/arm-protos.h (legitimize_pic_address): Adapt to prototype
603         change.
604         (arm_load_pic_register): Likewise.
605         * config/arm/predicated.md (guard_addr_operand): New predicate.
606         (guard_operand): New predicate.
607         * config/arm/arm.md (movsi expander): Adapt to legitimize_pic_address
608         prototype change.
609         (builtin_setjmp_receiver expander): Adapt to thumb1_expand_prologue
610         prototype change.
611         (stack_protect_combined_set): New expander..
612         (stack_protect_combined_set_insn): New insn_and_split pattern.
613         (stack_protect_set_insn): New insn pattern.
614         (stack_protect_combined_test): New expander.
615         (stack_protect_combined_test_insn): New insn_and_split pattern.
616         (arm_stack_protect_test_insn): New insn pattern.
617         * config/arm/thumb1.md (thumb1_stack_protect_test_insn): New insn pattern.
618         * config/arm/unspecs.md (UNSPEC_SP_SET): New unspec.
619         (UNSPEC_SP_TEST): Likewise.
620         * doc/md.texi (stack_protect_combined_set): Document new standard
621         pattern name.
622         (stack_protect_set): Clarify that the operand for guard's address is
623         legal.
624         (stack_protect_combined_test): Document new standard pattern name.
625         (stack_protect_test): Clarify that the operand for guard's address is
626         legal.
628 2018-11-22  Richard Biener  <rguenther@suse.de>
630         PR tree-optimization/88148
631         * tree-ssa-loop-niter.c (simplify_replace_tree): Get optional
632         valueization callback parameter and handle it.
633         * tree-ssa-loop-niter.h (simplify_replace_tree): Export.
634         * tree-ssa-sccvn.c (process_bb): Eliminate in loop niter trees.
636 2018-11-22  Richard Biener  <rguenther@suse.de>
638         PR lto/87229
639         PR lto/88112
640         * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs
641         which can appear in size expressions.
642         * tree-streamer-in.c (unpack_ts_base_value_fields): Stream
643         CALL_EXPR_BY_DESCRIPTOR.
644         (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN.
645         * tree-streamer-out.c (pack_ts_base_value_fields): Stream
646         CALL_EXPR_BY_DESCRIPTOR.
647         (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN.
649         Revert
650         PR lto/87229
651         * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
652         sizepos values.
654 2018-11-22  Richard Biener  <rguenther@suse.de>
656         PR tree-optimization/88069
657         * tree-ssa-sccvn.c (visit_phi): Tweak previous fix to not
658         apply to default defs.
660 2018-11-22  Jakub Jelinek  <jakub@redhat.com>
662         PR target/85644
663         PR target/86832
664         * config/i386/i386.c (ix86_option_override_internal): Default
665         ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
666         is defined.
667         * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
668         stack_protect_test, stack_protect_test_<mode>): Use empty condition
669         instead of TARGET_SSP_TLS_GUARD.
671 2018-11-22  Martin Liska  <mliska@suse.cz>
673         PR sanitizer/88017
674         * config/i386/i386.c (ix86_option_override_internal): Error
675         about usage -mabi=ms and -fsanitize=thread.
677 2018-11-21  Uros Bizjak  <ubizjak@gmail.com>
679         Revert the revert:
680         2013-10-26  Vladimir Makarov  <vmakarov@redhat.com>
682         Revert:
683         2013-10-25  Vladimir Makarov  <vmakarov@redhat.com>
685         * lra-spills.c (lra_final_code_change): Remove useless move insns.
687 2018-11-21  Uros Bizjak  <ubizjak@gmail.com>
689         PR middle-end/88129
690         * function.c (expand_function_end): Do not emit extra blockage insn.
692 2018-11-21  Lokesh Janghel  <lokeshjanghel91@gmail.com>
694         PR target/85667
695         * config/i386/i386.c (function_value_ms_64): Return AX_REG instead
696         of FIRST_SSE_REG for 4 or 8 byte modes.
698 2018-11-21  Jan Hubicka  <jh@suse.cz>
700         PR lto/87957
701         * tree.c (fld_decl_context): Break out from ...
702         (free_lang_data_in_decl): ... here; free TREE_PUBLIC, TREE_PRIVATE
703         DECL_ARTIFICIAL of TYPE_DECL; do not free TREE_TYPE of TYPE_DECL.
704         (fld_incomplete_type_of): Build copy of TYP_DECL.
705         * ipa-devirt.c (free_enum_values): Rename to ...
706         (free_odr_warning_data): ... this one; free also duplicated TYPE_DECLs
707         and TREE_TYPEs of TYPE_DECLs.
708         (get_odr_type): Initialize odr_vtable_hash if needed.
710 2018-11-21  Alexandre Oliva <oliva@adacore.com>
712         * final.c (compute_discriminator): Declare.  Renamed from...
713         (maybe_set_discriminator): ... this.  Set and return a local.
714         (override_discriminator): New.
715         (final_scan_insn_1): Set it.
716         (notice_source_line): Adjust.  Always set discriminator.
718 2018-11-21  Jakub Jelinek  <jakub@redhat.com>
720         PR target/87839
721         * config/aarch64/atomics.md (@aarch64_compare_and_swap<mode>): Use
722         rIJ constraint for aarch64_plus_operand rather than rn.
724 2018-11-21  Renlin Li  <renlin.li@arm.com>
726         PR middle-end/84877
727         * explow.h (get_dynamic_stack_size): Declare it as external.
728         * explow.c (record_new_stack_level): Remove function static attribute.
729         * function.c (assign_stack_local_1): Dynamically align the stack slot
730         addr for parameter copy on the stack.
732 2018-11-21  Richard Biener  <rguenther@suse.de>
734         PR bootstrap/88133
735         * bitmap.c (bitmap_last_set_bit): Refactor to avoid warning.
736         * Makefile.in (bitmap.o-warn): Remove again.
738 2018-11-20  Jeff Law  <law@redhat.com>
740         PR tree-optimization/88069
741         * tree-ssa-dom.c (record_equivalences_from_phis): Propagate away
742         degenerate virtual PHIs.
744 2018-11-21  H.J. Lu  <hongjiu.lu@intel.com>
746         PR target/87317
747         * config/i386/sse.md (sse4_1_<code>v8qiv8hi2<mask_name>): Replace
748         nonimmediate_operand with register_operand.
749         (avx2_<code>v8qiv8si2<mask_name>): Likewise.
750         (sse4_1_<code>v4qiv4si2<mask_name>): Likewise.
751         (sse4_1_<code>v4hiv4si2<mask_name>): Likewise.
752         (sse4_1_<code>v2qiv2di2<mask_name>): Likewise.
753         (avx512f_<code>v8qiv8di2<mask_name>): Likewise.
754         (avx2_<code>v4qiv4di2<mask_name>): Likewise.
755         (avx2_<code>v4hiv4di2<mask_name>): Likewise.
756         (sse4_1_<code>v2hiv2di2<mask_name>): Likewise.
757         (sse4_1_<code>v2siv2di2<mask_name>): Likewise.
758         (*sse4_1_<code>v8qiv8hi2<mask_name>_1): New pattern.
759         (*sse4_1_<code>v8qiv8hi2<mask_name>_2): Likewise.
760         (*avx2_<code>v8qiv8si2<mask_name>_1): Likewise.
761         (*avx2_<code>v8qiv8si2<mask_name>_2): Likewise.
762         (*sse4_1_<code>v4qiv4si2<mask_name>_1): Likewise.
763         (*sse4_1_<code>v4qiv4si2<mask_name>_2): Likewise.
764         (*sse4_1_<code>v4hiv4si2<mask_name>_1): Likewise.
765         (*sse4_1_<code>v4hiv4si2<mask_name>_2): Likewise.
766         (*avx512f_<code>v8qiv8di2<mask_name>_1): Likewise.
767         (*avx512f_<code>v8qiv8di2<mask_name>_2): Likewise.
768         (*avx2_<code>v4qiv4di2<mask_name>_1): Likewise.
769         (*avx2_<code>v4qiv4di2<mask_name>_2): Likewise.
770         (*avx2_<code>v4hiv4di2<mask_name>_1): Likewise.
771         (*avx2_<code>v4hiv4di2<mask_name>_2): Likewise.
772         (*sse4_1_<code>v2hiv2di2<mask_name>_1): Likewise.
773         (*sse4_1_<code>v2hiv2di2<mask_name>_2): Likewise.
774         (*sse4_1_<code>v2siv2di2<mask_name>_1): Likewise.
775         (*sse4_1_<code>v2siv2di2<mask_name>_2): Likewise.
777 2018-11-21  H.J. Lu  <hongjiu.lu@intel.com>
779         * read-rtl.c (apply_subst_iterator): Handle define_split and
780         define_insn_and_split.
782 2018-11-21  Jakub Jelinek  <jakub@redhat.com>
784         PR rtl-optimization/87817
785         * config/i386/i386.md (bmi2_bzhi_<mode>3, *bmi2_bzhi_<mode>3,
786         *bmi2_bzhi_<mode>3_1, *bmi2_bzhi_<mode>3_1_ccz): Use IF_THEN_ELSE
787         in the pattern to avoid triggering UB when operands[2] is zero.
788         (tbm_bextri_<mode>): New expander.  Renamed the old define_insn to ...
789         (*tbm_bextri_<mode>): ... this.
791 2018-11-21  Tom de Vries  <tdevries@suse.de>
793         PR driver/79855
794         * params.def (HOT_BB_COUNT_FRACTION): Terminate help message with
795         period.
797 2018-11-21  Andreas Krebbel  <krebbel@linux.ibm.com>
799         * configure.ac: Add check for Binutils to determine whether vector
800         load/store alignments hints are being supported.
801         * config.in: Regenerate.
802         * configure: Regenerate.
803         * config/s390/s390.c (print_operand): Support new output
804         modifier A.
805         * config/s390/s390.md ("movti"): Append alignment hint output
806         using the new output modifier 'A'.
807         * config/s390/vector.md ("mov<mode>", "*vec_tf_to_v1tf")
808         ("*vec_ti_to_v1ti"): Likewise.
810 2018-11-20  Martin Sebor  <msebor@redhat.com>
812         * doc/extend.texi (Other Builtins): Add __builtin_has_attribute.
814 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
816         PR lto/84044
817         * ipa-devirt.c (odr_types_equivalent_p): Use operand_equal_p to
818         compare ENUM values.
820 2018-11-20  Jakub Jelinek  <jakub@redhat.com>
822         PR tree-optimization/87895
823         * omp-simd-clone.c (ipa_simd_modify_function_body): When removing
824         or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT.
825         (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that
826         redirects edges to EXIT to edges to incr_bb, iterate while EXIT
827         has any preds and always use EDGE_PRED (, 0).
829 2018-11-20  Uros Bizjak  <ubizjak@gmail.com>
831         PR target/88070
832         * mode-switching.c (create_pre_exit): After reload, always split the
833         fallthrough edge to the exit block.
835 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
837         * ipa-devirt.c (add_type_duplicate): Do not ICE on incomplete enums.
838         * tree.c (build_array_type_1): Forward declare.
839         (fld_type_variant_equal_p): Add INNER_TYPE parameter.
840         (fld_type_variant): Likewise.
841         (fld_simplified_types): New hash.
842         (fld_process_array_type): New function.
843         (fld_incomplete_type_of): Handle array and enumeration types.
844         (fld_simplified_type): Handle simplification of arrays.
845         (free_lang_data): Allocate and free simplified types hash.
847 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
849         PR lto/87957
850         * ipa-devirt.c (odr_subtypes_equivalent_p): Report ODR violation
851         when sybtype already violates ODR.
852         (get_odr_type): Do not ICE when insert is false and type duplicate
853         is not registered yet.
854         (register_odr_type): Be sure to register subtypes first.
856 2018-11-20  Andreas Krebbel  <krebbel@linux.ibm.com>
858         * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the
859         DImode parts of the target operand.
861 2018-11-20  Nathan Sidwell  <nathan@acm.org>
863         PR 87926
864         * Makefile.in (bitmap.o-warn): Use -Wno-error=array-bounds.
866 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
868         PR lto/87997
869         * tree.c (free_lang_data_in_cgraph): Add argument fld; break out
870         type checking to...
871         (free_lang_data) ... here; update call of free_lang_data_in_cgraph.
873 2018-11-20  Jan Hubicka  <hubicka@ucw.cz>
875         PR ipa/87706
876         * ipa-fnsummary.c (pass_ipa_fnsummary): Do not remove functions
877         * ipa.c (possible_inline_candidate_p): Break out from ..
878         (process_references): ... here ; drop before_inlining_p;
879         cleanup handling of alises.
880         (walk_polymorphic_call_targets): Likewise.
881         (symbol_table::remove_unreachable_nodes): Likewise.
882         * passes.c (pass_data_ipa_remove_symbols): New structure.
883         (pass_ipa_remove_symbols): New pass.
884         (make_pass_ipa_remove_symbols): New function.
885         * tree-pass.h (make_pass_ipa_remove_symbols): Declare.
886         * passes.def (pass_ipa_remove_symbols): Schedule after early passes.
888 2018-11-20  Richard Biener  <rguenther@suse.de>
890         * tree-vect-stmts.c (vectorizable_condition): Do not get
891         at else_clause vect def for EXTRACT_LAST_REDUCTION.  Remove
892         pointless vect_is_simple_use calls.
894 2018-11-20  Richard Biener  <rguenther@suse.de>
896         PR tree-optimization/88074
897         * tree-ssa-dom.c (pass_dominator::execute): Do not walk
898         backedges.
900 2018-11-20  Richard Biener  <rguenther@suse.de>
902         PR tree-optimization/88069
903         * tree-ssa-sccvn.c (visit_phi): Do not value-number to unvisited
904         virtual PHI arguments.
906 2018-11-20  Ilya Leoshkevich  <iii@linux.ibm.com>
908         PR target/88083
909         * config/s390/s390.md: Skip LT(G) peephole when literal pool is
910         involved.
911         * rtl.h (contains_constant_pool_address_p): New function.
912         * rtlanal.c (contains_constant_pool_address_p): Likewise.
914 2018-11-20  Richard Biener  <rguenther@suse.de>
916         PR middle-end/83215
917         * alias.c (component_uses_parent_alias_set_from): Remove
918         alias-set zero and TYPE_TYPELESS_STORAGE case both already
919         handled in other ways.
921 2018-11-20  Richard Biener  <rguenther@suse.de>
923         PR tree-optimization/88087
924         * tree-ssa-pre.c (create_expression_by_pieces): Re-materialize
925         call fntype.
926         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Remember
927         call fntype.
929 2018-11-20  Richard Biener  <rguenther@suse.de>
931         PR middle-end/88089
932         * tree-data-ref.c (lambda_matrix_right_hermite): Use abs_hwi.
934 2018-11-19  Jan Hubicka  <hubicka@ucw.cz>
936         PR lto/87957
937         * ipa-devirt.c (free_enum_values): Do not ICE on ODR vilations.
939 2018-11-19  Sandra Loosemore  <sandra@codesourcery.com>
941         PR driver/50250
942         * doc/invoke.texi (Link Options): Mention shared libraries
943         in documentation for the -l option.  Simplify discussion and
944         point to the system linker documentation for details.
946 2018-11-19  Peter Bergner  <bergner@linux.ibm.com>
948         PR rtl-optimization/88033
949         * ira-lives.c (non_conflicting_reg_copy_p): Skip copies from a register
950         to itself.  Use HARD_REGISTER_NUM_P.
952 2018-11-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
954         * tree-vect-loop.c (vect_transform_loop): Disable further unrolling
955         of the loop if vf is non-constant.
957 2018-11-19  David Malcolm  <dmalcolm@redhat.com>
959         PR tree-optimization/87025
960         * dumpfile.c (dump_context::begin_scope): Call end_any_optinfo
961         immediately after creating the scope optinfo.
962         (selftest::test_pr87025): New function.
963         (selftest::dumpfile_c_tests): Call it.
964         * optinfo-emit-json.cc (optrecord_json_writer::pop_scope): Assert
965         that we're not popping the top-level records array.
966         * optinfo.cc (optinfo::handle_dump_file_kind): Assert that we're
967         not changing the kind of a "scope" optinfo.
969 2018-11-19  David Malcolm  <dmalcolm@redhat.com>
971         PR tree-optimization/87025
972         * dump-context.h: Include "optinfo.h".
973         (class optrecord_json_writer): New forward decl.
974         (dump_context::forcibly_enable_optinfo_p): Delete.
975         (dump_context::optinfo_enabled_p): New member function.
976         (dump_context::optimization_records_enabled_p): New member
977         function.
978         (dump_context::set_json_writer): New member function.
979         (dump_context::emit_optinfo): New member function.
980         (dump_context::m_forcibly_enable_optinfo): Delete.
981         (dump_context::m_json_writer): New member data.
982         * dumpfile.c (dump_context::set_json_writer): New member function.
983         (dump_context::finish_any_json_writer): New member function.
984         (dump_context::end_scope): Replace call to
985         optimization_records_maybe_pop_dump_scope with call to
986         m_json_writer->pop_scope.
987         (dump_context::optinfo_enabled_p): New member function.
988         (dump_context::end_any_optinfo): Replace call to optinfo::emit with call
989         to dump_context::emit_optinfo.
990         (dump_context::emit_optinfo): New member function.
991         (temp_dump_context::temp_dump_context): Replace
992         m_forcibly_enable_optinfo with call to set_json_writer.
993         (temp_dump_context::~temp_dump_context): Clean up any json writer.
994         * optinfo-emit-json.cc (class optrecord_json_writer): Move to
995         optinfo-emit-json.h
996         (the_json_writer): Delete.
997         (optimization_records_start): Delete.
998         (optimization_records_finish): Delete.
999         (optimization_records_enabled_p): Delete, in favor of
1000         dump_context::optimization_records_enabled_p.
1001         (optimization_records_maybe_record_optinfo): Delete.
1002         (optimization_records_maybe_pop_dump_scope): Delete.
1003         * optinfo-emit-json.h: Include "json.h".  Delete forward
1004         decl of opt_pass.
1005         (optimization_records_start): Delete.
1006         (optimization_records_finish): Delete.
1007         (optimization_records_enabled_p): Delete.
1008         (optimization_records_maybe_record_optinfo): Delete.
1009         (optimization_records_maybe_pop_dump_scope): Delete.
1010         (class optrecord_json_writer): Move here from
1011         optinfo-emit-json.cc.
1012         * optinfo.cc (optinfo::emit_for_opt_problem): Replace call
1013         to optinfo::emit with call to dump_context::emit_optinfo.
1014         (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
1015         (optinfo_enabled_p): Delete, in favor of
1016         dump_context::optinfo_enabled_p.
1017         (optinfo_wants_inlining_info_p): Update for conversion o
1018         optimization_records_enabled_p to a member function of
1019         dump_context.
1020         * optinfo.h (optinfo_enabled_p): Delete, in favor of
1021         dump_context::optinfo_enabled_p.
1022         (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
1023         * toplev.c: Include "dump-context.h".
1024         (compile_file): Replace call to optimization_records_finish with
1025         dump_context::finish_any_json_writer.
1026         (do_compile): Replace call to optimization_records_start with
1027         conditionally creating a optrecord_json_writer for the
1028         dump_context.
1030 2018-11-19  Jakub Jelinek  <jakub@redhat.com>
1032         PR tree-optimization/88071
1033         * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead
1034         of false as last argument to gsi_remove.
1035         * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of
1036         false as last argument to gsi_replace.
1038         PR debug/87039
1039         * omp-expand.c: Don't include debug.h.
1040         (adjust_context_and_scope): Add REGION argument.  Find DECL_CONTEXT
1041         from innermost outer parallel, task, teams or target that has a
1042         child_fn set, or, if there is no such outer region, use
1043         current_function_decl.  Do the DECL_CONTEXT adjustment regardless of
1044         whether a suitable BLOCK is found or not.
1045         (expand_parallel_call, expand_teams_call): Don't call
1046         adjust_context_and_scope here.
1047         (grid_expand_target_grid_body): Revert 2017-01-25 changes.
1048         (expand_omp_taskreg, expand_omp_target): Likewise.  Call
1049         adjust_context_and_scope.
1050         * dwarf2out.c (dwarf2out_early_global_decl): For
1051         decl_function_context recurse instead of calling dwarf2out_decl.
1053 2018-11-20  Eric Botcazou  <ebotcazou@adacore.com>
1055         PR rtl-optimization/85925
1056         * rtl.h (word_register_operation_p): New predicate.
1057         * combine.c (record_dead_and_set_regs_1): Only apply specific handling
1058         for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX.
1059         * rtlanal.c (nonzero_bits1): Likewise.  Adjust couple of comments.
1060         (num_sign_bit_copies1): Likewise.
1062 2018-11-19  Richard Biener  <rguenther@suse.de>
1064         PR lto/87229
1065         * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
1066         sizepos values.
1068 2018-11-19  Eric Botcazou  <ebotcazou@adacore.com>
1070         * compare-elim.c (struct comparison): Add not_in_a field.
1071         (is_not): New static function.
1072         (strip_not): Likewise.
1073         (conforming_compare): Handle a NOT in the first operand.
1074         (can_eliminate_compare): Likewise.
1075         (find_comparison_dom_walker::before_dom_children): Likewise.
1076         (try_eliminate_compare): Likewise.
1077         * config/visium/visium.md (negsi2_insn_set_carry): Turn into...
1078         (neg<mode>2_insn_set_carry): ...this and add missing NEG operation.
1080 2018-11-19  Jonathan Wakely  <jwakely@redhat.com>
1082         * doc/extend.texi (Common Type Attributes): Fix typos.
1084 2018-11-19  Martin Liska  <mliska@suse.cz>
1086         PR gcov-profile/88045
1087         * coverage.c (coverage_begin_function): Add assert.
1089 2018-11-18  Sandra Loosemore  <sandra@codesourcery.com>
1091         PR other/40498
1092         * doc/extend.texi (Common Function Attributes): Document that
1093         no_instrument_function applies to -p and -pg, too.
1094         * doc/invoke.texi (Instrumentation Options): Add cross-references
1095         to docs for -p, -pg, and -finstrument-functions.
1097 2018-11-18  Alan Modra  <amodra@gmail.com>
1099         * config/gnu-user.h (GNU_USER_TARGET_CRTI): Define.
1100         (GNU_USER_TARGET_STARTFILE_SPEC): Use it here.
1101         (GNU_USER_TARGET_CRTN): Define.
1102         (GNU_USER_TARGET_ENDFILE_SPEC): Use it here.
1103         (GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC): Define.
1105 2018-11-18  Alan Modra  <amodra@gmail.com>
1107         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Delete
1108         !HAVE_LD_PIE variant.
1109         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
1111 2018-11-17  Nick Clifton  <nickc@redhat.com>
1112             Sandra Loosemore  <sandra@codesourcery.com>
1114         PR driver/31357
1115         * doc/invoke.texi (Overall Options): Document that --help and
1116         --help=<value> options cannot be combined.
1118 2018-11-17  Jim Wilson  <jimw@sifive.com>
1120         * config/riscv/riscv.c (epilogue_cfa_sp_offset): New.
1121         (riscv_restore_reg): If restoring HARD_FRAME_POINTER_REGNUM, and
1122         epilogue_cfa_sp_offset set, then add REG_CFA_DEF_CFA regnote.
1123         (riscv_expand_epilogue): Initialize epilogue_cfa_sp_offset.  Set it
1124         to step2 if frame_pointer_needed and step1 is 0.
1126 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
1128         PR c++/4225
1129         * doc/cpp.texi (System Headers): Add note about implicit
1130         extern "C" block on targets that define SYSTEM_IMPLICIT_EXTERN_C.
1132 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
1134         PR c++/4025
1135         * doc/invoke.texi (C++ Dialect Options): Clarify usage of
1136         -fno-implicit-templates.
1138 2018-11-17  Jakub Jelinek  <jakub@redhat.com>
1140         PR tree-optimization/87546
1141         * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
1142         min_precision variable, initially set it to orig_precision, only does
1143         something if op_type's precision is <= min_precision and update
1144         min_precision whenever calling set_op.
1146 2018-11-16  Jan Hubicka  <hubicka@ucw.cz>
1148         PR ipa/87957
1149         * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
1151 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
1153         * doc/invoke.texi (Option Summary): Fix whitespace and line
1154         breaks in @gccoptlist environments.
1155         (Warning Options): Likewise.
1156         (Optimize Options): Likewise.
1157         (PowerPC SPE Options): Likewise.
1158         (RS/6000 and PowerPC Options): Likewise.
1160 2018-11-16  Jeff Law  <law@redhat.com>
1162         * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
1163         gcc_assert.
1165 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
1167         PR middle-end/23197
1168         * doc/invoke.texi (Optimize Options): Update options enabled by
1169         fprofile-generate, -fprofile-use, and -fauto-profile.
1171 2018-11-16  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1173         PR target/87927
1174         * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
1175         Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
1176         * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
1177         the byte_op hook.
1178         * target.h: Add psi, pdi, pti to struct asm_int_op definition.
1179         * targhooks.c (default_print_patchable_function_entry): Assert
1180         asm_int_op does not return a NULL string.
1181         * varasm.c (integer_asm_op): Return the op for a partial int type
1182         when the requested size does not correspond to an integer type.
1183         * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
1184         * doc/tm.texi: Regenerate.
1186 2018-11-12  Jason Merrill  <jason@redhat.com>
1188         * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
1190 2018-11-16  Michael Meissner  <meissner@linux.ibm.com>
1192         * config/rs6000/constraints.md (wF constraint): Remove power9
1193         fusion documentation.  Just document wF's use for power8 fusion.
1194         * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
1195         (fusion_gpr_addis): Delete power9 fusion support.  Change power8
1196         fusion support to require the upper 12 bits to be all 0's or all
1197         1's.
1198         (fusion_gpr_mem_load): Add comment.
1199         (fusion_addis_mem_combo_load): Remove power9 fusion support.
1200         (fusion_addis_mem_combo_store): Delete.
1201         (fusion_offsettable_mem_operand): Delete.
1202         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
1203         power8 fusion here.
1204         (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
1205         (POWERPC_MASKS): Delete power9 fusion.
1206         * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
1207         (fusion_p9_p): Delete.
1208         (expand_fusion_p9_load): Delete.
1209         (expand_fusion_p9_store): Delete.
1210         (emit_fusion_p9_load): Delete.
1211         (emit_fusion_p9_store): Delete.
1212         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
1213         fusion support.
1214         (rs6000_option_override_internal): Set power8 fusion based on
1215         whether we are tuning for power8.  Delete power9 fusion support.
1216         (rs6000_opt_masks): Delete -mpower9-fusion switch.
1217         (emit_fusion_load): Rename emit_fusion_load_store to
1218         emit_fusion_load, and drop fusion store support.  Update callers.
1219         (emit_fusion_load_store): Likewise.
1220         (emit_fusion_gpr_load): Likewise.
1221         (fusion_p9_p): Delete.
1222         (expand_fusion_p9_load): Delete.
1223         (expand_fusion_p9_store): Delete.
1224         (emit_fusion_p9_load): Delete.
1225         (emit_fusion_p9_store): Delete.
1226         * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
1227         (GPR_FUSION): Delete.
1228         (FPR_FUSION): Delete.
1229         (power9 fusion peephole2s): Delete.
1230         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
1231         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
1232         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
1233         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
1234         (fusion_p9_<mode>_constant): Delete.
1235         * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
1236         power9 fusion switch.
1237         * doc/md.texi (PowerPC constraints): Update wF constraint
1238         documentation for power8 fusion only.
1240 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
1242         PR rtl-optimization/87475
1243         * cfgrtl.c (patch_jump_insn): Allow redirection failure for
1244         CROSSING_JUMP_P insns.
1245         (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
1247 2018-11-16  Uros Bizjak  <ubizjak@gmail.com>
1249         PR target/88051
1250         * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
1251         * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
1252         (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
1254 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
1256         PR middle-end/88032
1257         * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
1258         is VOIDmode, in which case use int_mode.  Similarly for op1_mode.
1260 2018-11-16  Eric Botcazou  <ebotcazou@adacore.com>
1262         * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
1263         location_t parameter.
1264         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1265         (switch_decision_tree::do_jump_if_equal): Likewise.
1266         * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
1267         of switch statement to emit_case_nodes.
1268         (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
1269         set it on the newly built GIMPLE comparison statement.
1270         (switch_decision_tree::do_jump_if_equal): Likewise.
1271         (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
1272         in calls to do_jump_if_equal as well as recursive calls.
1274 2018-11-16  Andi Kleen  <ak@linux.intel.com>
1276         * config/i386/i386.md: Allow memory operands to ptwrite.
1278 2018-11-16  Michael Matz  <matz@suse.de>
1280         * system.h (PRsa): New macro.
1281         (SIZE_AMOUNT): Cast number to uint64_t.
1282         * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
1283         (pool_usage::dump_footer): Likewise and also use PRIu64.
1284         * bitmap.h (bitmap_usage::dump): Likewise.
1285         * ggc-common.c (ggc_usage::dump): Likewise.
1286         * ggc-page.c (ggc_print_statistics): Likewise.
1287         * input.c (dump_line_table_statistics): Likewise.
1288         * mem-stats.h (mem_usage::dump): Likewise.
1289         (mem_usage::dump_footer): Likewise.
1290         * rtl.c (dump_rtx_statistics): Likewise.
1291         * tree-cfg.c (dump_cfg_stats): Likewise.
1292         * tree-dfa.c (dump_dfa_stats): Likewise.
1293         * tree-phinodes.c (phinodes_print_statistics): Likewise.
1294         * tree-ssanames (ssanames_print_statistics): Likewise.
1295         * vec.c (vec_usage::dump): Likewise.
1296         (vec_usage::dump_footer): Likewise.
1298 2018-11-16  Richard Biener  <rguenther@suse.de>
1300         PR tree-optimization/88011
1301         * tree-vrp.c (extract_range_from_binary_expr): Fix error in
1302         replacing set_value_range_to_undefined and
1303         set_value_range_to_varying with method calls.
1305 2018-11-16  Ilya Leoshkevich  <iii@linux.ibm.com>
1307         * config/s390/s390.md
1308         (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
1309         assembler.
1310         (*r<noxa>sbg_<mode>_sll): Likewise.
1311         (*r<noxa>sbg_<mode>_srl): Likewise.
1312         (*r<noxa>sbg_sidi_srl): New pattern.
1314 2018-11-16  Jerome Lambourg  <lambourg@adacore.com>
1316         * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
1317         Default to TARGET_VXWORKS_RTP.
1318         (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
1319         of TARGET_VXWORKS_RTP.
1320         * config/vxworksae.h: Also define TARGET_VXWORKS_HAVE_CTORS_DTORS.
1321         * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
1322         of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
1324 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
1326         PR c++/25759
1327         * doc/extend.texi (Common Type Attributes): Make it explicit
1328         that attribute "packed" can apply to C++ classes.
1330 2018-11-15  Martin Sebor  <msebor@redhat.com>
1332         PR c++/87541
1333         PR c++/87542
1334         * tree.c (type_argument_type): New function.
1335         * tree.h (type_argument_type): Declare it.
1336         * gcc/doc/extend.texi (alloc_align): Update and clarify.
1337         (alloc_size, nonnull, sentinel): Same.
1339 2018-11-15  Andrew Stubbs  <ams@codesourcery.com>
1340             Kwok Cheung Yeung  <kcy@codesourcery.com>
1342         * tree-vect-stmts.c (vectorizable_store): Don't ICE when
1343         int_mode_for_size fails.
1344         (vectorizable_load): Likewise.
1346 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
1348         * doc/ux.texi (Group logically-related diagnostics): Move
1349         discussion of auto_diagnostic_group into this new subsection.
1350         Give an example of where this grouping is used.
1352 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
1354         PR other/19165
1355         * Makefile.in (OBJS): Move json.o to...
1356         (OBJS-libcommon): ...here and add diagnostic-format-json.o.
1357         * common.opt (fdiagnostics-format=): New option.
1358         (diagnostics_output_format): New enum.
1359         * diagnostic-format-json.cc: New file.
1360         * diagnostic.c (default_diagnostic_final_cb): New function, taken
1361         from start of diagnostic_finish.
1362         (diagnostic_initialize): Initialize final_cb to
1363         default_diagnostic_final_cb.
1364         (diagnostic_finish): Move "being treated as errors" messages to
1365         default_diagnostic_final_cb.  Call any final_cb.
1366         (default_diagnostic_finalizer): Add diagnostic_t param.
1367         (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
1368         diagnostic_finalizer callback.
1369         * diagnostic.h (enum diagnostics_output_format): New enum.
1370         (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
1371         (struct diagnostic_context): Add "final_cb".
1372         (default_diagnostic_finalizer): Add diagnostic_t param.
1373         (diagnostic_output_format_init): New decl.
1374         * doc/invoke.texi (-fdiagnostics-format): New option.
1375         * dwarf2out.c (gen_producer_string): Ignore
1376         OPT_fdiagnostics_format_.
1377         * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
1378         * lto-wrapper.c (append_diag_options): Ignore it.
1379         * opts.c (common_handle_option): Handle it.
1381 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
1383         PR tree-optimization/88015
1384         * graphite-isl-ast-to-gimple.c
1385         (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
1386         for dump_enabled_p.
1387         * graphite-sese-to-poly.c (build_poly_scop): Likewise.
1389 2018-11-15  Richard Biener  <rguenther@suse.de>
1391         PR middle-end/88029
1392         * gimple.c (gimple_call_flags): Union flags from decl, type
1393         and call fntype.
1394         * trans-mem.c (is_tm_pure_call): Simplify.
1396 2018-11-15  Richard Biener  <rguenther@suse.de>
1398         PR tree-optimization/88031
1399         * tree-vect-loop.c (vectorizable_reduction): Move check
1400         for multiple types earlier so we get the expected dump.
1401         Simplify calls to vectorizable_condition.
1402         * tree-vect-stmts.h (vectorizable_condition): Update prototype.
1403         * tree-vect-stmts.c (vectorizable_condition): Instead of
1404         reduc_def and reduc_index take just a flag.  Simplify
1405         code-generation now that we can rely on the defs being set up.
1406         (vectorizable_comparison): Remove unused argument.
1408 2018-11-15  Richard Biener  <rguenther@suse.de>
1410         PR tree-optimization/88030
1411         * tree-complex.c (need_eh_cleanup): New global.
1412         (update_complex_assignment): Mark blocks that need EH update.
1413         (expand_complex_comparison): Likewise.
1414         (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
1415         perform EH cleanup and schedule CFG cleanup if that did anything.
1417 2018-11-15  Jakub Jelinek  <jakub@redhat.com>
1419         PR rtl-optimization/88018
1420         * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
1421         on the presence of fallthru edge, rather than if it is a USE or not.
1423 2018-11-15  Richard Biener  <rguenther@suse.de>
1425         PR middle-end/87917
1426         * tree-data-ref.c (analyze_miv_subscript): Guard calls to
1427         analyze_subscript_affine_affine properly.
1429 2018-11-15  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1431         * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
1432         (GLIBC_DYNAMIC_LINKER): Define.
1433         (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
1435 2018-11-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
1437         PR tree-optimization/84648
1438         * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
1439         (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
1440         by calling adjust_cond_for_loop_until_wrap.
1442 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
1444         PR other/56334
1445         * doc/extend.texi (Common Function Attributes): Clarify linker
1446         restrictions on "aligned" attribute.
1447         (Common Variable Attributes): Likewise.  Mention that linker
1448         restrictions don't apply to stack-allocated variables.
1450 2018-11-15  Joern Rennecke  <joern.rennecke@riscy-ip.com>
1452         * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
1453         as an lvalue.
1455 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
1457         PR lto/55102
1458         PR lto/56700
1459         * doc/invoke.texi (Optimize Options): Remove bad example about
1460         interaction between -flto and -O.  Replace it with a note that
1461         you need to compile with -O and not just link.  Copy-edit -flto
1462         discussion to reduce verbiage and improve flow.
1464 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
1466         PR middle-end/59658
1467         * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
1468         also suppress many optimizations.  Alphabetize option lists for
1469         -O1, -O2, and -Os.  Add list of options disabled with -Og, and
1470         correct documentation for those options to say that.
1471         * opts.c (default_options_table): Sort table by level and option
1472         name, to make it easier to correlate to the manual.
1474 2018-11-14  Uros Bizjak  <ubizjak@gmail.com>
1476         * config/i386/i386.c (ix86_print_operand_address_as): Simplify
1477         printing of the names of segment registers.
1479 2018-11-14  Nathan Sidwell  <nathan@acm.org>
1481         PR debug/88006
1482         PR debug/87462
1483         * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
1484         type list.
1486 2018-11-14  David Malcolm  <dmalcolm@redhat.com>
1488         * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
1490 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
1492         P1236R1 - Signed integers are two's complement
1493         * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
1495         PR bootstrap/86739
1496         * hash-map.h (hash_map::iterator::reference_pair): New class.
1497         (hash_map::iterator::operator*): Return it rather than std::pair.
1499 2018-11-14  Jeff Law  <law@redhat.com>
1501         * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
1502         iff the operand is a constant.
1504 2018-11-14  Aldy Hernandez  <aldyh@redhat.com>
1506         * gimple-ssa-evrp-analyze.c
1507         (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
1508         ignore_equivs_equal_p to equal_p.
1509         * ipa-cp.c (meet_with_1): Use equal_p instead of
1510         ignore_equivs_equal_p.
1511         * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
1512         * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
1513         (value_range::operator==): Remove.
1514         (value_range::operator!=): Remove.
1515         (vrp_prop::visit_stmt): Use equal_p.
1516         * tree-vrp.h (value_range): Remove operator==, operator!=,
1517         ignore_equivs_equal_p.
1518         * vr-values.c (update_value_range): Use equal_p.
1520 2018-11-14  Michael Matz  <matz@suse.de>
1522         PR middle-end/86575
1523         * gimplify.c (collect_fallthrough_labels): Add new argument,
1524         return location via that, don't modify statements.
1525         (warn_implicit_fallthrough_r): Adjust call, don't use
1526         statement location directly.
1528 2018-11-14  Richard Biener  <rguenther@suse.de>
1530         PR middle-end/87985
1531         * tree-data-ref.c (split_constant_offset): Add wrapper
1532         allocating a cache hash-map.
1533         (split_constant_offset_1): Cache results of expanding
1534         expressions from SSA def stmts.
1536 2018-11-14  Richard Biener  <rguenther@suse.de>
1538         PR middle-end/88021
1539         * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
1540         to lambda_int.
1541         (lambda_vector_mult_const): Likewise.
1542         (lambda_matrix_right_hermite): Use lambda_int temporaries.
1544 2018-11-14  Wilco Dijkstra  <wdijkstr@arm.com>
1545             Jackson Woodruff  <jackson.woodruff@arm.com>
1547         PR 71026/tree-optimization
1548         * match.pd: Simplify floating point comparisons.
1550 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
1552         PR rtl-optimization/87817
1553         * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
1554         last argument has low 8 bits clear, fold to 0.
1556 2018-11-14  Iain Sandoe  <iain@sandoe.co.uk>
1558         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
1559         MAX_OFILE_ALIGNMENT as needed.
1561 2018-11-14  Richard Biener  <rguenther@suse.de>
1563         PR tree-optimization/88019
1564         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
1565         COND_EXPR arguments to non-trapping overflow.
1567 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
1569         * config/arc/arc.h (reg_class): Reorder registers classes, remove
1570         unused register classes.
1571         (REG_CLASS_NAMES): Likewise.
1572         (REG_CLASS_CONTENTS): Likewise.
1573         (FIXED_REGISTERS): Make lp_count fixed.
1574         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
1575         (PROGRAM_COUNTER_REGNO): Remove.
1576         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
1577         register classes, use constants for register numbers, remove
1578         reg_class_contents references.
1579         (arc_process_double_reg_moves): Add asserts.
1580         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
1581         lra_in_progress predicate.
1582         (arc_init_reg_tables): Remove unused register classes.
1583         (arc_register_move_cost): Likewise.
1584         (arc_preferred_reload_class): Likewise.
1585         (hwloop_optimize): Update rtx patterns involving lp_count register.
1586         (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
1587         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
1588         CHEAP_CORE_REGS.  Former one will be used for LRA.
1589         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
1590         one will be used for LRA.
1591         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
1592         one will be used for LRA.
1593         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
1594         Former one will be used for LRA.
1595         ("f"): Delete constraint.
1596         ("k"): Likewise.
1597         ("e"): Likewise.
1598         ("l"): Change it from register constraint to constraint.
1599         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
1600         (movhi_insn): Likewise.
1601         (movsi_insn): Update pattern.
1602         (arc_lp): Likewise.
1603         (dbnz): Likewise.
1604         (stack_tie): Remove 'b' constraint letter.
1605         (R4_REG): Define.
1606         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
1607         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
1608         (R57_REG, R59_REG, PCL_REG): Likewise.
1609         (ILINK1_REGNUM): Renamed to ILINK1_REG.
1610         (ILINK2_REGNUM): Renamed to ILINK2_REG.
1611         (Rgp): Remove.
1612         (SP_REGS): Likewise.
1613         (Rcw): Remove unused reg classes.
1614         * config/arc/predicates.md (dest_reg_operand): Just default on
1615         register_operand predicate.
1616         (mpy_dest_reg_operand): Likewise.
1617         (move_dest_operand): Use macros instead of constants.
1618         * config/arc/arc.opt (mlra): Switch to lra as default.
1620 2018-11-14  Richard Biener  <rguenther@suse.de>
1622         PR tree-optimization/87974
1623         * tree-vect-loop.c (vectorizable_reduction): When computing
1624         the vectorized reduction PHI vector type ignore constant
1625         and external defs.
1627 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
1629         PR tree-optimization/87977
1630         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
1631         stmt, build a new one and replace the old one with it.  Formatting fix.
1632         Call release_ssa_name (x) if !has_other_use and !delete_div.
1633         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
1634         verify lhs of stmt is still def.
1636 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
1638         PR rtl-optimization/87507
1639         * lower-subreg.c (operand_for_swap_move_operator): New function.
1640         (simple_move): Strip simple operators.
1641         (find_pseudo_copy): Likewise.
1642         (resolve_operand_for_swap_move_operator): New function.
1643         (resolve_simple_move): Strip simple operators and swap operands.
1645 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
1647         PR tree-optimization/87898
1648         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
1649         (ipa_simd_modify_function_body): Remove debug stmts where the first
1650         argument was changed into a non-decl.
1652 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
1654         PR rtl-optimization/87899
1655         * lra-lives.c (start_living): Update white space in comment.
1656         (enum point_type): New.
1657         (sparseset_contains_pseudos_p): New function.
1658         (update_pseudo_point): Likewise.
1659         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
1660         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
1661         handling.  Move early exit after adding conflicts.
1662         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
1663         if regno is already live.  Remove all handling of program points.
1664         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
1665         after adding conflicts.  Remove all handling of program points and
1666         ignore_reg_for_conflicts.
1667         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
1668         and do not guard call to mark_pseudo_live.
1669         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
1670         and do not guard call to mark_pseudo_dead.
1671         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
1672         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
1673         Use new function update_pseudo_point.  Handle register copies by
1674         removing the source register from the live set.  Handle INOUT operands.
1675         Update to the next program point using the unused_set, dead_set and
1676         start_dying sets.
1677         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
1679 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
1681         * builtins.c: Replace "source_location" with "location_t".
1682         * diagnostic-show-locus.c: Likewise.
1683         * diagnostic.c: Likewise.
1684         * dumpfile.c: Likewise.
1685         * gcc-rich-location.h: Likewise.
1686         * genmatch.c: Likewise.
1687         * gimple.h: Likewise.
1688         * gimplify.c: Likewise.
1689         * input.c: Likewise.
1690         * input.h: Likewise.  Eliminate the typedef.
1691         * omp-expand.c: Likewise.
1692         * selftest.h: Likewise.
1693         * substring-locations.h (get_source_location_for_substring):
1694         Rename to...
1695         (get_location_within_string): ...this.
1696         * tree-cfg.c: Replace "source_location" with "location_t".
1697         * tree-cfgcleanup.c: Likewise.
1698         * tree-diagnostic.c: Likewise.
1699         * tree-into-ssa.c: Likewise.
1700         * tree-outof-ssa.c: Likewise.
1701         * tree-parloops.c: Likewise.
1702         * tree-phinodes.c: Likewise.
1703         * tree-phinodes.h: Likewise.
1704         * tree-ssa-loop-ivopts.c: Likewise.
1705         * tree-ssa-loop-manip.c: Likewise.
1706         * tree-ssa-phiopt.c: Likewise.
1707         * tree-ssa-phiprop.c: Likewise.
1708         * tree-ssa-threadupdate.c: Likewise.
1709         * tree-ssa.c: Likewise.
1710         * tree-ssa.h: Likewise.
1711         * tree-vect-loop-manip.c: Likewise.
1713 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
1715         * doc/extend.texi: Fix typo in the weakref description.
1717 2018-11-13  Richard Biener  <rguenther@suse.de>
1719         PR tree-optimization/86991
1720         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
1721         group building until we have successfully detected the SLP
1722         reduction.
1723         (vect_is_simple_reduction): Remove fixup code here.
1725 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
1727         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
1728         (dump_gimple_stmt): Use it.
1729         (dump_gimple_stmt_loc): Likewise.
1730         (dump_gimple_expr): Likewise.
1731         (dump_gimple_expr_loc): Likewise.
1732         (dump_generic_expr): Likewise.
1733         (dump_generic_expr_loc): Likewise.
1734         (dump_printf): Likewise.
1735         (dump_printf_loc): Likewise.
1736         (dump_dec): Likewise.
1737         (dump_dec): Likewise.
1738         (dump_hex): Likewise.
1739         (dump_symtab_node): Likewise.
1740         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
1741         Guard dump call with dump_enabled_p.
1742         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
1743         * graphite-optimize-isl.c (optimize_isl): Likewise.
1744         * graphite.c (graphite_transform_loops): Likewise.
1745         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
1746         * tree-parloops.c (parallelize_loops): Likewise.
1747         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
1748         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
1749         (vect_prune_runtime_alias_test_list): Likewise.
1750         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
1751         (vect_estimate_min_profitable_iters): Likewise.
1752         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
1753         (vect_build_slp_tree_2): Likewise.
1754         (vect_supported_load_permutation_p): Likewise.
1755         (vect_slp_analyze_operations): Likewise.
1756         (vect_slp_analyze_bb_1): Likewise.
1757         (vect_slp_bb): Likewise.
1758         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
1759         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
1760         (pass_slp_vectorize::execute): Likewise.
1761         (increase_alignment): Likewise.
1763 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
1765         PR ipa/87955
1766         * ipa-inline.c (report_inline_failed_reason): Guard calls to
1767         cl_target_option_print_diff and cl_optimization_print_diff with
1768         if (dump_file).
1770 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
1772         * doc/invoke.texi (-fsave-optimization-record): Note that the
1773         output is compressed.
1774         * optinfo-emit-json.cc: Include <zlib.h>.
1775         (optrecord_json_writer::write): Compress the output.
1777 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
1779         * tree-vrp.c (value_range_base::dump): Dump type.
1780         Do not use INF nomenclature for 1-bit types.
1781         (dump_value_range): Group all variants to common dumping code.
1782         (debug): New overloaded functions for value_ranges.
1783         (value_range_base::dump): Remove no argument version.
1784         (value_range::dump): Same.
1786 2018-11-13  Richard Biener  <rguenther@suse.de>
1788         PR tree-optimization/87931
1789         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
1790         nested cycles we support to latch computations vectorizable_reduction
1791         handles.
1793 2018-11-13  Martin Liska  <mliska@suse.cz>
1795         PR tree-optimization/87885
1796         * cfghooks.c (account_profile_record): Rename
1797         to ...
1798         (profile_record_check_consistency): ... this.
1799         Calculate missing num_mismatched_freq_in.
1800         (profile_record_account_profile): New function
1801         that calculates time and size of a function.
1802         * cfghooks.h (struct profile_record): Remove
1803         all tuples.
1804         (struct cfg_hooks): Remove after_pass flag.
1805         (account_profile_record): Rename to ...
1806         (profile_record_check_consistency): ... this.
1807         (profile_record_account_profile): New.
1808         * cfgrtl.c (rtl_account_profile_record): Remove
1809         after_pass flag.
1810         * passes.c (check_profile_consistency): Do only
1811         checking.
1812         (account_profile): Calculate size and time of
1813         function only.
1814         (pass_manager::dump_profile_report): Reformat
1815         output.
1816         (execute_one_ipa_transform_pass): Call
1817         consistency check before clean upand call account_profile
1818         after a clean up is done.
1819         (execute_one_pass): Call check_profile_consistency and
1820         account_profile instead of using after_pass flag..
1821         * tree-cfg.c (gimple_account_profile_record): Likewise.
1823 2018-11-13  Martin Liska  <mliska@suse.cz>
1825         PR sanitizer/87930
1826         * config/i386/i386.c (ix86_option_override_internal): Error
1827         about usage -mabi=ms and -fsanitize={,kernel-}address.
1829 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
1831         * config/aarch64/aarch64.c
1832         (aarch64_vectorize_preferred_vector_alignment): Change return type to
1833         poly_uint64.
1834         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
1835         alignment being a poly int.
1836         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
1837         return type to poly_uint64.
1838         * target.def (default_preferred_vector_alignment): Likewise.
1839         * targhooks.c (default_preferred_vector_alignment): Likewise.
1840         * targhooks.h (default_preferred_vector_alignment): Likewise.
1841         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
1842         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
1843         poly int.
1844         (vect_update_misalignment_for_peel): Likewise.
1845         (vect_enhance_data_refs_alignment): Likewise.
1846         (vect_find_same_alignment_drs): Likewise.
1847         (vect_duplicate_ssa_name_ptr_info): Likewise.
1848         (vect_setup_realignment): Likewise.
1849         (vect_can_force_dr_alignment_p): Change alignment parameter type to
1850         poly_uint64.
1851         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
1852         mask with a compile time variable vector alignment.
1853         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
1854         int.
1855         (vect_do_peeling): Exit early if vector alignment is not constant.
1856         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
1857         a poly int.
1858         (vectorizable_store): Likewise.
1859         (vectorizable_load): Likweise.
1860         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
1861         poly_uint64.
1862         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
1863         poly int.
1864         (vect_can_force_dr_alignment_p): Change alignment parameter type to
1865         poly_uint64.
1867 2018-11-13  Richard Biener  <rguenther@suse.de>
1869         PR tree-optimization/87962
1870         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
1871         detect outer reduction for disqualifying in-loop uses.
1873 2018-11-13  Richard Biener  <rguenther@suse.de>
1875         PR tree-optimization/87967
1876         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
1877         for constants for the scalar loop.
1879 2018-11-13  Alan Modra  <amodra@gmail.com>
1881         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
1882         integer constants when -mcmodel=medium.
1884 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
1886         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
1887         * config/arc/elf.h (PROFILE_HOOK): Define.
1888         * config/arc/linux.h (PROFILE_HOOK): Likewise.
1890 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
1892         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
1893         jump table data in the text section.
1895 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
1897         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
1898         to fit the eh_return pattern.
1899         * config/arc/arc.md (eh_return): Define.
1900         (VUNSPEC_ARC_EH_RETURN): Likewise.
1901         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
1902         new implementation.
1903         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
1906 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
1908         * common/config/arc/arc-common.c (arc_option_optimization_table):
1909         Millicode optimization is default on for size optimizations.
1910         * config/arc/arc-protos.h (arc_check_multi): New function.
1911         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
1912         (ENTER_LEAVE_START_REG): Define.
1913         (ENTER_LEAVE_END_REG): Likewise.
1914         (arc_override_options): Disable millicode when long calls option
1915         is on.
1916         (arc_frame_info): Change it from int to bool.
1917         (arc_compute_frame_size): Clean up.
1918         (arc_save_restore): Remove.
1919         (frame_save_reg): New function.
1920         (frame_restore_reg): Likewise.
1921         (arc_enter_leave_p): Likewise.
1922         (arc_save_callee_saves): Likewise.
1923         (arc_restore_callee_saves): Likewise.
1924         (arc_save_callee_enter): Likewise.
1925         (arc_restore_callee_leave): Likewise.
1926         (arc_save_callee_milli): Likewise.
1927         (arc_restore_callee_milli): Likewise.
1928         (arc_expand_prologue): Reimplement to emit enter/leave
1929         instructions.
1930         (arc_expand_epilogue): Likewise.
1931         (arc_check_multi): New function.
1932         * config/arc/arc.md (push_multi_fp): New pattern.
1933         (push_multi_fp_blink): Likewise.
1934         (pop_multi_fp): Likewise.
1935         (pop_multi_fp_blink): Likewise.
1936         (pop_multi_fp_ret): Likewise.
1937         (pop_multi_fp_blink_ret): Likewise.
1938         * config/arc/arc.opt (mmillicode): Update option.
1939         (mcode-density-frame): New option.
1940         * config/arc/predicates.md (push_multi_operand): New predicate.
1941         (pop_multi_operand): Likewise.
1942         * doc/invoke.texi (ARC): Update ARC options information.
1945 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
1947         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
1948         * config/arc/arc.c (operands_ok_ldd_std): New function.
1949         (mem_ok_for_ldd_std): Likewise.
1950         (gen_operands_ldd_std): Likewise.
1951         * config/arc/arc.md: Add peephole2 rules for std/ldd.
1953 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
1955         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
1956         test on stack_usage_file.
1957         (lang_dependent_init): Do not open the .su file if generating LTO.
1959 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
1961         PR rtl-optimization/87918
1962         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
1963         simplify_gen_relational rather than simplify_gen_binary.
1965 2018-11-13  Richard Biener  <rguenther@suse.de>
1967         * tree-ssanames.h (set_range_info): Use value_range_base.
1968         (get_range_info): Likewise.
1969         * tree-ssanames.c (set_range_info): Likewise.
1970         (get_range_info): Likewise.
1971         * tree-vrp.c (value_range_base::union_helper): Split
1972         out common parts of value_range[_base]::union_.
1973         (value_range_base::union_): Update.
1974         (value_range::union_): Likewise.
1975         (determine_value_range_1): Use value_range_base.
1976         (determine_value_range): Likewise.
1977         * tree-vrp.h (value_range_base::union_helper): Move ...
1978         (value_range::union_helper): ... from here.
1980 2018-11-13  Alan Modra  <amodra@gmail.com>
1982         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
1983         offset for PRE_DEC.
1984         (rs6000_secondary_reload_gpr): Don't call find_replacement.
1986 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
1988         PR middle-end/59634
1989         * doc/invoke.texi (Optimize Options): Clarify that the
1990         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
1991         apply to data cache size.
1993 2018-11-13  Alan Modra  <amodra@gmail.com>
1995         * config/rs6000/predicates.md (logical_const_operand),
1996         (logical_operand): Correct comment.
1997         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
1999 2018-11-13  Alan Modra  <amodra@gmail.com>
2001         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
2003 2018-11-13  Alan Modra  <amodra@gmail.com>
2005         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
2006         description.
2007         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
2008         covered by alternative.
2009         (movcc_internal1): Ignore h for register preference.
2010         (mov<mode>_hardfloat64): Likewise.
2011         (mov<mode>_softfloat): Ignore c, l, h for register preference.
2013 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
2015         PR preprocessor/47823
2016         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
2017         documentation to...
2018         (Pragmas): ...here.
2019         * doc/extend.texi (Pragmas): Note additional pragmas documented
2020         in the CPP manual.
2022 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
2024         PR middle-end/86677
2025         PR middle-end/87528
2026         * tree-scalar-evolution.c (expression_expensive_p): Make
2027         BUILTIN POPCOUNT as expensive when backend does not define it.
2029 2018-11-12  Fredrik Noring  <noring@nocrew.org>
2031         * config.gcc: Update with-llsc defaults for MIPS r5900.
2033 2018-11-12  Martin Liska  <mliska@suse.cz>
2035         PR gcov-profile/87442
2036         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
2037         options.
2038         * doc/invoke.texi: Document them.
2039         * tree-profile.c (parse_profile_filter): New.
2040         (parse_profile_file_filtering): Likewise.
2041         (release_profile_file_filtering): Likewise.
2042         (include_source_file_for_profile): Likewise.
2043         (tree_profiling): Filter source files based on the
2044         newly added options.
2046 2018-11-12  Martin Liska  <mliska@suse.cz>
2048         PR target/87903
2049         * doc/extend.texi: Add missing values for __builtin_cpu_is and
2050         __builtin_cpu_supports for x86 target.
2052 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
2054         PR middle-end/21110
2055         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
2056         have to be Pmode.
2057         (Arithmetic): Likewise for "lo_sum".
2059 2018-11-12  Renlin Li  <renlin.li@arm.com>
2061         PR target/87815
2062         * dse.c (get_stored_val): Add check for compile-time constantness
2063         of gap.
2065 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
2067         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
2068         (ARMv8_5a): New fgroup.
2069         (armv8.5-a): New arch.
2070         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
2071         sb and predres.
2072         * config/arm/arm-tables.opt: Regenerate.
2073         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
2074         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
2075         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
2076         Add matching rules for -march=armv8.5-a and extensions.
2077         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
2078         Add sb and predres to all armv8-a except armv8.5-a.
2080 2018-11-12  Richard Biener  <rguenther@suse.de>
2082         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
2083         overload for single value.
2084         (value_range[_base]::set_nonnull): New.
2085         (value_range[_base]::set_null): Likewise.
2086         (value_range): Document bitmap copying behavior, mark
2087         copy constructor and assignment operator deleted.
2088         (value_range::move): New.
2089         (value_range::set_and_canonicalize): Default bitmap to zero.
2090         (set_value_range_to_nonnull): Remove.
2091         (set_value_range_to_null): Likewise.
2092         (set_value_range): Likewise.
2093         (set_value_range_to_value): Likewise.
2094         (extract_range_from_unary_expr): Work on value_range_base.
2095         (extract_range_from_binary_expr_1): Likewise.  Rename to...
2096         (extract_range_from_binary_expr): ... this.
2097         * tree-vrp.c (value_range::update): Clear equiv bitmap
2098         if required.
2099         (value_range::move): New, move equiv bitmap.
2100         (value_range_base::set_undefined): Avoid assignment.
2101         (value_range::set_undefined): Likewise.
2102         (value_range_base::set_varying): Likewise.
2103         (value_range::set_varying): Likewise.
2104         (set_value_range): Remove.
2105         (value_range_base::set): New overload for value.
2106         (value_range::set): Likewise.
2107         (set_value_range_to_nonnull): Remove.
2108         (value_range_base::set_nonnull): New.
2109         (value_range::set_nonnull): Likewise.
2110         (set_value_range_to_null): Remove.
2111         (value_range_base::set_null): New.
2112         (value_range::set_null): Likewise.
2113         (range_is_null): Work on value_range_base.
2114         (range_is_nonnull): Likewise.
2115         (ranges_from_anti_range): Likewise.
2116         (extract_range_into_wide_ints): Likewise.
2117         (extract_range_from_multiplicative_op): Likewise.
2118         (extract_range_from_binary_expr): Likewise.  Update for API changes.
2119         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
2120         handling.
2121         (value_range::intersect_helper): Avoid copy and assignment.
2122         (value_range::union_helper): Likewise.
2123         (determine_value_range_1): Adjust.
2124         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
2125         Avoid assignment by using move.
2126         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
2127         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
2128         Likewise.
2129         * tree-ssanames.c (get_range_info): Likewise.
2130         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
2131         * vr-values.c (vr_values::get_value_range): Adjust.
2132         (vr_values::update_value_range): Likewise.
2133         (symbolic_range_based_on_p): Work on value_range_base.
2134         (vr_values::extract_range_from_binary_expr): Use value_range_base.
2135         (vr_values::extract_range_from_unary_expr): Likewise.
2136         (vr_values::extract_range_from_cond_expr): Avoid assignment.
2137         (vr_values::extract_range_from_comparison): Adjust.
2138         (vr_values::check_for_binary_op_overflow): Use value_range_base.
2139         (vr_values::extract_range_basic): Adjust.
2140         (vr_values::adjust_range_with_scev): Likewise.
2141         (vr_values::vrp_visit_assignment_or_call): Likewise.
2142         (vr_values::get_vr_for_comparison): Change API to avoid
2143         assignment and copy construction.
2144         (vr_values::compare_name_with_value): Adjust accordingly.
2145         (vr_values::compare_names): Likewise.
2146         (vr_values::extract_range_from_phi_node): Avoid assignment and
2147         bogus in-place modify of equiv bitmap.
2148         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
2149         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
2150         for extract_range_from_unary_expr API change.
2151         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
2153 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
2155         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
2157 2018-11-12  Richard Biener  <rguenther@suse.de>
2159         * tree-vrp.h (value_range_base::symbolic_p,
2160         value_range_base::constant_p, value_range_base::zero_p,
2161         value_range_base::singleton_p): Move from value_range.
2162         (value_range::dump): Add.
2163         * gimple-ssa-evrp-analyze.c
2164         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
2165         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
2166         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
2167         Use set_varying.
2168         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
2169         (value_range::constant_p): Likewise.
2170         (value_range::singleton_p): Likewise.
2171         (value_range_base::dump): Add.
2172         (set_value_range_to_undefined): Remove.
2173         (set_value_range_to_varying): Likewise.
2174         (range_int_cst_p): Take value_range_base argument.
2175         (range_int_cst_singleton_p): Likewise.
2176         (value_range_constant_singleton): Likewise.
2177         (vrp_set_zero_nonzero_bits): Likewise.
2178         (extract_range_from_multiplicative_op): Use set_varying.
2179         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
2180         (extract_range_from_unary_expr): Likewise.
2181         (dump_value_range_base): Change to overload of dump_value_range.
2182         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
2183         (vrp_prop::visit_stmt): Likewise.
2184         (value_range::intersect_helper): Likewise.
2185         (value_range::union_helper): Likewise.
2186         (determine_value_range_1): Likewise.
2188 2018-11-12  Richard Biener  <rguenther@suse.de>
2190         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
2191         (set_value_range_to_null): Likewise.
2192         * vr-values.c (vr_values::extract_range_from_comparison):
2193         Clear equiv for constant singleton ranges.
2195 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
2197         * config/i386/sse.md: Combine VFIXUPIMM* patterns
2198         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2199         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2200         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
2201         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2202         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2203         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
2205 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
2207         PR c/69502
2208         * doc/extend.texi (Common Type Attributes): For the align type
2209         attribute, copy language about decreasing alignment from the
2210         corresponding variable attribute.
2212 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
2214         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
2215         -frounding-math.
2217 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
2219         PR c++/43105
2220         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
2221         -frtti and -fno-rtti code.
2223 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
2225         PR c/26366
2226         * doc/extend.texi (Other Builtins): Document probability associated
2227         with __builtin_expect.
2229 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
2231         PR target/87928
2232         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
2233         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
2234         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
2235         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
2237 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
2239         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
2241 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
2243         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
2245 2018-11-11  Richard Biener  <rguenther@suse.de>
2247         * tree-vrp.h (class value_range_base): New base class for
2248         value_range containing all but the m_equiv member.
2249         (dump_value_range_base): Add.
2250         (range_includes_zero_p): Work on value_range_base.
2251         * tree-vrp.c (value_range_base::set): Split out base handling
2252         from...
2253         (value_range::set): this.
2254         (value_range::set_equiv): New.
2255         (value_range_base::value_range_base): New constructors.
2256         (value_range_base::check): Split out base handling from...
2257         (value_range::check): this.
2258         (value_range::equal_p): Refactor in terms of
2259         ignore_equivs_equal_p which is now member of the base.
2260         (value_range_base::set_undefined): New.
2261         (value_range_base::set_varying): Likewise.
2262         (value_range_base::dump):Split out base handling from...
2263         (value_range::dump): this.
2264         (value_range_base::set_and_canonicalize): Split out base handling
2265         from...
2266         (value_range::set_and_canonicalize): this.
2267         (value_range_base::union_): New.
2268         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
2269         for m_vr.
2270         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
2271         instead of value_range everywhere.
2272         (ipcp_vr_lattice::print): Use dump_value_range_base.
2273         (ipcp_vr_lattice::meet_with): Adjust.
2274         (ipcp_vr_lattice::meet_with_1): Likewise.
2275         (ipa_vr_operation_and_type_effects): Likewise.
2276         (propagate_vr_across_jump_function): Likewise.
2277         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
2278         (ipa_get_value_range): Likewise.
2279         (ipa_set_jfunc_vr): Likewise.
2280         (ipa_compute_jump_functions_for_edge): Likewise.
2282 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
2284         PR middle-end/65703
2285         * doc/invoke.texi (Optimize Options): Add @opindex entries
2286         for the positive forms of -fno-xxx and -mno-xxx options
2287         that were lacking them.
2289 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
2291         * combine.c (make_more_copies): Only make an intermediate copy if the
2292         dest of a move is a pseudo.
2294 2018-11-09  Maya Rashish  <coypu@sdf.org>
2296         PR target/87221
2297         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
2298         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
2300 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
2302         PR driver/41179
2303         PR middle-end/65703
2304         * doc/invoke.texi (Optimize Options): Clarify default behavior
2305         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
2307 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2309         PR tree-optimization/87940
2310         * expr.c (string_constant): Don't strip NOPS in subexpressions.
2311         Fold PLUS_EXPR correctly.
2313 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
2315         PR target/87762
2316         * config/s390/s390.c (s390_safe_relative_long_p): New function.
2317         (annotate_constant_pool_refs): Skip insns which support
2318         relative addressing.
2319         (annotate_constant_pool_refs_1): New helper function.
2320         (find_constant_pool_ref): Skip insns which support relative
2321         addression.
2322         (find_constant_pool_ref_1): New helper function.
2323         (replace_constant_pool_ref): Skip insns which support
2324         relative addressing.
2325         (replace_constant_pool_ref_1): New helper function.
2326         (s390_mainpool_start): Adapt to the new signature.
2327         (s390_mainpool_finish): Likewise.
2328         (s390_chunkify_start): Likewise.
2329         (s390_chunkify_finish): Likewise.
2330         (pass_s390_early_mach::execute): Likewise.
2331         (s390_prologue_plus_offset): Likewise.
2332         (s390_emit_prologue): Likewise.
2333         (s390_emit_epilogue): Likewise.
2335 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
2337         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
2338         but unsupported lastprivate with conditional modifier.
2340 2018-11-09  Jeff Law  <law@redhat.com>
2342         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
2343         unused argument better.  Add gcc_unreachable to silence warning.
2345 2018-11-09  Martin Sebor  <msebor@redhat.com>
2347         PR middle-end/81824
2348         * attribs.c (has_attribute): New helper function.
2349         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
2350         * attribs.h (decls_mismatched_attributes): Declare.
2351         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
2352         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
2353         * common.opt (-Wattribute-alias): Take an argument.
2354         (-Wno-attribute-alias): New option.
2355         * doc/extend.texi (Common Function Attributes): Document copy.
2356         (Common Variable Attributes): Same.
2357         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
2358         (-Wattribute-alias): Document new option argument.
2360 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
2362         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
2363         matching criteria.  Remove unused array initializer.
2365 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
2366             Jinsong Ji  <jji@us.ibm.com>
2368         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
2369         constraints by introducing a new temporary.
2370         (_mm_cvtss_si64): Likewise.
2372 2018-11-09  Martin Liska  <mliska@suse.cz>
2374         * common.opt: Add -fipa-stack-alignment flag.
2375         * doc/invoke.texi: Document it.
2376         * final.c (rest_of_clean_state): Guard stack
2377         shrinking with flag.
2379 2018-11-09  Martin Liska  <mliska@suse.cz>
2381         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
2382         to ...
2383         (ipa_discover_variable_flags): ... this.
2384         * common.opt: Come up with new flag -fipa-reference-addressable.
2385         * doc/invoke.texi: Document it.
2386         * ipa-reference.c (propagate): Call the renamed fn.
2387         * ipa-visibility.c (whole_program_function_and_variable_visibility):
2388         Likewise.
2389         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
2390         ...
2391         (ipa_discover_variable_flags): ... this.  Discover
2392         non-addressable variables only with the newly added flag.
2393         * opts.c: Enable the newly added flag with -O1 and higher
2394         optimization level.
2396 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
2398         * json.cc (selftest::test_writing_literals): Fix comment.
2400 2018-11-09  Martin Liska  <mliska@suse.cz>
2402         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
2403         string to a stack buffer.
2404         (aarch64_parse_cpu): Likewise.
2405         (aarch64_parse_tune): Likewise.
2407 2018-11-09  Richard Biener  <rguenther@suse.de>
2409         PR tree-optimization/87953
2410         * tree-vect-loop.c (vectorizable_reduction): For analysis
2411         always pass ops[0] to vectorizable_condition.
2413 2018-11-09  Stafford Horne  <shorne@gmail.com>
2414             Richard Henderson  <rth@twiddle.net>
2415             Joel Sherrill  <joel@rtems.org>
2417         * common/config/or1k/or1k-common.c: New file.
2418         * config/or1k/*: New.
2419         * config.gcc (or1k*-*-*): New.
2420         * configure.ac (or1k*-*-*): New test for openrisc tls.
2421         * configure: Regenerated.
2422         * doc/install.texi: Document OpenRISC triplets.
2423         * doc/invoke.texi: Document OpenRISC arguments.
2424         * doc/md.texi: Document OpenRISC.
2426 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
2428         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
2429         (arm7tdmi-s): Delete CPU.
2430         (arm710t): Add aliases for arm720t and arm740t.
2431         (arm720t, arm740t): Delete CPUs.
2432         (arm920t): Add aliases for arm920, arm922t and arm940t.
2433         (arm920, arm922t, arm940t): Delete CPUs.
2434         (arm10tdmi): Add alias for arm1020t.
2435         (arm1020t): Delete CPU.
2436         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
2437         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
2438         (arm10e): Add aliases for arm1020e and arm1022e.
2439         (arm1020e, arm1022e): Delete CPU.
2440         * config/arm/arm.md (generic_sched): Remove entries that are now
2441         handled by aliases.
2442         (generic_vfp): Likewise.
2443         * config/arm/arm1020e.md: Simplify tuning selection based on alias
2444         changes.
2445         * config/arm/arm-tune.md: Regenerated.
2446         * config/arm/arm-tables.opt: Regenerated.
2448 2018-11-09  Richard Biener  <rguenther@suse.de>
2450         PR tree-optimization/87621
2451         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
2452         op with only phi inputs.
2453         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
2454         (ch_base::copy_headers): Run CSE on copied loop headers.
2455         (pass_ch_vect::process_loop_p): Simplify.
2457 2018-11-09  Alexandre Oliva <oliva@adacore.com>
2459         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
2460         for non-w64 x86_64 biarch.
2462 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
2464         PR rtl-optimization/86438
2465         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
2466         of in_b for the compare if in_b is SET_DEST.
2468         PR target/87793
2469         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
2470         non-toplevel UNSPEC.
2472 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
2474         * tree-vrp.c (value_range::check): Do not access internals
2475         directly.
2476         (value_range::singleton_p): Same.
2477         (value_range::type): Same.
2478         (vrp_finalize): Use value_range API.
2480 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
2482         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
2484 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
2486         * vr-values.c (vr_values::get_value_range): Use value_range API
2487         instead of piecing together ranges.
2488         (vr_values::update_value_range): Same.
2490 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
2492         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
2493         of performing ad-hoc calculations.
2494         * tree-ssanames.c (set_range_info): New overloaded function
2495         accepting value_range &.
2496         (get_range_info): Same.
2497         * tree-ssanames.h (set_range_info_raw): Remove.
2498         (set_range_info): New prototype.
2499         (get_range_info): Same.
2500         * tree-vrp.h (value_range::null_p): Rename to zero_p.
2501         * tree-vrp.c (value_range::null_p): Same.
2503 2018-11-09  Jan Hubicka  <jh@suse.cz>
2505         * tree.c (fld_type_variant_equal_p): Test user align flag.
2506         (flt_type_variant): Copy user align flag.
2507         (fld_incomplete_type_of): Clear it.
2509 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
2511         * config/arm/neon.md (div<mode>3): New pattern.
2513 2018-11-08  Andi Kleen  <ak@linux.intel.com>
2515         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
2516         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
2517         (ix86_handle_option): Handle OPT_mptwrite.
2518         * config/i386/cpuid.h (bit_PTWRITE): Add.
2519         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
2520         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
2521         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
2522         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
2523         (ix86_option_override_internal): Handle PTA_PTWRITE.
2524         (ix86_valid_target_attribute_inner_p): Define ptwrite.
2525         (def_builtin2): Force UINT64 to be 64bit only.
2526         * config/i386/i386.h (TARGET_PTWRITE): Add.
2527         (TARGET_PTWRITE_P): Add.
2528         (PTA_PTWRITE): Add.
2529         * config/i386/i386.md: Define ptwrite.
2530         * config/i386/i386.opt: Add -mptwrite.
2531         * config/i386/immintrin.h (_ptwrite64): Add.
2532         (_ptwrite32): Add
2533         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
2534         * doc/invoke.texi: Document -mptwrite.
2536 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
2538         PR rtl-optimization/87600
2539         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
2540         * lra-constraints.c (process_alt_operands): Skip illegal hard
2541         register usage.  Prefer reloading non hard register operands.
2543 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
2545         PR other/36572
2546         * doc/invoke.texi (Optimize Options): Clarify default behavior
2547         for -fno-sched-interblock and -fno-sched-spec.
2549 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
2551         * collect2.c (linker_select):  Add USE_LLD_LD.
2552         (ld_suffixes): Add ld.lld.
2553         (main): Handle -fuse-ld=lld.
2554         * common.opt (-fuse-ld=lld): New option.
2555         * doc/invoke.texi (-fuse-ld=lld): Document.
2556         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
2558 2018-11-08  Paul Koning  <ni1d@arrl.net>
2560         * config/pdp11/constraints.md: Add "Z" series constraints for use
2561         with pre-dec and post-inc addressing.
2562         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
2563         (pdp11_expand_operands): Add int argument (word count).
2564         (pdp11_sp_frame_offset): Delete.
2565         (pdp11_cmp_length): New function.
2566         (pushpop_regeq): New function.
2567         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
2568         Add hook.
2569         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
2570         frame layout.
2571         (pdp11_initial_elimination_offset): Ditto.
2572         (pdp11_expand_operands): Add word count argument.  Bugfixes.
2573         (output_move_multiple): Change how pointer adjustment is done.
2574         (pdp11_gen_int_label): Correct format.
2575         (output_ascii): Ditto.
2576         (pdp11_asm_output_var): Add code for DEC assembler case.
2577         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
2578         value.
2579         (legitimate_const_double_p): Ditto.
2580         (pdp11_register_move_cost): Adjust for new register classes.
2581         (pdp11_regno_reg_class): Ditto.
2582         (expand_block_move): Delete.
2583         (pushpop_regeq): New function.
2584         (pdp11_legitimate_address_p): Bugfix in check for constant
2585         offset.
2586         (pdp11_sp_frame_offset): Delete.
2587         (pdp11_reg_save_size): New helper function for new frame layout.
2588         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
2589         (pdp11_expand_shift): Bugfix in check for constant shift count.
2590         (pdp11_shift_length): Ditto.
2591         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
2592         (pdp11_cmp_length): New function.
2593         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
2594         some compile options.
2595         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
2596         (CALL_USED_REGISTERS): Ditto.
2597         (ELIMINABLE_REGS): Ditto.
2598         (REGISTER_NAMES): Ditto.
2599         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
2600         constraints.
2601         (REG_CLASS_NAMES): Ditto.
2602         (REG_CLASS_CONTENTS): Ditto.  Also remove
2603         HARD_FRAME_POINTER_REGNUM.
2604         (CPU_REG_CLASS): New macro.
2605         (CLASS_MAX_NREGS): Adjust for new register classes.
2606         (FUNCTION_PROFILER): Make no-op.
2607         (may_call_alloca): Remove unused declaration.
2608         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
2609         (ASM_OUTPUT_SKIP): Fix format.
2610         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
2611         (HARD_FRAME_POINTER_REGNUM): Remove.
2612         (return): Delete.
2613         (*rts): Rename.  Remove epilogue related checks.
2614         (cmpsi, cmpdi): New insn.
2615         (cbranch<mode>4): Change to apply to SI and DI modes as well.
2616         (mov<mode>): Change constraints to enforce that push/pop
2617         destination cannot use the same register as source.
2618         (*mov<mode><cc_cc>): Ditto.
2619         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
2620         at assembly output rather than as RTL expander.
2621         (zero_extendqihi2): Bugfix in check for same registers.
2622         (adddi3_nocc): Bugfix in check for constant operand.
2623         (addsi3_nocc): Ditto.
2624         (subdi3_nocc): Ditto.
2625         (subsi3_nocc): Ditto.
2626         (negdi2_nocc): Copy input to pdp11_expand_operands.
2627         (negsi2_nocc): Ditto.
2628         (bswap2_nocc): Ditto.
2629         * config/pdp11/pdp11.opt (mlra): Fix documentation.
2630         * config/pdp11/t-pdp11: Use -Os.
2632 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
2634         * config/arm/parsecpu.awk (/alias/): New parsing rule.
2635         (/begin cpu/): Check that the cpu name hasn't been previously defined.
2636         (gen_comm_data): Print out CPU alias tables.
2637         (check_cpu): Match aliases when checking the CPU name.
2638         * config/arm/arm-protos.h (cpu_alias): New structure.
2639         (cpu_option): Add entry for aliases.
2640         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
2641         strongarm1100 and strongarm1110.
2642         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
2643         (config/arm/arm-generic.md): Remove redundant references to
2644         strongarm110, strongarm1100 and strongarm1110.
2645         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2646         Scan aliases for additional hints.
2647         (arm_parse_cpu_option_name): Also match a cpu name against the list
2648         of aliases.
2649         * config/arm/arm-tables.opt: Regenerated.
2650         * config/arm/arm-tune.md: Regenerated.
2652 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
2654         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
2655         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
2656         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
2657         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
2658         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
2659         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
2660         * gengtype.c (open_base_files): Add omp-general.h.
2661         * gimple.c (gimple_build_omp_critical):
2662         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
2663         gimple_omp_taskgroup_set_clauses.
2664         (gimple_build_omp_atomic_load): Add mo argument, call
2665         gimple_omp_atomic_set_memory_order.
2666         (gimple_build_omp_atomic_store): Likewise.
2667         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
2668         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
2669         instead of GSS_OMP.
2670         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
2671         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
2672         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
2673         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
2674         different value for GF_OMP_ATOMIC_NEED_VALUE.
2675         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
2676         comments.
2677         (struct gimple_statement_omp_single_layout): And remove here.
2678         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
2679         than gimple_statement_omp_single_layout.
2680         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
2681         GIMPLE_OMP_TEAMS.
2682         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
2683         (gimple_omp_subcode): Formatting fix.
2684         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
2685         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
2686         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
2687         gimple_omp_teams_host, gimple_omp_teams_set_host,
2688         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
2689         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
2690         gimple_omp_taskgroup_set_clauses): New inline functions.
2691         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
2692         (gimple_build_omp_atomic_store): Likewise.
2693         (gimple_omp_atomic_seq_cst_p): Remove.
2694         (gimple_omp_atomic_memory_order): New function.
2695         (gimple_omp_atomic_set_seq_cst): Remove.
2696         (gimple_omp_atomic_set_memory_order): New function.
2697         (gimple_build_omp_taskgroup): Add clauses argument.
2698         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
2699         (dump_gimple_omp_task): Print taskwait with depend clauses.
2700         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
2701         dump_omp_atomic_memory_order.
2702         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
2703         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
2704         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
2705         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
2706         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
2707         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
2708         ORT_UNTIED_TASKLOOP enumerators.
2709         (enum gimplify_defaultmap_kind): New.
2710         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
2711         target_map_pointers_as_0len_arrays members, add defaultmap.
2712         (new_omp_context): Initialize defaultmap member.
2713         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
2714         (maybe_fold_stmt): Don't fold even in host teams regions.
2715         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
2716         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
2717         ctx->omp_firstprivatize_variable.
2718         (omp_add_variable): Don't add private/firstprivate for VLAs in
2719         ORT_TASKGROUP.
2720         (omp_default_clause): Print "taskloop" rather than "task" if
2721         ORT_*TASKLOOP.
2722         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
2723         Handle new defaultmap clause kinds.
2724         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
2725         iterator to be lastprivate or private.  Fix up diagnostics if linear
2726         is used on collapse>1 simd iterator.
2727         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
2728         (gimplify_omp_depend): New function.
2729         (gimplify_scan_omp_clauses): Add shared clause on parallel for
2730         combined parallel master taskloop{, simd} if taskloop has
2731         firstprivate, lastprivate or reduction clause.  Handle
2732         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
2733         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
2734         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
2735         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
2736         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
2737         lastprivate.
2738         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
2739         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
2740         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
2741         OMP_CLAUSE_{TASK,IN}_REDUCTION.
2742         (gimplify_omp_task): Handle taskwait with depend clauses.
2743         (gimplify_omp_for): Add shared clause on parallel for combined
2744         parallel master taskloop{, simd} if taskloop has firstprivate,
2745         lastprivate or reduction clause.  Use ORT_TASKLOOP or
2746         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
2747         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
2748         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
2749         __for_end and __for_range temporaries on OMP_PARALLEL for
2750         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
2751         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
2752         sandwiched in between two taskloops.
2753         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
2754         instead of ctx->omp_firstprivatize_variable.
2755         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
2756         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
2757         host teams, use gimplify_and_return_first etc. for body like
2758         for target or target data constructs, and at the end call
2759         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
2760         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
2761         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
2762         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
2763         gimple_omp_atomic_set_seq_cst calls.
2764         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
2765         case, handle taskgroup clauses.
2766         * lto-streamer-out.c (hash_tree): Handle
2767         OMP_CLAUSE_{TASK,IN}_REDUCTION.
2768         * Makefile.in (GTFILES): Add omp-general.h.
2769         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
2770         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
2771         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
2772         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
2773         BUILT_IN_GOMP_LOOP_DOACROSS_START,
2774         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
2775         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
2776         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
2777         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
2778         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
2779         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
2780         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
2781         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
2782         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
2783         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
2784         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
2785         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
2786         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
2787         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
2788         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
2789         * omp-expand.c (workshare_safe_to_combine_p): Return false for
2790         non-worksharing loops.
2791         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
2792         (determine_parallel_type): Don't combine parallel with worksharing
2793         which has _reductemp_ clause.
2794         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
2795         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
2796         if there is nonmonotonic modifier or if there is no modifier and no
2797         ordered clause.  For dynamic and guided schedule without monotonic
2798         and nonmonotonic modifier, default to nonmonotonic.
2799         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
2800         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
2801         task reductions.
2802         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
2803         there are any reduction clauses.
2804         (expand_taskwait_call): New function.
2805         (expand_teams_call): New function.
2806         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
2807         expand_teams_call for it.  Formatting fix.  Handle taskwait with
2808         depend clauses.
2809         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
2810         of worksharing loops with task reductions.
2811         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
2812         expansion of worksharing loops with task reductions.
2813         (expand_omp_sections): Handle expansion of sections with task
2814         reductions.
2815         (expand_omp_synch): For host teams call expand_omp_taskreg.
2816         (omp_memory_order_to_memmodel): New function.
2817         (expand_omp_atomic_load, expand_omp_atomic_store,
2818         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
2819         instead of gimple_omp_atomic_seq_cst_p.
2820         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
2821         depend clauses as a standalone directive.
2822         * omp-general.c (enum omp_requires): New variable.
2823         (omp_extract_for_data): Initialize have_reductemp member.  Allow
2824         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
2825         GT_EXPR loops depending on incr sign.  Formatting fixes.
2826         * omp-general.h (struct omp_for_data): Add have_reductemp member.
2827         (enum omp_requires): New enum.
2828         (omp_requires_mask): Declare.
2829         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
2830         Fix comment typos.
2831         * omp-low.c (struct omp_context): Add task_reductions and
2832         task_reduction_map fields.
2833         (is_host_teams_ctx): New function.
2834         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
2835         (use_pointer_for_field): Use is_global_var instead of
2836         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
2837         in outer contexts.
2838         (build_outer_var_ref): Ignore taskgroup outer contexts.
2839         (delete_omp_context): Release task_reductions and task_reduction_map.
2840         (scan_sharing_clauses): Don't add any fields for reduction clause on
2841         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
2842         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
2843         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
2844         Handle OMP_CLAUSE_NONTEMPORAL.
2845         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
2846         needed.
2847         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
2848         clauses with task modifier.
2849         (scan_omp_task): Handle taskwait with depend clauses.
2850         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
2851         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
2852         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
2853         (scan_omp_for): Fix comment formatting.
2854         (scan_omp_teams): Handle host teams constructs.
2855         (check_omp_nesting_restrictions): Allow teams with no outer
2856         OpenMP context.  Adjust diagnostics for teams strictly nested into
2857         some explicit OpenMP construct other than target.  Allow OpenMP atomics
2858         inside of simd regions.
2859         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
2860         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
2861         taskreg_nesting_level while scanning host teams construct.
2862         (task_reduction_read): New function.
2863         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
2864         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
2865         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
2866         second argument create_tmp_var if it is NULL.  Don't ignore shared
2867         clauses in is_host_teams_ctx contexts.  Handle
2868         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
2869         clauses.
2870         (lower_reduction_clauses): Ignore reduction clauses with task
2871         modifier.  Remove second argument create_tmp_var if it is NULL.
2872         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
2873         (lower_send_clauses): Ignore reduction clauses with task modifier.
2874         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
2875         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
2876         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
2877         rely that it is the last stmt in body so far.  Ignore outer taskgroup
2878         contexts.
2879         (omp_task_reductions_find_first, omp_task_reduction_iterate,
2880         lower_omp_task_reductions): New functions.
2881         (lower_omp_sections): Handle reduction clauses with taskgroup
2882         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
2883         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
2884         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
2885         modifiers.
2886         (lower_omp_taskgroup): Handle taskgroup reductions.
2887         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
2888         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
2889         (lower_depend_clauses): If there are any
2890         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
2891         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
2892         seen, assume lowering is done already and return early.  Set kind
2893         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
2894         (lower_omp_taskreg): Handle reduction clauses with task modifier on
2895         parallel construct.  Handle reduction clause on taskloop construct.
2896         Handle taskwait with depend clauses.
2897         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
2898         for host teams constructs.
2899         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
2900         nontemporal and _reductemp_ clause entries.
2901         (omp_clause_code_name): Likewise.
2902         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
2903         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
2904         * tree-core.h (enum omp_clause_code): Add
2905         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
2906         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
2907         (struct tree_base): Add omp_atomic_memory_order field into union.
2908         Remove OMP_ATOMIC_SEQ_CST comment.
2909         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
2910         and OMP_CLAUSE_DEPEND_DEPOBJ.
2911         (struct tree_omp_clause): Add subcode.defaultmap_kind.
2912         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
2913         OpenMP constructs with body and clauses operands.
2914         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
2915         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
2916         (OMP_TASKGROUP_CLAUSES): Define.
2917         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
2918         OMP_CLAUSE__LOOPTEMP_.
2919         (OMP_ATOMIC_SEQ_CST): Remove.
2920         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
2921         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
2922         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
2923         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
2924         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
2925         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
2926         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
2927         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
2928         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
2929         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
2930         Define.
2931         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
2932         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2933         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
2934         (convert_local_omp_clauses): Likewise.  Remove useless test.
2935         * tree-parloops.c (create_call_for_reduction_1): Pass
2936         OMP_MEMORY_ORDER_RELAXED as new argument to
2937         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
2938         * tree-pretty-print.c (dump_omp_iterators): New function.
2939         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
2940         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
2941         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
2942         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
2943         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
2944         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
2945         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
2946         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
2947         (dump_omp_atomic_memory_order): New function.
2948         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
2949         taskwait with depend clauses.
2950         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
2951         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
2952         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
2953         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
2954         write_ts_omp_clause_tree_pointers): Likewise.
2956 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
2958         PR ipa/86395
2959         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
2960         "-missed", "-note", and "-all" sub-options.
2961         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
2962         API.
2963         (can_early_inline_edge_p): Likewise.
2964         (want_early_inline_function_p): Likewise.
2965         (want_inline_self_recursive_call_p): Likewise.
2966         (recursive_inlining): Likewise.
2967         (inline_small_functions): Likewise.
2968         (flatten_function): Likewise.
2969         (ipa_inline): Likewise.
2970         (inline_always_inline_functions): Likewise.
2971         (early_inline_small_functions): Likewise.
2972         (early_inliner): Likewise.
2973         * tree-inline.c (expand_call_inline): Likewise.
2975 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
2977         * pretty-print.c (pp_format): Handle %f.
2978         (selftest::test_pp_format): Add test of %f.
2979         * pretty-print.h (pp_double): New macro.
2981 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
2983         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
2984         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
2985         (dump_pretty_printer::decode_format): Implement "%C" for
2986         cgraph_node *.
2987         (selftest::test_capture_of_dump_calls): Rename "where" to
2988         "stmt_loc".  Convert test_decl to a function decl and set its
2989         location.  Add a symbol_table_test RAII instance and a
2990         cgraph_node, using it to test "%C" and dump_symtab_node.
2992 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
2994         PR middle-end/87916
2995         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
2997 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
2999         * cgraph.c: Include "selftest.h".
3000         (saved_symtab): New variable.
3001         (selftest::symbol_table_test::symbol_table_test): New ctor.
3002         (selftest::symbol_table_test::~symbol_table_test): New dtor.
3003         (selftest::test_symbol_table_test): New test.
3004         (selftest::cgraph_c_tests): New.
3005         * cgraph.h (saved_symtab): New decl.
3006         (selftest::symbol_table_test): New class.
3007         * selftest-run-tests.c (selftest::run_tests): Call
3008         selftest::cgraph_c_tests.
3009         * selftest.h (selftest::cgraph_c_tests): New decl.
3011 2018-11-08  Richard Biener  <rguenther@suse.de>
3013         * tree-data-ref.h (lambda_int): New typedef.
3014         (lambda_vector_gcd): Adjust.
3015         (lambda_vector_new): Likewise.
3016         (lambda_matrix_new): Likewise.
3017         * tree-data-ref.c  (print_lambda_vector): Adjust.
3019 2018-11-08  Richard Biener  <rguenther@suse.de>
3021         PR tree-optimization/87929
3022         * tree-complex.c (expand_complex_comparison): Clean EH.
3024 2018-11-08  Martin Liska  <mliska@suse.cz>
3026         * doc/extend.texi: Reword.
3027         * predict.c (expr_expected_value_1): Likewise.
3029 2018-11-08  Richard Biener  <rguenther@suse.de>
3031         PR tree-optimization/87913
3032         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
3033         of extreme values to ordered comparisons.
3035 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
3037         PR middle-end/42726
3038         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
3040 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
3042         * doc/invoke.texi: Remove leading dash from @opindex entries
3043         throughout the file.
3045 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
3047         PR driver/80828
3048         * doc/invoke.texi (Option Summary): Add -e and --entry.
3049         (Link Options): Likewise.
3051 2018-11-07  Nathan Sidwell  <nathan@acm.org>
3053         PR 87926
3054         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
3055         --disable-checking bootstrap.
3057 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
3059         * configure: Regenerated.
3061 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
3063         PR c/87691
3064         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
3065         to the mode of the widest field iff the widest field has mode class
3066         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
3067         reference.
3069 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
3071         * common.opt: Add -Wattribute-warning.
3072         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
3073         * expr.c (expand_expr_real_1): Add new attribute to warning_at
3074         call to allow user configure behavior of "warning" attribute.
3076 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
3078         * target.def: Put @: after every vs., e.g., and i.e. where it is
3079         followed by whitespace.
3080         * doc/extend.texi: Ditto.
3081         * doc/fragments.texi: Ditto.
3082         * doc/gimple.texi: Ditto.
3083         * doc/implement-c.texi: Ditto.
3084         * doc/install.texi: Ditto.
3085         * doc/invoke.texi: Ditto.
3086         * doc/md.texi: Ditto.
3087         * doc/plugins.texi: Ditto.
3088         * doc/rtl.texi: Ditto.
3089         * doc/sourcebuild.texi: Ditto.
3090         * doc/tm.texi.in: Ditto.
3091         * doc/ux.texi: Ditto.
3092         * doc/tm.texi: Regenerate.
3094 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3096         * config/arm/arm-cpus.in (ares): New entry.
3097         * config/arm/arm-tables.opt: Regenerate.
3098         * config/arm/arm-tune.md: Likewise.
3099         * doc/invoke.texi (ARM Options): Document ares.
3101 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3103         * config/aarch64/aarch64-cores.def (ares): Define.
3104         * config/aarch64/aarch64-tune.md: Regenerate.
3105         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
3107 2018-11-07  Jan Hubicka  <jh@suse.cz>
3109         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
3110         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
3111         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
3112         TYPE_DECL.
3114 2018-11-07  Richard Biener  <rguenther@suse.de>
3116         PR tree-optimization/87914
3117         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
3118         of nested cycles.
3119         (vectorizable_reduction): Handle shifts and rotates by dispatching
3120         to vectorizable_shift.
3121         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
3122         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
3123         def cases.
3124         (vectorizable_shift): Export and handle being called as
3125         vect_nested_cycle.
3126         (vect_analyze_stmt): Call vectorizable_shift after
3127         vectorizable_reduction.
3128         * tree-vectorizer.h (vectorizable_shift): Declare.
3130 2018-11-07  Jan Hubicka  <jh@suse.cz>
3132         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
3133         than const decls in TREE_VALUE of enum.
3134         (dump_type_inheritance_graph): Improve duplicate dumping.
3135         (free_enum_values): New.
3136         (build_type_inheritance_graph): Use it.
3137         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
3138         which are not main variants or not ODR types.
3139         (verify_type_variant): Expect variants to have no TYPE_VALUES.
3141 2018-11-07  Richard Biener  <rguenther@suse.de>
3143         * ipa-inline.c (want_inline_small_function_p): Compute
3144         big_speedup_p lazily and last.
3146 2018-11-07  Jan Hubicka  <jh@suse.cz>
3148         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
3149         building incomplete variant of complete type.
3150         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
3151         variant of complete type.
3153 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
3155         * config/mips/mips.c: Fix typo in documentation of
3156         mips_loongson_ext2_prefetch_cookie.
3157         (mips_option_override): fix brain twister logical.
3158         * config/mips/mips.h: Fix typo in documentation of
3159         ISA_HAS_CTZ_CTO and define pattern.
3160         * config/mips/mips.md (prefetch): Hoist EXT2 above
3161         the 2EF/EXT block.
3162         (prefetch_indexed): Hoist EXT2 above the EXT block.
3164 2018-11-07  Jan Hubicka  <jh@suse.cz>
3166         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
3167         return and parameter types of function and method types.
3168         (free_lang_data_in_cgraph): Update.
3170 2018-11-07  Martin Liska  <mliska@suse.cz>
3172         PR rtl-optimization/87868
3173         * postreload-gcse.c (eliminate_partially_redundant_load): Set
3174         threshold to max_count if we would overflow.
3175         * profile-count.h: Make max_count a public constant.
3177 2018-11-07  Martin Liska  <mliska@suse.cz>
3179         * mem-stats.h: Fix GNU coding style.
3181 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
3183         * config/mips/gs264e.md: New.
3184         * config/mips/mips-cpus.def: Define gs264e.
3185         * config/mips/mips-tables.opt: Regenerate.
3186         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
3187         gs264e.
3188         (mips_issue_rate): Add support for gs264e.
3189         (mips_multipass_dfa_lookahead): Likewise.
3190         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
3191         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
3192         (MIPS_ASE_MSA_SPEC): New.
3193         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
3194         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
3195         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
3196         * config/mips/mips.md: Include gs264e.md.
3197         (processor): Add gs264e.
3198         * config/mips/mips.opt (MSA): Use Mask instead of Var.
3199         * doc/invoke.texi: Add gs264e to supported architectures.
3201 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
3203         * config/mips/gs464e.md: New.
3204         * config/mips/mips-cpus.def: Define gs464e.
3205         * config/mips/mips-tables.opt: Regenerate.
3206         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
3207         gs464e.
3208         (mips_issue_rate): Add support for gs464e.
3209         (mips_multipass_dfa_lookahead): Likewise.
3210         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
3211         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
3212         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
3213         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
3214         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
3215         * config/mips/mips.md: Include gs464e.md.
3216         (processor): Add gs464e.
3217         * doc/invoke.texi: Add gs464e to supported architectures.
3219 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
3221         * config/mips/loongson3a.md: Rename to ...
3222         * config/mips/gs464.md: ... here.
3223         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
3224         as an alias of gs464 processor.
3225         * config/mips/mips-tables.opt: Regenerate.
3226         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
3227         instead of PROCESSOR_LOONGSON_3A.
3228         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
3229         TUNE_LOONGSON_3A.
3230         (mips_option_override): Enable MMI and EXT for gs464.
3231         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
3232         Rename TUNE_LOONGSON_3A to TUNE_GS464.
3233         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
3234         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
3235         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
3236         TARGET_LOONGSON_3A.
3237         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
3238         (processor): Add gs464;
3239         * doc/invoke.texi: Add gs464 to supported architectures.
3241 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
3243         * config/mips/mips-protos.h
3244         (mips_loongson_ext2_prefetch_cookie): New prototype.
3245         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
3246         (mips_option_override): Enable TARGET_LOONGSON_EXT when
3247         TARGET_LOONGSON_EXT2 is true.
3248         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
3249         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
3250         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
3251         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
3252         TARGET_LOONGSON_EXT2.
3253         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
3254         (define_insn "ctz<mode>2"): New insn pattern.
3255         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
3256         (define_insn "prefetch_indexed_<mode>"): Include
3257         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
3258         * config/mips/mips.opt (-mloongson-ext2): Add option.
3259         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
3261 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
3263         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
3264         __mips_loongson_ext.
3265         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
3266         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
3267         -mloongson-ext.
3268         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
3269         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
3270         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
3271         instead of TARGET_LOONGSON_3A.
3272         * config/mips/mips.opt (-mloongson-ext): Add option.
3273         * gcc/doc/invoke.texi (-mloongson-ext): Document.
3275 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
3277         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
3278         * config/mips/loongson.md: Move to ...
3279         * config/mips/loongson-mmi.md: here; Adjustment.
3280         * config/mips/loongson.h: Move to ...
3281         State as deprecated. Include loongson-mmiintrin.h for back
3282         compatibility and warning.
3283         * config/mips/loongson-mmiintrin.h: ... here.
3284         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
3285         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
3286         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
3287         (mips_option_override): Make sure MMI use hard float;
3288         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
3289         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
3290         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
3291         TARGET_LOONGSON_VECTORS.
3292         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
3293         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
3294         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
3295         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
3296         -mloongson-mmi.
3297         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
3298         TARGET_LOONGSON_VECTORS.
3299         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
3300         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
3301         (Loongson MMI patterns): Include loongson-mmi.md instead of
3302         loongson.md.
3303         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
3304         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
3306 2018-11-07  Richard Biener  <rguenther@suse.de>
3308         PR lto/87906
3309         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
3310         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
3312 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
3314         PR rtl-optimization/87874
3315         * lra.c (lra_substitute_pseudo): Do not create a subreg for
3316         const wide ints.
3318 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
3320         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
3321         if not in indexed or indirect form.
3322         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
3323         (bswap<mode>2_store): Ditto.
3325 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
3327         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
3328         the UNSPEC.
3330 2018-11-06  Richard Biener  <rguenther@suse.de>
3332         PR tree-optimization/86850
3333         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
3334         instead of src.m_vec.
3336 2018-11-06  Jan Hubicka  <jh@suse.cz>
3338         * tree.c (fld_simplified_type_name): Break out form ...
3339         (free_lang_data_in_type): ... here.
3340         (fld_type_variant_equal_p): Use it.
3342 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3344         * config/default-d.c: Include memmodel.h.
3346         * config/sol2-d.c: New file.
3347         * config/t-sol2 (sol2-d.o): New rule.
3348         * config.gcc <*-*-solaris2*>: Set d_target_objs,
3349         target_has_targetdm.
3351 2018-11-06  Jan Hubicka  <jh@suse.cz>
3353         * tree.c (fld_type_variant): Also copy alignment; be sure that
3354         new variant is equal.
3356 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
3358         PR target/87762
3359         * config/s390/s390.md: Add relative_long attribute.
3361 2018-11-06  Jan Hubicka  <jh@suse.cz>
3363         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
3364         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
3365         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
3366         stream TYPE_NEEDS_CONSTRUCTING.
3367         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
3368         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
3370 2018-11-06  Richard Biener  <rguenther@suse.de>
3372         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
3373         dump-scope ...
3374         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
3376 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
3378         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
3379         REG_CLASS_CONTENTS[GEN_REGS].
3380         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
3382 2018-11-06  Jan Hubicka  <jh@suse.cz>
3384          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
3386 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3388         PR sanitizer/80953
3389         * config/sol2.h (ASAN_CC1_SPEC): Define.
3390         (LD_WHOLE_ARCHIVE_OPTION): Define.
3391         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
3392         (ASAN_REJECT_SPEC): Provide default.
3393         (LIBASAN_EARLY_SPEC): Define.
3394         (LIBTSAN_EARLY_SPEC): Define.
3395         (LIBLSAN_EARLY_SPEC): Define.
3396         * config/i386/sol2.h (CC1_SPEC): Redefine.
3397         (ASAN_REJECT_SPEC): Define.
3399         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
3400         (TARGET_ASAN_SHADOW_OFFSET): Define.
3401         (sparc_asan_shadow_offset): New function.
3402         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
3403         (ASAN_REJECT_SPEC): Define.
3405 2018-11-06  Jan Hubicka  <jh@suse.cz>
3407         * tree.c (fld_type_variant): Copy canonical type.
3408         (fld_incomplete_type_of): Check that canonical types looks sane;
3409         copy canonical type.
3410         (verify_type): Accept when incomplete type has complete canonical type.
3412 2018-11-06  Jan Hubicka  <jh@suse.cz>
3414         * tree.c (free_lang_data): Reset overwite_assembler_name,
3415         print_xnode, print_decl, print_type and print_identifier of
3416         langhooks.
3418 2018-11-06  Richard Biener  <rguenther@suse.de>
3420         PR tree-optimization/87889
3421         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
3422         Do nothing if old and new arg are the same
3424 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
3426         PR target/87723
3427         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
3428         attributes for operands 3 and 4.
3430 2018-11-06  Richard Biener  <rguenther@suse.de>
3432         PR middle-end/18041
3433         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
3434         matching bitfield insertion.
3436 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
3438         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
3439         (reg_next_debug_use): New.
3440         (attempt_change): Propagate adjusted expression into affected
3441         debug insns.
3442         (merge_in_block): Track uses in debug insns.
3443         (pass_inc_dec::execute): Allocate and release
3444         reg_next_debug_use.
3446 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
3448         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
3449         (_mm512_fixupimm_round_pd): Update parameters and builtin.
3450         (_mm512_maskz_fixupimm_round_pd): Ditto.
3451         (_mm512_fixupimm_round_ps): Ditto.
3452         (_mm512_maskz_fixupimm_round_ps): Ditto.
3453         (_mm_fixupimm_round_sd): Ditto.
3454         (_mm_maskz_fixupimm_round_sd): Ditto.
3455         (_mm_fixupimm_round_ss): Ditto.
3456         (_mm_maskz_fixupimm_round_ss): Ditto.
3457         (_mm512_fixupimm_pd): Ditto.
3458         (_mm512_maskz_fixupimm_pd): Ditto.
3459         (_mm512_fixupimm_ps): Ditto.
3460         (_mm512_maskz_fixupimm_ps): Ditto.
3461         (_mm_fixupimm_sd): Ditto.
3462         (_mm_maskz_fixupimm_sd): Ditto.
3463         (_mm_fixupimm_ss): Ditto.
3464         (_mm_maskz_fixupimm_ss): Ditto.
3465         (_mm512_mask_fixupimm_round_pd): Update builtin.
3466         (_mm512_mask_fixupimm_round_ps): Ditto.
3467         (_mm_mask_fixupimm_round_sd): Ditto.
3468         (_mm_mask_fixupimm_round_ss): Ditto.
3469         (_mm512_mask_fixupimm_pd): Ditto.
3470         (_mm512_mask_fixupimm_ps): Ditto.
3471         (_mm_mask_fixupimm_sd): Ditto.
3472         (_mm_mask_fixupimm_ss): Ditto.
3473         * config/i386/avx512vlintrin.h:
3474         (_mm256_fixupimm_pd): Update parameters and builtin.
3475         (_mm256_maskz_fixupimm_pd): Ditto.
3476         (_mm256_fixupimm_ps): Ditto.
3477         (_mm256_maskz_fixupimm_ps): Ditto.
3478         (_mm_fixupimm_pd): Ditto.
3479         (_mm_maskz_fixupimm_pd): Ditto.
3480         (_mm_fixupimm_ps): Ditto.
3481         (_mm_maskz_fixupimm_ps): Ditto.
3482         (_mm256_mask_fixupimm_pd): Update builtin.
3483         (_mm256_mask_fixupimm_ps): Ditto.
3484         (_mm_mask_fixupimm_pd): Ditto.
3485         (_mm_mask_fixupimm_ps): Ditto.
3486         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
3487         * config/i386/i386-builtin.def: Update builtin definitions.
3488         * config/i386/i386.c: Handle new builtin types and remove useless ones.
3489         * config/i386/sse.md: Update VFIXUPIMM* patterns.
3490         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
3491         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
3492         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
3493         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
3494         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
3495         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
3496         * config/i386/subst.md:
3497         (round_saeonly_sd_mask_operand4): Add new subst_attr.
3498         (round_saeonly_sd_mask_op4): Ditto.
3499         (round_saeonly_expand_operand5): Ditto.
3500         (round_saeonly_expand): Update.
3502 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
3504         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
3506 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
3508         PR rtl-optimization/87871
3509         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
3511 2018-11-05  Paul Koning  <ni1d@arrl.net>
3513         * doc/sourcebuild.texi (target attributes): Document new "inf"
3514         effective target keyword.
3516 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
3518         * config/s390/s390.c (s390_register_move_cost): Increase costs for
3519         moves involving the CC reg.
3521 2018-11-05  Richard Biener  <rguenther@suse.de>
3523         PR tree-optimization/87873
3524         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
3525         argument.
3526         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
3527         * tree-vect-loop.c (vect_transform_loop): When splitting the
3528         loop exit also create forwarder PHIs for constants.
3529         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
3530         Handle constant to_arg, add extra checking we match up the correct
3531         PHIs.
3533 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
3535         * config/s390/s390.md: QImode and HImode for load on condition.
3537 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
3539         * config/s390/predicates.md: Fix typo.
3540         * config/s390/s390.md: Allow immediates for load on condition.
3542 2018-11-05  Martin Liska  <mliska@suse.cz>
3544         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
3545         * bitmap.h (struct bitmap_usage): Likewise.
3546         * ggc-common.c (SCALE): Remove.
3547         (LABEL): Likewise.
3548         (struct ggc_usage): Use SIZE_AMOUNT. And update
3549         compare method.
3550         * ggc-page.c (SCALE): Remove.
3551         (STAT_LABEL): Remove.
3552         (ggc_print_statistics): Use SIZE_AMOUNT.
3553         * gimple.h (SCALE): Remove.
3554         (LABEL): Likewise.
3555         * input.c (ONE_K): Remove.
3556         (ONE_M): Likewise.
3557         (SCALE): Likewise.
3558         (STAT_LABEL): Likewise.
3559         (FORMAT_AMOUNT): Likewise.
3560         (dump_line_table_statistics): Use SIZE_AMOUNT.
3561         * mem-stats.h (struct mem_usage): Likewise.
3562         * rtl.c (dump_rtx_statistics): Likewise.
3563         (rtx_alloc_counts): Change type to size_t.
3564         (rtx_alloc_sizes): Likewise.
3565         (rtx_count_cmp): New.
3566         (dump_rtx_statistics): Sort first based on counts.
3567         * tree.c (tree_nodes_cmp): New.
3568         (tree_codes_cmp): New.
3569         (dump_tree_statistics): Sort first based on counts.
3570         * system.h (ONE_K): New.
3571         (ONE_M): Likewise.
3572         (SIZE_SCALE): Likewise.
3573         (SIZE_LABEL): Likewise.
3574         (SIZE_AMOUNT): Likewise.
3575         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
3576         * tree-dfa.c (dump_dfa_stats): Likewise.
3577         * tree-phinodes.c (phinodes_print_statistics): Likewise.
3578         * tree-ssanames.c (ssanames_print_statistics): Likewise.
3579         * tree.c (dump_tree_statistics): Likewise.
3580         * vec.c (struct vec_usage): Likewise.
3581         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
3582         trigger a -Werror=format-overflow with
3583         --enable-gather-detailed-stats.
3585 2018-11-05  Martin Liska  <mliska@suse.cz>
3587         * mem-stats.h (mem_alloc_description::release_instance_overhead):
3588         Return T *.
3589         * vec.c (struct vec_usage): Register m_element_size.
3590         (vec_prefix::register_overhead): New arguments: elements and
3591         element_size.
3592         (vec_prefix::release_overhead): Subtract elements.
3593         * vec.h (struct vec_prefix): Change signature.
3594         (va_heap::reserve): Pass proper arguments.
3595         (va_heap::release): Likewise.
3597 2018-11-05  Martin Liska  <mliska@suse.cz>
3599         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
3600         style.
3601         * vec.c: Likewise.
3603 2018-11-05  Richard Biener  <rguenther@suse.de>
3605         * tree-scalar-evolution.h (final_value_replacement_loop): Update
3606         prototype.
3607         * tree-scalar-evolution.c (final_value_replacement_loop): Return
3608         whether anything was done.
3609         (scev_const_prop): Remove constant propagation part, fold
3610         remains into ...
3611         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
3612         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
3613         conditionally.
3615 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
3617         PR tree-optimization/87859
3618         * gimple-ssa-store-merging.c (struct merged_store_group): Add
3619         only_constants and first_nonmergeable_order members.
3620         (merged_store_group::merged_store_group): Initialize them.
3621         (merged_store_group::do_merge): Clear only_constants member if
3622         adding something other than INTEGER_CST store.
3623         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
3624         stores with order >= first_nonmergeable_order.  Use
3625         merged_store->only_constants instead of always recomputing it.
3626         Set merged_store->first_nonmergeable_order if we've skipped any
3627         stores.  Attempt to merge overlapping INTEGER_CST stores that
3628         we would otherwise skip.
3630         PR sanitizer/87837
3631         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
3633 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
3635         PR target/87853
3636         * config/i386/emmintrin.h (__v16qs): New to cope with option
3637         -funsigned-char.
3638         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
3639         (_mm_cmplt_epi8): Likewise.
3640         (_mm_cmpgt_epi8): Likewise.
3642 2018-11-05  Richard Biener  <rguenther@suse.de>
3644         PR rtl-optimization/87852
3645         * fwprop.c (use_killed_between): Only consider single-defs of the
3646         use whose definition statement dominates the use.
3648 2018-11-05  Martin Liska  <mliska@suse.cz>
3650         PR web/87829
3651         * doc/invoke.texi: Remove options that are
3652         not disabled with -Os.
3654 2018-11-05  Martin Liska  <mliska@suse.cz>
3656         PR c/87811
3657         * doc/extend.texi: Update constrain about the last argument
3658         of __builtin_expect_with_probability.
3660 2018-11-05  Martin Liska  <mliska@suse.cz>
3662         PR c/87811
3663         * predict.c (expr_expected_value_1): Verify
3664         that last argument is a real constants and emit
3665         error.
3667 2018-11-05  Martin Liska  <mliska@suse.cz>
3669         PR gcov-profile/77698
3670         * ipa-profile.c (ipa_profile): Adjust hotness threshold
3671         only in LTO mode.
3673 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3675         PR tree-optimization/86572
3676         * builtins.c (c_strlen): Handle negative offsets in a safe way.
3678 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3680         PR tree-optimization/87672
3681         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
3682         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
3684 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
3686         PR middle-end/58372
3687         * cfgexpand.c (pass_expand::execute): Move the call to
3688         finish_eh_generation in front of the call to expand_stack_alignment.
3690 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
3692         * common/config/i386/i386-common.c (processor_alias_table): Add
3693         znver2 entry.
3694         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
3695         (case ${target}): Add znver2.
3696         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
3697         -march=native recognize znver2 processors.
3698         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
3699         * config/i386/i386.c (m_znver2): New definition.
3700         (m_ZNVER): New definition.
3701         (m_AMD_MULTIPLE): Includes m_znver2.
3702         (processor_cost_table): Add znver2 entry.
3703         (processor_target_table): Add znver2 entry.
3704         (get_builtin_code_for_version): Set priority for
3705         PROCESSOR_ZNVER2.
3706         (processor_model): Add M_AMDFAM17H_ZNVER2.
3707         (arch_names_table): Ditto.
3708         (ix86_reassociation_width): Include znver2.
3709         * config/i386/i386.h (TARGET_znver2): New definition.
3710         (struct ix86_size_cost): Add TARGET_ZNVER2.
3711         (enum processor_type): Add PROCESSOR_ZNVER2.
3712         * config/i386/i386.md (define_attr "cpu"): Add znver2.
3713         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
3714         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
3715         (ix86_adjust_cost): Add znver2.
3716         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
3717         * gcc/doc/extend.texi: Add details about znver2.
3718         * gcc/doc/invoke.texi: Add details about znver2.
3720 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
3722         PR target/87079
3724         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
3725         pattern.
3727 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
3729         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
3730         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
3732 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
3734         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
3735         shrink-wrapping when -mtrack-speculation.
3737 2018-11-02  Richard Biener  <rguenther@suse.de>
3739         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
3740         (pop_cost_one_pair): Do not free pair.
3741         (pop_best_coalesce): Likewise.
3742         (create_coalesce_list): Initialize obstack.
3743         (delete_coalesce_list): Free obstack.
3744         (find_coalesce_pair): Obstack-allocate coalesce pairs.
3745         (add_cost_one_coalesce): Likewise.
3746         (struct live_track): Remove bitmap pointer indirections.
3747         (new_live_track): Adjust.
3748         (delete_live_track): Likewise.
3749         (live_track_remove_partition): Likewise.
3750         (live_track_add_partition): Likewise.
3751         (live_track_live_p): Likewise.
3752         (live_track_process_def): Likewise.
3753         (live_track_clear_base_vars): Likewise.
3755 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3757         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
3758         * configure: Regenerate.
3759         * config.in: Regenerate.
3760         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
3761         HAVE_AS_REGISTER_PSEUDO_OP guard.
3762         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
3763         !HAVE_AS_REGISTER_PSEUDO_OP support.
3764         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
3765         guard.
3767 2018-11-02  Richard Biener  <rguenther@suse.de>
3769         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
3770         with is_type_die.
3772 2018-11-02  Richard Biener  <rguenther@suse.de>
3774         PR tree-optimization/87776
3775         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
3776         executable when iterating but running into rpo-vn-max-loop-depth
3777         and not eliding the iteration.
3779 2018-11-30  Jan Hubicka  <jh@suse.cz>
3781         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
3782         TYPE_DECL.
3784 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
3786         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
3787         prototype.
3788         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
3789         Combine with rs6000_address_for_fpconvert.
3790         (rs6000_address_for_fpconvert): Combine with
3791         rs6000_force_indexed_or_indirect_mem.
3792         (rs6000_expand_vector_init): Change function call from
3793         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
3794         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
3795         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
3796         (floatsi<mode>2_lfiwax_mem): Ditto.
3797         (floatunssi<mode>2_lfiwzx): Ditto.
3798         (floatunssi<mode>2_lfiwzx_mem): Ditto.
3799         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
3800         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
3801         (fix_trunc<mode>si2_stfiwx): Ditto.
3802         (fixuns_trunc<mode>si2_stfiwx): Ditto.
3803         (float_<mode>si2_hw): Ditto.
3804         (floatuns_<mode>si2_hw): Ditto.
3805         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
3806         (vsx_splat_<mode>): Ditto.
3808 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
3810         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
3811         at end of assembler input text.
3812         * configure: Regenerate.
3814 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
3816         PR tree-optimization/87826
3817         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
3818         negative or larger or equal to type's precision.
3820 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
3822         * opts.c (default_options_table): Do not enable
3823         OPT_fdelayed_branch at -Og.
3824         * doc/invoke.texi (-fdelayed-branch): Document it.
3826 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
3828         * optabs-libfuncs.c (build_libfunc_function_visibility):
3829         New, split out from...
3830         (build_libfunc_function): ... here.
3831         (init_one_libfunc_visibility): New, split out from ...
3832         (init_one_libfunc): ... here.
3834         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
3835         scratch register need not be early-clobber.  Document the reason
3836         why we cannot use ST<OP>.
3838 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
3840         PR bootstrap/82856
3841         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
3842         line for second argument of AC_DEFINE_UNQUOTED.
3843         * doc/install.texi (Tools/packages necessary for modifying GCC):
3844         Update to autoconf 2.69 and automake 1.15.1.
3845         * aclocal.m4, config.in, configure: Regenerate.
3847 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
3849         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
3850         initialization.
3852 2018-10-31  Martin Liska  <mliska@suse.cz>
3854         PR driver/83193
3855         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
3856         Add new argument invalid_extension.
3857         (aarch64_get_all_extension_candidates): New function.
3858         (aarch64_rewrite_selected_cpu): Add NULL to function call.
3859         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
3860         new argument.
3861         (aarch64_get_all_extension_candidates): New function.
3862         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
3863         argument invalid_extension.
3864         (aarch64_parse_cpu): Likewise.
3865         (aarch64_print_hint_for_extensions): New function.
3866         (aarch64_validate_mcpu): Provide hint about invalid extension.
3867         (aarch64_validate_march): Likewise.
3868         (aarch64_handle_attr_arch): Pass new argument.
3869         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
3870         (aarch64_handle_attr_isa_flags): Likewise.
3872 2018-10-31  Richard Biener  <rguenther@suse.de>
3874         PR middle-end/70359
3875         PR middle-end/86270
3876         * tree-outof-ssa.c (insert_backedge_copies): Restrict
3877         copy generation to useful cases.  Place the copy before
3878         the definition of the backedge value when possible.
3880 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
3882         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
3883         * config/arc/arc.c (arc_active_insn): New function.
3884         (check_store_cacheline_hazard): Likewise.
3885         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
3886         (arc_override_options): Disable delay slot scheduler for older
3887         A7.
3888         (arc_store_addr_hazard_p): New implementation, old one renamed to
3889         ...
3890         (arc_store_addr_hazard_internal_p): Renamed.
3891         (arc_reorg): Don't combine into brcc instructions which are part
3892         of hardware hazard solution.
3893         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
3894         (tune_arc700): Likewise.
3895         * config/arc/arc.opt (arc7xx): New tune value.
3896         * config/arc/arc700.md: Improve A7 scheduler.
3898 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
3900         * config/arc/arc.c (arc_override_options): Remove
3901         TARGET_COMPACT_CASESI.
3902         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
3903         (CASE_VECTOR_MODE): Likewise.
3904         (CASE_VECTOR_PC_RELATIVE): Likewise.
3905         (CASE_VECTOR_SHORTEN_MODE): Likewise.
3906         (CASE_VECTOR_SHORTEN_MODE1): Delete.
3907         (ADDR_VEC_ALIGN): Update.
3908         (ASM_OUTPUT_CASE_LABEL): Undefine.
3909         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
3910         (TARGET_BI_BIH): Define.
3911         (DEFAULT_BRANCH_INDEX): Likewise.
3912         * config/arc/arc.md (casesi): Rework to accept BI/BIH
3913         instructions, remove compact_casesi use case.
3914         (casesi_compact_jump): Remove.
3915         (casesi_dispatch): New pattern.
3916         * config/arc/arc.opt: Add mbranch-index option. Deprecate
3917         compact_casesi option.
3918         * doc/invoke.texi: Document mbranch-index option.
3920 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
3922         * config/arc/arc.c (arc_get_tp): Remove function.
3923         (arc_emit_call_tls_get_addr): Likewise.
3924         (arc_call_tls_get_addr): New function.
3925         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
3926         * config/arc/arc.md (tls_load_tp_soft): Remove.
3927         (tls_gd_get_addr): Likewise.
3929 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
3931         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
3932         (mulsi3_600_lib): Remove pattern.
3933         (umulsi3_highpart_600_lib_le): Likewise.
3934         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
3935         (umulsidi3): Remove call to umulsidi3_600_lib.
3936         (umulsidi3_600_lib): Remove pattern.
3937         (peephole2): Remove peephole using the above deprecated patterns.
3939 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
3941         PR target/87374
3942         * config/arm/arm.c (arm_option_check_internal): Disable the combined
3943         use of -mslow-flash-data and -mword-relocations.
3944         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
3945         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
3946         flag_pic.
3947         * doc/invoke.texi (-mword-relocations): Mention conflict with
3948         -mslow-flash-data.
3949         (-mslow-flash-data): Reciprocally.
3951 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
3953         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
3954         16-byte modes held in GP registers to use an even regno.
3956         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
3957         (aarch64_atomic_ldop_supported_p): Remove.
3958         (aarch64_gen_atomic_ldop): Remove.
3959         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
3960         Fully expand LSE operations here.
3961         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
3962         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
3963         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
3964         and use ATOMIC_LDOP instead; use register_operand for the input;
3965         drop the split and emit insns directly.
3966         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
3967         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
3968         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
3970         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
3971         (aarch64_gen_atomic_ldop): Don't call it.
3972         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
3973         Use aarch64_reg_or_zero.
3974         (aarch64_atomic_exchange<ALLI>): Likewise.
3975         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
3976         operand 0; use aarch64_reg_or_zero for input; merge ...
3977         (@aarch64_atomic_swp<ALLI>): ... this and remove.
3979         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
3980         (aarch64_split_compare_and_swap): Use it.
3981         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
3982         test oldval against the proper predicate.
3983         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
3984         Use nonmemory_operand for expected.
3985         (cas_short_expected_pred): New.
3986         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
3987         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
3988         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
3989         (aarch64_plushi_operand): New.
3991         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
3992         Force oldval into the rval register for TARGET_LSE; emit the compare
3993         during initial expansion so that it may be deleted if unused.
3994         (aarch64_gen_atomic_cas): Remove.
3995         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
3996         Change =&r to +r for operand 0; use match_dup for operand 2;
3997         remove is_weak and mod_f operands as unused.  Drop the split
3998         and merge with...
3999         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
4000         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
4001         (@aarch64_atomic_cas<GPI>): Similarly.
4003 2018-10-31  Richard Biener  <rguenther@suse.de>
4005         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
4006         using ABSU_EXPR.
4008 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4010         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
4011         * config/aarch64/aarch64.md: Include saphira.md
4012         * config/aarch64/saphira.md: New file for pipeline description.
4014 2018-10-30  Martin Sebor  <msebor@redhat.com>
4016         PR middle-end/87041
4017         * gimple-ssa-sprintf.c (format_directive): Use %G to include
4018         inlining context.
4019         (sprintf_dom_walker::compute_format_length):
4020         Avoid setting POSUNDER4K here.
4021         (get_destination_size): Handle null argument values.
4022         (get_user_idx_format): New function.
4023         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
4024         functions, including user-defined with attribute format printf.
4025         Use %G to include inlining context.
4026         Set POSUNDER4K here.
4028 2018-10-30  Jan Hubicka  <jh@suse.cz>
4030         * params.def (lto-partitions): Bump from 32 to 128.
4032 2018-10-30  Jan Hubicka  <jh@suse.cz>
4034         * tree.c
4035         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
4036         head in file.
4037         (free_lang_data_in_type): Forward declare.
4038         (fld_type_variant_equal_p): New function.
4039         (fld_type_variant): New function
4040         (fld_incomplete_types): New hash.
4041         (fld_incomplete_type_of): New function
4042         (fld_simplfied-type): New function.
4043         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
4044         (free_lang_data): Allocate and free fld_incomplete_type; update call
4045         of free_lang_data_in_decl.
4047 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
4049         * gcov.c (output_lines): Remove duplicate line.
4051 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
4053         * config/rs6000/rs6000.md (bswapdi2): Force address into register
4054         if not in indexed or indirect form.
4055         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
4056         (bswapdi2_store): Ditto.
4057         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
4058         helper function.
4059         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
4060         Prototype for helper function.
4062 2018-10-30  Martin Sebor  <msebor@redhat.com>
4064         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
4065         (target, pragma GCC optimize, pragma GCC target): Ditto.
4067 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
4069         * doc/extend.texi: Fix prototype and description of
4070         __builtin_expect_with_probability.
4072 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
4074         * cgraph.h (clone_function_name_1): Replaced by new
4075           clone_function_name_numbered that takes name as string; for
4076           privatize_symbol_name_1 use only.
4077           (clone_function_name): Renamed to
4078           clone_function_name_numbered to be explicit about numbering.
4079           (clone_function_name): New two-argument function that does
4080           not number its output.
4081           (clone_function_name): New three-argument function that
4082           takes a number to append to its output.
4083         * cgraphclones.c (duplicate_thunk_for_node):
4084           (clone_function_name_1): Renamed.
4085           (clone_function_name_numbered): Two new functions.
4086           (clone_function_name): Improved documentation.
4087           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
4088         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
4089         * final.c (final_scan_insn_1): Use the new clone_function_name
4090           without numbering.
4091         * multiple_target.c (create_dispatcher_calls): Ditto.
4092           (create_target_clone): Ditto.
4093         * omp-expand.c (grid_expand_target_grid_body): Ditto.
4094         * omp-low.c (create_omp_child_function_name): Ditto.
4095         * omp-simd-clone.c (simd_clone_create): Ditto.
4096         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
4097           new clone_function_name without numbering.
4099 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
4101         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
4102         Assert that the allocation size is not zero.
4104 2018-10-30  Richard Biener  <rguenther@suse.de>
4106         PR tree-optimization/87800
4107         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
4108         non-induction or reduction PHIs.
4110 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
4112         * config/aarch64/falkor-tag-collision-avoidance.c
4113         (execute_tag_collision_avoidance): Call df_note_add_problem.
4115 2018-10-30  Martin Liska  <mliska@suse.cz>
4117         * doc/extend.texi: Fix typo in documentation
4118         of __builtin_expect_with_probability.
4120 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
4122         PR c++/87721
4123         * input.c (get_substring_ranges_for_loc): Detect if
4124         linemap_resolve_location gives us a NULL map, and reject
4125         this case.
4127 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
4129         * config.gcc (xstormy16-*-elf): Set tm_d_file.
4131 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
4132             Martin Sebor  <msebor@redhat.com>
4133             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
4135         * doc/ux.texi (Quoting): New subsection, adapted from material at
4136         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
4137         MartinSebor and ManuelLopezIbanez.
4138         (Fix-it hints): Note that fix-it hints shouldn't be marked for
4139         translation.
4141 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
4143         PR middle-end/87469
4144         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
4145         max value.
4147 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
4149         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
4151 2018-10-29  Paul Koning  <ni1d@arrl.net>
4153         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
4155 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
4157         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
4158         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
4159         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
4160         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
4161         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
4162         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
4163         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
4164         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
4165         Change 'vector' to '__vector'.
4166         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
4167         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
4168         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
4169         _mm_avg_pu16): Likewise.  And, whitespace corrections.
4171 2018-10-29  Richard Biener  <rguenther@suse.de>
4173         PR tree-optimization/87785
4174         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
4175         internal defs.
4177 2018-10-29  Olivier Hainque  <hainque@adacore.com>
4179         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
4181 2018-10-29  Olivier Hainque  <hainque@adacore.com>
4183         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
4184         ports configuration macro, defaults to "gnu".
4185         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
4186         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
4188 2018-10-29  Olivier Hainque  <hainque@adacore.com>
4190         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
4192 2018-10-29  Richard Biener  <rguenther@suse.de>
4194         PR tree-optimization/87790
4195         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
4196         (vect_make_slp_decision): Adjust.
4197         (vect_slp_analyze_bb_1): Likewise.
4198         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
4199         edges.
4201 2018-10-29  Richard Biener  <rguenther@suse.de>
4203         PR tree-optimization/87785
4204         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
4205         and processing.
4206         (vect_build_slp_tree): Likewise.
4207         (vect_gather_slp_loads): New function.
4208         (vect_analyze_slp_instance): Gather loads separately from the
4209         SLP tree build.
4211 2018-10-29  Martin Liska  <mliska@suse.cz>
4213         * Makefile.in: Make dependency to json.o.
4214         * doc/gcov.texi: Document new JSON format, remove
4215         old intermediate format documentation.
4216         * gcov.c (struct function_info): Come up with m_name and
4217         m_demangled_name.
4218         (function_info::function_info): Initialize it.
4219         (function_info::~function_info): Release it.
4220         (main): Rename flag_intermediate_format to flag_json_format.
4221         (print_usage): Describe --json-format.
4222         (process_args): Set flag_json_format.
4223         (output_intermediate_line): Remove.
4224         (output_intermediate_json_line): Likewise.
4225         (get_gcov_intermediate_filename): Return new extension
4226         ".gcov.json.gz".
4227         (output_intermediate_file): Implement JSON emission.
4228         (output_json_intermediate_file): Implement JSON emission.
4229         (generate_results): Use ::get_name for function name.
4230         Handle JSON output file.
4231         (read_graph_file): Use ::get_name instead of cplus_demangle.
4232         (read_count_file): Likewise.
4233         (solve_flow_graph): Likewise.
4234         (add_line_counts): Likewise.
4235         (accumulate_line_counts): Use new flag_json_format.
4236         (output_function_details): Use ::get_name instead of cplus_demangle.
4237         (output_lines): Likewise.
4238         * json.cc (test_writing_literals): Add new tests.
4239         * json.h (class literal): Add new boolean constructor.
4241 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
4243         PR rtl-optimization/87701
4244         PR rtl-optimization/87780
4245         * combine.c (make_more_copies): Rewrite.
4247 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
4249         * doc/generic.texi (ABSU_EXPR): Document.
4250         * match.pd (absu(x)*absu(x) -> x*x): Handle.
4251         (absu(absu(X)) -> absu(X)): Likewise.
4252         (absu(-X) -> absu(X)): Likewise.
4253         (absu(X)  where X is nonnegative -> X): Likewise.
4255 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
4257         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
4258         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
4259         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
4260         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
4261         (s-tm-texi): Also check timestamp on d-target.def.
4262         (generated_files): Add TM_D_H and d-target-hooks-def.h.
4263         (build/genhooks.o): Also depend on D_TARGET_DEF.
4264         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
4265         variables.
4266         * config/aarch64/aarch64-d.c: New file.
4267         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
4268         Define.
4269         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
4270         prototype.
4271         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
4272         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
4273         * config/arm/arm-d.c: New file.
4274         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
4275         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
4276         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
4277         * config/arm/t-arm (arm-d.o): New rule.
4278         * config/default-d.c: New file.
4279         * config/glibc-d.c: New file.
4280         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4281         * config/i386/i386-d.c: New file.
4282         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
4283         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
4284         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
4285         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
4286         * config/i386/t-i386 (i386-d.o): New rule.
4287         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4288         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4289         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
4290         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4291         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
4292         * config/mips/mips-d.c: New file.
4293         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
4294         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
4295         * config/mips/t-mips (mips-d.o): New rule.
4296         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4297         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4298         * config/powerpcspe/powerpcspe-d.c: New file.
4299         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
4300         New prototype.
4301         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
4302         Support GNU D by using 0 as the language type.
4303         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
4304         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
4305         * config/riscv/riscv-d.c: New file.
4306         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
4307         prototype.
4308         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
4309         * config/riscv/t-riscv (riscv-d.o): New rule.
4310         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4311         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4312         * config/rs6000/rs6000-d.c: New file.
4313         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
4314         prototype.
4315         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
4316         Support GNU D by using 0 as the language type.
4317         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
4318         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
4319         * config/s390/s390-d.c: New file.
4320         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
4321         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
4322         * config/s390/t-s390 (s390-d.o): New rule.
4323         * config/sparc/sparc-d.c: New file.
4324         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
4325         prototype.
4326         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
4327         * config/sparc/t-sparc (sparc-d.o): New rule.
4328         * config/t-glibc (glibc-d.o): New rule.
4329         * configure: Regenerated.
4330         * configure.ac (tm_d_file): New variable.
4331         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
4332         * doc/contrib.texi (Contributors): Add self for the D frontend.
4333         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
4334         * doc/install.texi (Configuration): Mention libphobos as an option for
4335         --enable-shared.  Mention d as an option for --enable-languages.
4336         (Testing): Mention check-d as a target.
4337         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
4338         name suffixes.  Mention d as a -x option.
4339         * doc/sourcebuild.texi (Top Level): Mention libphobos.
4340         * doc/standards.texi (Standards): Add section on D language.
4341         * doc/tm.texi: Regenerated.
4342         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
4343         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
4344         * dwarf2out.c (is_dlang): New function.
4345         (gen_compile_unit_die): Use DW_LANG_D for D.
4346         (declare_in_namespace): Return module die for D, instead of adding
4347         extra declarations into the namespace.
4348         (gen_namespace_die): Generate DW_TAG_module for D.
4349         (gen_decl_die): Handle CONST_DECLSs for D.
4350         (dwarf2out_decl): Likewise.
4351         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
4352         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
4353         kinds of aggregates.
4354         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
4355         * genhooks.c: Include d/d-target.def.
4357 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
4359         PR target/85669
4360         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
4361         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
4362         (STACK_DYNAMIC_OFFSET): Likewise.
4364 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
4366         PR target/80024
4367         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
4368         error message.
4370 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
4372         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
4373         return type and other typos.
4375 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
4377         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
4378         a shorter sequence with fewer branches.
4379         (emit_final_str_compare_gpr): Ditto.
4381 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
4383         * config/rs6000/tmmintrin.h: New file.
4384         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
4386 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
4388         * config/rs6000/mmintrin.h: Enable 32bit compilation.
4389         * config/rs6000/xmmintrin.h: Likewise.
4391 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
4393         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
4395 2018-10-26  Richard Biener  <rguenther@suse.de>
4397         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
4398         and wrapper.
4399         (vect_mark_slp_stmts_relevant): Likewise.
4400         (vect_detect_hybrid_slp_stmts): Likewise.
4401         (vect_bb_slp_scalar_cost): Likewise.
4402         (vect_remove_slp_scalar_calls): Likewise.
4404 2018-10-26  Jan Hubicka  <jh@suse.cz>
4406         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
4407         (warn_types_mismatch): Fix walk of DECL_NAME.
4408         (odr_types_equivalent_p): Fix overactive assert.
4410 2018-10-26  Richard Biener  <rguenther@suse.de>
4412         PR tree-optimization/87105
4413         * tree-vectorizer.h (_slp_tree::refcnt): New member.
4414         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
4415         refcnt.
4416         (vect_create_new_slp_node): Initialize refcnt to one.
4417         (bst_traits): Move.
4418         (scalar_stmts_set_t, bst_fail): Remove.
4419         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
4420         (vect_build_slp_tree): Add bst_map argument and lookup
4421         already created SLP nodes.
4422         (vect_print_slp_tree): Handle a SLP graph, print SLP node
4423         addresses.
4424         (vect_slp_rearrange_stmts): Handle a SLP graph.
4425         (vect_analyze_slp_instance): Adjust and free SLP nodes from
4426         the CSE map.  Fix indenting.
4427         (vect_schedule_slp_instance): Add short-cut.
4429 2018-10-26  Martin Liska  <mliska@suse.cz>
4431         PR testsuite/86158
4432         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
4433         addr_expr and not with pointers.
4435 2018-10-26  Jan Hubicka  <jh@suse.cz>
4437         * tree.c (free_lang_data_in_type): Only check main variants.
4438         * ipa-devirt.c (warn_odr): Make static.
4439         (types_same_for_odr): Drop strict variant.
4440         (types_odr_comparable): Likewise.
4441         (odr_or_derived_type_p): Look for main variants.
4442         (odr_name_hasher::equal): Cleanup comment.
4443         (odr_subtypes_equivalent): Add warn and warned arguments; check main
4444         variants.
4445         (type_variants_equivalent_p): break out from ...
4446         (odr_types_equivalent): ... here; go for main variants where needed.
4447         (warn_odr): ... here; turn static.
4448         (warn_types_mismatch): Compare mangled names of main variants.
4449         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
4450         (type_with_linkage_p): Sanity check that we look at main variant.
4451         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
4452         * tree.h (types_same_for_odr): Drop strict argument.
4454 2018-10-26  Richard Biener  <rguenther@suse.de>
4456         PR tree-optimization/87746
4457         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
4458         Simplify and fix WRT strided store groups with size not
4459         equal to step in element count.
4460         (vect_analyze_group_access_1): Dump the whole group.
4462 2018-10-25  Carl Love  <cel@us.ibm.com>
4464         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
4465         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
4466         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
4467         precicion entry for each overloaded builtin.
4468         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
4469         VSCEDPUO): Rename overloaded name.
4470         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
4471         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
4472         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
4473         define_expand for xscmpexqp instruction.
4474         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
4476 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
4477             Jinsong Ji <jji@us.ibm.com>
4479         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
4480         function with vec_sl.
4481         (_mm_slli_epi32): Likewise.
4482         (_mm_slli_epi64): Likewise.
4483         (_mm_srai_epi16): Replace deprecated function with vec_sra.
4484         (_mm_srai_epi32): Likewise.
4485         (_mm_srli_epi16): Replace deprecated function with vec_sr.
4486         (_mm_srli_epi32): Likewise.
4487         (_mm_srli_epi64): Likewise.
4488         (_mm_sll_epi16): Replace deprecated function with vec_sl.
4489         (_mm_sll_epi32): Likewise.
4490         (_mm_sll_epi64): Likewise.
4491         (_mm_sra_epi16): Replace deprecated function with vec_sra.
4492         (_mm_sra_epi32): Likewise.
4493         (_mm_srl_epi16): Replace deprecated function with vec_sr.
4494         (_mm_srl_epi32): Likewise.
4495         (_mm_srl_epi64): Likewise.
4497 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
4498             Jinsong Ji <jji@us.ibm.com>
4500         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
4501         comparison operators with vec_cmp* for compatibility due to
4502         unfortunate history; clean up formatting and use types more
4503         appropriately.
4504         (_mm_sll_epi32): Likewise.
4505         (_mm_sll_epi64): Likewise.
4506         (_mm_srl_epi16): Likewise.
4507         (_mm_srl_epi32): Likewise.
4508         (_mm_srl_epi64): Likewise.
4510 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
4511             Jinsong Ji <jji@us.ibm.com>
4513         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
4514         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
4515         __vector __bool int.  Use vec_cmpgt in preference to deprecated
4516         function vec_vcmpgtfp.
4517         (_mm_max_ps): Likewise.
4519 2018-10-25  Jeff Law  <law@redhat.com>
4521         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
4522         if returning false.
4524 2018-10-25  Martin Sebor  <msebor@redhat.com>
4526         * doc/extend.texi (aligned): Expand attribute description.
4527         (Alignment): Rename section.  Discuss function arguments.
4529 2018-10-25  Jan Hubicka  <jh@suse.cz>
4531         * ipa-devirt.c (main_odr_variant): Remove.
4532         (hash_odr_name, types_same_for_odr, types_odr_comparable,
4533         odr_name_hasher::equal, odr_subtypes_equivalent_p):
4534         Drop use of main_odr_variant.
4535         (add_type_duplicate): Silence confused warnings on integer types.
4536         (get_odr_type): Always look for main variant.
4537         (register_odr_type): Simplify.
4539 2018-10-25  Richard Biener  <rguenther@suse.de>
4541         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
4542         Initialize ng to silence error with release checking bootstrap.
4544 2018-10-25  Richard Biener  <rguenther@suse.de>
4546         * tree-if-conv.c: Include tree-ssa-sccvn.h.
4547         (tree_if_conversion): Run CSE on the if-converted loop body.
4549 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
4551         * config/s390/constraints.md (ZL): New constraint.
4552         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
4553         operands.
4554         * config/s390/s390.md (movdi_larl): Remove.
4555         (movdi_64): Add the LARL alternative.
4557 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
4559         PR bootstrap/87747
4560         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
4561         (RTX_CODE_HWINT_P): New macro.
4562         (rtx_code_size): Use RTX_CODE_HWINT_P ().
4564 2018-10-25  Jan Hubicka  <jh@suse.cz>
4566         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
4567         is anonymous.
4569 2018-10-25  Richard Biener  <rguenther@suse.de>
4571         PR tree-optimization/87665
4572         PR tree-optimization/87745
4573         * tree-vectorizer.h (get_earlier_stmt): Remove.
4574         (get_later_stmt): Pick up UID from the original non-pattern stmt.
4576 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
4578         * options.texi (Deprecated): Move list to Var section.
4580 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
4581             Jinsong Ji <jji@us.ibm.com>
4583         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
4584         __vector long to __vector long long.
4585         (_mm_cvtpd_ps): Likewise.
4586         (_mm_cvttpd_epi32): Likewise.
4587         (_mm_cvtpi32_pd): Likewise.
4588         (_mm_unpackhi_epi64): Likewise.
4589         (_mm_unpacklo_epi64): Likewise.
4591 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
4593         PR rtl-optimization/87720
4594         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
4596 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
4598         * gimple-ssa-isolate-paths.c
4599         (find_implicit_erroneous_behavior): Do not change code if the
4600         pass is running for warnings only.
4601         (find_explicit_erroneous_behavior): Likewise.
4603 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
4605         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
4606         Define as rs6000_mangle_decl_assembler_name.
4607         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
4608         long double to IEEE long double, switch the names of the long
4609         double built-in functions to be <func>f128 instead of <func>l.
4611 2018-10-24  Martin Sebor  <msebor@redhat.com>
4613         * doc/extend.texi (nonnull): List no-argument form.  Reference
4614         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
4616 2018-10-24  Richard Biener  <rguenther@suse.de>
4618         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
4620 2018-10-24  Martin Liska  <mliska@suse.cz>
4622         PR tree-optimization/84436
4623         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
4624         Remove.
4625         (switch_conversion::contains_linear_function_p): New.
4626         (switch_conversion::build_one_array): Support linear
4627         transformation on input.
4628         * tree-switch-conversion.h (struct switch_conversion): Add
4629         contains_linear_function_p declaration.
4631 2018-10-24  Richard Biener  <rguenther@suse.de>
4633         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
4634         if its argument is CONSTANT_CLASS_P.
4636 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
4638         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
4639         it is wrong for forward declarations.
4641 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
4643         * config/s390/s390.c (s390_check_qrst_address): Add the missing
4644         SYMBOL_REF_P () check.
4646 2018-10-24  Richard Biener  <rguenther@suse.de>
4648         PR tree-optimization/87105
4649         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
4650         dump classification.
4651         (vect_analyze_data_ref_accesses): Handle duplicate loads and
4652         stores by splitting the affected group after the fact.
4653         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
4654         fail the SLP build because of size constraints.
4656 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4658         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
4659         * configure: Regenerate.
4660         * config.in: Regenerate.
4661         * varasm.c (mergeable_string_section): Use readonly_data_section
4662         if linker doesn't support SHF_MERGE with alignment > 8.
4663         (mergeable_constant_section): Likewise.
4665 2018-10-24  Richard Biener  <rguenther@suse.de>
4667         PR tree-optimization/84013
4668         * tree-ssa-structalias.c (struct msdi_data): New struct for
4669         marshalling data to walk_stmt_load_store_ops.
4670         (maybe_set_dependence_info): Refactor as callback for
4671         walk_stmt_load_store_ops.
4672         (compute_dependence_clique): Set restrict info on all stmt kinds.
4674 2018-10-24  Martin Liska  <mliska@suse.cz>
4676         * cgraph.c (cgraph_node::dump):
4677         Remove reduntant dumps and make tp_first_run dump more compact.
4679 2018-10-24  Richard Biener  <rguenther@suse.de>
4681         PR tree-optimization/87665
4682         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
4683         to reflect reality.
4685 2018-10-12  Jeff Law  <law@redhat.com>
4687         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
4688         for H8/S.
4690 2018-10-23  Richard Biener  <rguenther@suse.de>
4692         * tree-vrp.c (add_assert_info): Guard dump_printf with
4693         dump_enabled_p.
4694         * gimple-ssa-evrp-analyze.c
4695         (evrp_range_analyzer::record_ranges_from_incoming_edge):
4696         Use value_range::ignore_equivs_equal_p.
4698 2018-10-23  Richard Biener  <rguenther@suse.de>
4700         PR tree-optimization/87105
4701         PR tree-optimization/87608
4702         * passes.def (pass_all_early_optimizations): Add early phi-opt
4703         after dce.
4704         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
4705         addition to debug stmts.
4706         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
4707         and abs replacement early.
4708         * tree-cfg.c (gimple_empty_block_p): Likewise.
4710 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
4712         PR target/86383
4713         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
4714         specified to configure.
4715         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
4717 2018-10-23  Richard Biener  <rguenther@suse.de>
4719         PR tree-optimization/87700
4720         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
4722 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
4724         PR target/87674
4725         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
4726         second argument from __mmask16 to __mmask8.
4727         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
4728         _mm_mask_packs_epi32): Likewise.
4729         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
4730         Likewise.
4731         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
4733 2018-10-23  Richard Biener  <rguenther@suse.de>
4735         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
4737 2018-10-23  Richard Biener  <rguenther@suse.de>
4739         PR tree-optimization/86144
4740         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
4741         over simd attribute.
4743 2018-10-23  Richard Biener  <rguenther@suse.de>
4745         PR tree-optimization/87693
4746         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
4747         the case we do not find the taken edge.
4749 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
4750             Jinsong Ji  <jji@us.ibm.com>
4752         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
4753         (_mm_store_pd): Use unaligned vector type for pointer cast.
4754         (_mm_maskmoveu_si128): Likewise.
4755         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
4756         (_mm_store_ps): Use unaligned vector type for pointer cast.
4758 2018-10-22  Paul Koning  <ni1d@arrl.net>
4760         * symtab.c (symtab_node::increase_alignment): Correct max
4761         alignment check.
4763 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
4765         PR tree-optimization/87633
4766         * match.pd: Do not generate unordered integer comparisons.
4768 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
4770         PR rtl-optimization/87600
4771         * combine.c: Add include of expr.h.
4772         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
4773         register to a pseudo.
4774         (make_more_copies): New function, add a copy to a new pseudo after
4775         the moves from hard registers into pseudos.
4776         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
4777         later.  Call make_more_copies.
4779 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
4781         * lra-constraints.c (process_alt_operands): New local array,
4782         matching_early_clobber.  Check matching_early_clobber before
4783         decrementing reject, and set matching_early_clobber after.
4785 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
4787         PR target/87598
4788         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
4789         call output_addr_const and hope for the best.
4791 2018-10-22  Richard Biener  <rguenther@suse.de>
4793         * gimple-ssa-evrp-analyze.c
4794         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
4795         smarter about what ranges to use.
4796         * tree-vrp.c (add_assert_info): Dump here.
4797         (register_edge_assert_for_2): Instead of here at multiple but
4798         not all places.
4800         * gcc.dg/tree-ssa/evrp12.c: New testcase.
4801         * gcc.dg/predict-6.c: Adjust.
4802         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
4803         * gcc.dg/tree-ssa/vrp02.c: Likewise.
4804         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
4806 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
4807         Richard Biener  <rguenther@suse.de>
4809         * bitmap.h: Update data structure documentation, including a
4810         description of bitmap views as either linked-lists or splay trees.
4811         (struct bitmap_element_def): Update comments for splay tree bitmaps.
4812         (struct bitmap_head_def): Likewise.
4813         (bitmap_list_view, bitmap_tree_view): New prototypes.
4814         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
4815         tree_form fields.
4816         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
4817         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
4818         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
4819         released bitmap element here.
4820         (bitmap_element_free): Remove.
4821         (bitmap_elt_clear_from): Work on splay tree bitmaps.
4822         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
4823         this function similar ones such that linked-list bitmap implementation
4824         functions are grouped.
4825         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
4826         and moved for grouping.
4827         (bitmap_list_insert_element_after): Renamed from
4828         bitmap_elt_insert_after, and moved for grouping.
4829         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
4830         (bitmap_tree_link_left, bitmap_tree_link_right,
4831         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
4832         bitmap_tree_link_element, bitmap_tree_unlink_element,
4833         bitmap_tree_find_element): New functions for splay-tree bitmap
4834         implementation.
4835         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
4836         Renamed and moved, see above entries.
4837         (bitmap_tree_listify_from): New function to convert part of a splay
4838         tree bitmap to a linked-list bitmap.
4839         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
4840         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
4841         (bitmap_find_bit): Remove.
4842         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
4843         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
4844         Handle splay tree bitmaps.
4845         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
4846         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
4847         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
4848         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
4849         bitmap_intersect_compl_p, bitmap_ior_and_compl,
4850         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
4851         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
4852         corresponding changes to use linked-list specific bitmap_element
4853         manipulation functions as applicable for efficiency.
4854         (bitmap_tree_to_vec): New function.
4855         (debug_bitmap_elt_file): New function split out from ...
4856         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
4857         (bitmap_print): Likewise.
4859         PR tree-optimization/63155
4860         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
4861         SSA edge worklists.
4862         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
4863         in tree-view.
4865 2018-10-22  Martin Liska  <mliska@suse.cz>
4867         PR tree-optimization/87686
4868         Revert
4869         2018-08-29  Martin Liska  <mliska@suse.cz>
4871         * tree-switch-conversion.c (switch_conversion::expand):
4872         Strenghten assumption about gswitch statements.
4874 2018-10-22  Martin Liska  <mliska@suse.cz>
4876         * ipa-icf.c (sem_item::compare_attributes): Remove.
4877         (sem_item::compare_referenced_symbol_properties): Use
4878         attribute_list_equal instead.
4879         (sem_function::equals_wpa): Likewise.
4880         * ipa-icf.h: Remove compare_attributes.
4882 2018-10-22  Richard Biener  <rguenther@suse.de>
4884         PR middle-end/87682
4885         * mem-stats.h (mem_usage::operator==): Fix pasto.
4887 2018-10-22  Richard Biener  <rguenther@suse.de>
4889         PR tree-optimization/87640
4890         * tree-vrp.c (set_value_range_with_overflow): Decompose
4891         incomplete result.
4892         (extract_range_from_binary_expr_1): Adjust.
4894 2018-10-22  Martin Jambor  <mjambor@suse.cz>
4896         * tree-eh.h (stmt_could_throw_p): Add function parameter.
4897         (stmt_can_throw_external): Likewise.
4898         (stmt_can_throw_internal): Likewise.
4899         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
4900         (lower_eh_constructs_2): Likewise.
4901         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
4902         (stmt_can_throw_external): Likewise.
4903         (stmt_can_throw_internal): Likewise.
4904         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
4905         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
4906         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
4907         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
4908         (pass_lower_eh_dispatch::execute): Pass cfun to
4909         stmt_can_throw_external.
4910         (cleanup_empty_eh): Likewise.
4911         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
4912         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
4913         stmt_can_throw_external instead of pushing it to cfun.
4914         (symbol_table::create_edge): Likewise.
4915         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
4916         stmt_can_throw_internal.
4917         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
4918         to stmt_could_throw_p.
4919         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
4920         stmt_can_throw_internal.
4921         (pass_store_merging::execute): Likewise.
4922         * gimple-ssa-strength-reduction.c
4923         (find_candidates_dom_walker::before_dom_children): Pass cfun to
4924         stmt_could_throw_p.
4925         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
4926         stmt_can_throw_internal.
4927         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
4928         to stmt_can_throw_external.
4929         (check_stmt): Pass cfun to stmt_could_throw_p.
4930         (check_stmt): Pass cfun to stmt_can_throw_external.
4931         (pass_nothrow::execute): Likewise.
4932         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
4933         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
4934         stmt_can_throw_internal.
4935         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
4936         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
4937         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
4938         * tree-complex.c (expand_complex_libcall): Pass cfun to
4939         stmt_could_throw_p and to stmt_can_throw_internal.
4940         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
4941         * tree-inline.c (copy_edges_for_bb): Likewise.
4942         (maybe_move_debug_stmts_to_successors): Likewise.
4943         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
4944         stmt_could_throw_p.
4945         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
4946         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
4947         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
4948         stmt_can_throw_internal.
4949         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
4950         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
4951         stmt_could_throw_p.
4952         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
4953         stmt_can_throw_internal.
4954         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
4955         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
4956         stmt_could_throw_p.
4957         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
4958         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
4959         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
4960         (convert_mult_to_fma_1): Likewise.
4961         (convert_to_divmod): Likewise.
4962         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
4963         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
4964         * tree-ssa-propagate.c
4965         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
4966         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
4967         (maybe_optimize_range_tests): Likewise.
4968         (linearize_expr_tree): Likewise.
4969         (reassociate_bb): Likewise.
4970         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4971         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
4972         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
4973         (handle_char_store): Likewise.
4974         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
4975         stmt_can_throw_internal.
4976         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
4977         stmt_could_throw_p.
4978         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4979         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
4980         (vectorizable_simd_clone_call): Likewise.
4981         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
4982         (gimple_stringop_fixed_value): Likewise.
4984 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
4986         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
4987         literal pool references.
4988         (s390_check_qrst_address): Adapt to the new behavior of
4989         s390_loadrelative_operand_p ().
4991 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
4993         PR target/72782
4994         * config/i386/sse.md (*andnot<mode>3_bcst): New.
4996 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
4998         PR target/72782
4999         * config/i386/sse.md (*<code><mode>3_bcst): New.
5001 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
5003         PR target/72782
5004         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
5005         V4DI, V16SI and V8DI.
5006         (*sub<mode>3<mask_name>_bcst): New.
5007         (*add<mode>3<mask_name>_bcst): Likewise.
5009 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
5010             Jinsong Ji  <jji@us.ibm.com>
5012         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
5013         __m64 with __vector unsigned long long for compatibility.
5014         (_mm_movemask_epi8): Likewise.
5015         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
5016         (_mm_cvttps_pi32): Likewise.
5017         (_mm_cvtpi32_ps): Likewise.
5018         (_mm_cvtps_pi16): Likewise.
5019         (_mm_loadh_pi): Likewise.
5020         (_mm_storeh_pi): Likewise.
5021         (_mm_movehl_ps): Likewise.
5022         (_mm_movelh_ps): Likewise.
5023         (_mm_loadl_pi): Likewise.
5024         (_mm_storel_pi): Likewise.
5025         (_mm_movemask_ps): Likewise.
5026         (_mm_shuffle_pi16): Likewise.
5028 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
5030         PR target/72782
5031         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
5032         __builtin_ia32_vfnmsubpd512_mask.
5033         (_mm512_mask_fnmsub_round_pd): Likewise.
5034         (_mm512_fnmsub_pd): Likewise.
5035         (_mm512_mask_fnmsub_pd): Likewise.
5036         (_mm512_maskz_fnmsub_round_pd): Use
5037         __builtin_ia32_vfnmsubpd512_maskz.
5038         (_mm512_maskz_fnmsub_pd): Likewise.
5039         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
5040         (_mm512_mask_fnmsub_round_ps): Likewise.
5041         (_mm512_fnmsub_ps): Likewise.
5042         (_mm512_mask_fnmsub_ps): Likewise.
5043         (_mm512_maskz_fnmsub_round_ps): Use
5044         __builtin_ia32_vfnmsubps512_maskz.
5045         (_mm512_maskz_fnmsub_ps): Likewise.
5046         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
5047         __builtin_ia32_vfnmsubpd256_mask.
5048         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
5049         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
5050         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
5051         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
5052         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
5053         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
5054         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
5055         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
5056         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
5057         __builtin_ia32_vfnmsubpd.
5058         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
5059         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
5060         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
5061         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
5062         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
5063         * config/i386/i386-builtin.def: Add
5064         __builtin_ia32_vfnmsubpd256_mask,
5065         __builtin_ia32_vfnmsubpd256_maskz,
5066         __builtin_ia32_vfnmsubpd128_mask,
5067         __builtin_ia32_vfnmsubpd128_maskz,
5068         __builtin_ia32_vfnmsubps256_mask,
5069         __builtin_ia32_vfnmsubps256_maskz,
5070         __builtin_ia32_vfnmsubps128_mask,
5071         __builtin_ia32_vfnmsubps128_maskz,
5072         __builtin_ia32_vfnmsubpd512_mask,
5073         __builtin_ia32_vfnmsubpd512_maskz,
5074         __builtin_ia32_vfnmsubps512_mask,
5075         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
5076         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
5077         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
5078         __builtin_ia32_vfnmsubpd256.
5079         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
5080         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
5081         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
5082         Likewise.
5083         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
5084         Likewise.
5085         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
5086         Likewise.
5087         (fmai_vmfnmsub_<mode><round_name>): Likewise.
5089 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
5091         PR target/72782
5092         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
5093         __builtin_ia32_vfnmaddpd512_mask.
5094         (_mm512_mask_fnmadd_round_pd): Likewise.
5095         (_mm512_fnmadd_pd): Likewise.
5096         (_mm512_mask_fnmadd_pd): Likewise.
5097         (_mm512_maskz_fnmadd_round_pd): Use
5098         __builtin_ia32_vfnmaddpd512_maskz.
5099         (_mm512_maskz_fnmadd_pd): Likewise.
5100         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
5101         (_mm512_mask_fnmadd_round_ps): Likewise.
5102         (_mm512_fnmadd_ps): Likewise.
5103         (_mm512_mask_fnmadd_ps): Likewise.
5104         (_mm512_maskz_fnmadd_round_ps): Use
5105         __builtin_ia32_vfnmaddps512_maskz.
5106         (_mm512_maskz_fnmadd_ps): Likewise.
5107         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
5108         __builtin_ia32_vfnmaddpd256_mask.
5109         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
5110         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
5111         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
5112         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
5113         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
5114         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
5115         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
5116         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
5117         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
5118         __builtin_ia32_vfnmaddpd.
5119         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
5120         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
5121         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
5122         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
5123         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
5124         * config/i386/i386-builtin.def: Add
5125         __builtin_ia32_vfnmaddpd256_mask,
5126         __builtin_ia32_vfnmaddpd256_maskz,
5127         __builtin_ia32_vfnmaddpd128_mask,
5128         __builtin_ia32_vfnmaddpd128_maskz,
5129         __builtin_ia32_vfnmaddps256_mask,
5130         __builtin_ia32_vfnmaddps256_maskz,
5131         __builtin_ia32_vfnmaddps128_mask,
5132         __builtin_ia32_vfnmaddps128_maskz,
5133         __builtin_ia32_vfnmaddpd512_mask,
5134         __builtin_ia32_vfnmaddpd512_maskz,
5135         __builtin_ia32_vfnmaddps512_mask,
5136         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
5137         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
5138         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
5139         __builtin_ia32_vfnmaddpd256.
5140         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
5141         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
5142         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
5143         Likewise.
5144         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
5145         Likewise.
5146         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
5147         Likewise.
5148         (fmai_vmfnmadd_<mode><round_name>): Likewise.
5150 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
5152         PR target/72782
5153         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
5154         __builtin_ia32_vfmsubpd512_mask.
5155         (_mm512_mask_fmsub_round_pd): Likewise.
5156         (_mm512_fmsub_pd): Likewise.
5157         (_mm512_mask_fmsub_pd): Likewise.
5158         (_mm512_maskz_fmsub_round_pd): Use
5159         __builtin_ia32_vfmsubpd512_maskz.
5160         (_mm512_maskz_fmsub_pd): Likewise.
5161         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
5162         (_mm512_mask_fmsub_round_ps): Likewise.
5163         (_mm512_fmsub_ps): Likewise.
5164         (_mm512_mask_fmsub_ps): Likewise.
5165         (_mm512_maskz_fmsub_round_ps): Use
5166         __builtin_ia32_vfmsubps512_maskz.
5167         (_mm512_maskz_fmsub_ps): Likewise.
5168         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
5169         __builtin_ia32_vfmsubpd256_mask.
5170         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
5171         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
5172         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
5173         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
5174         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
5175         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
5176         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
5177         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
5178         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
5179         __builtin_ia32_vfmsubpd.
5180         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
5181         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
5182         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
5183         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
5184         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
5185         * config/i386/i386-builtin.def: Add
5186         __builtin_ia32_vfmsubpd256_mask,
5187         __builtin_ia32_vfmsubpd256_maskz,
5188         __builtin_ia32_vfmsubpd128_mask,
5189         __builtin_ia32_vfmsubpd128_maskz,
5190         __builtin_ia32_vfmsubps256_mask,
5191         __builtin_ia32_vfmsubps256_maskz,
5192         __builtin_ia32_vfmsubps128_mask,
5193         __builtin_ia32_vfmsubps128_maskz,
5194         __builtin_ia32_vfmsubpd512_mask,
5195         __builtin_ia32_vfmsubpd512_maskz,
5196         __builtin_ia32_vfmsubps512_mask,
5197         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
5198         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
5199         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
5200         __builtin_ia32_vfmsubpd256.
5201         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
5202         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
5203         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
5204         Likewise.
5205         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
5206         Likewise.
5207         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
5208         Likewise.
5209         (fmai_vmfmsub_<mode><round_name>): Likewise.
5211 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
5213         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
5214         Remove plus.  Renamed to ...
5215         (*sub<mode>3<mask_name>_bcst): This.
5216         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
5217         (*add<mode>3<mask_name>_bcst): This.
5219 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
5221         PR target/72782
5222         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
5224 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
5226         PR target/87662
5227         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
5228         (_mm_or_epi32): Likewise.
5229         (_mm256_xor_epi32): Likewise.
5230         (_mm_xor_epi32): Likewise.
5231         (_mm256_or_epi64): Likewise.
5232         (_mm_or_epi64): Likewise.
5233         (_mm256_xor_epi64): Likewise.
5234         (_mm_xor_epi64): Likewise.
5236 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
5238         PR target/72782
5239         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
5241 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
5243         PR middle-end/87647
5244         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
5246 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
5248         * doc/ux.texi: Move @section directly after @node.
5250 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
5252         PR middle-end/85488
5253         PR middle-end/87649
5254         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
5255         depend closely nested inside of loop with ordered clause with
5256         a parameter.
5258 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
5260         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
5261         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
5262         * doc/ux.texi: New file.
5264 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
5266         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
5267         be the first CR field allocated.
5269 2018-10-19  Richard Biener  <rguenther@suse.de>
5271         PR target/87657
5272         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
5273         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
5275 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
5277         PR target/72782
5278         * config/i386/sse.md
5279         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
5280         (*add<mode>3<mask_name>_bcst_2): Likewise.
5282 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
5284         * config/i386/sse.md
5285         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
5286         Replace nonimmediate_operand with register_operand.
5287         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
5288         Likewise.
5289         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
5290         Likewise.
5292 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
5294         PR rtl-optimization/87596
5295         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
5296         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
5297         for instructions in FROM..TO range.
5299 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
5301         * cfgexpand.c (expand_one_var): Use specific wording in error message
5302         for non-local frame variables.
5303         * stor-layout.c (layout_decl): Do not issue a warning for them.
5305 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
5307         * haifa-sched.c (priority): Add force_recompute parameter.
5308         (apply_replacement): Call priority () with force_recompute = true.
5309         (restore_pattern): Likewise.
5311 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
5313         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
5314         HOST_BITS_PER_WIDE_INT.
5315         (test_vector_ops_duplicate): Likewise.
5317 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
5319         PR target/72782
5320         * config/i386/sse.md (VF_AVX512): New.
5321         (avx512bcst): Likewise.
5322         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
5323         Likewise.
5324         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
5325         Likewise.
5326         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
5327         Likewise.
5329 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
5331         * doc/invoke.texi (-dumpversion): Improve grammar.
5332         (-dumpfullversion): Make more consistent with -dumpversion.
5334 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
5336         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
5337         Set JUMP_LABEL to the jump insn.
5338         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
5339         Predict emitted jump and add label to jump insn.
5341 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
5343         PR tree-optimization/87562
5344         * input.c (get_substring_ranges_for_loc): Use
5345         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
5346         getting the linemap for the endpoint.  Verify that it's either
5347         in the same linemap as the start point's spelling location, or
5348         at least in the same file.
5350 2018-10-18  Richard Biener  <rguenther@suse.de>
5352         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
5353         feed width-specific load/store costs through ix86_vec_cost.
5354         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
5355         (k8_cost): Likewise.
5356         (bdver_cost): Likewise.
5357         (znver1_cost): Likewise.
5358         (btver1_cost): Likewise.
5359         (btver2_cost): Likewise.
5361 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
5363         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
5364         to simplify subreg of vec_merge.
5366 2018-10-18  Richard Biener  <rguenther@suse.de>
5368         * config/i386/i386.c: Fix costing of vector FMA.
5370 2018-10-18  Richard Biener  <rguenther@suse.de>
5372         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
5373         and argument.
5374         (ix86_builtin_vectorization_cost): For vec_construct properly
5375         cost insertion into SSE regs.
5376         (...): Adjust calls to ix86_vec_cost.
5378 2018-10-18  Richard Biener  <rguenther@suse.de>
5380         PR middle-end/87087
5381         Revert
5382         2018-02-07  Richard Biener  <rguenther@suse.de>
5384         PR tree-optimization/84204
5385         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5386         this place.
5388 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
5390         PR target/87537
5391         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
5392         of vec_duplicate.
5393         (test_vector_ops_duplicate): Add test for a scalar subreg of a
5394         VEC_MERGE of a VEC_DUPLICATE.
5396 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
5398         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
5399         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
5400         * doc/standards.texi (C Language): Document C2X.
5401         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
5402         (rl78_option_override): Handle "GNU C2X" language name.
5404 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
5406         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
5407         Document C17 as published in 2018.
5409 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
5411         PR middle-end/87623
5412         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
5413         bail out if both sides do not have the same storage order.
5415 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
5417         * bitmap.c (bitmap_head::dump): New.
5418         * bitmap.h (bitmap_head): Add dump().
5419         * gimple-ssa-evrp-analyze.c
5420         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
5421         (evrp_range_analyzer::set_ssa_range_info): Same.
5422         (evrp_range_analyzer::record_ranges_from_phis): Same.
5423         (evrp_range_analyzer::record_ranges_from_stmt): Same.
5424         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
5425         * gimple-ssa-sprintf.c (get_int_range): Same.
5426         (format_integer): Same.
5427         (sprintf_dom_walker::handle_gimple_call): Same.
5428         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
5429         (ipcp_vr_lattice::top_p): Same.
5430         (ipcp_vr_lattice::bottom_p): Same.
5431         (ipcp_vr_lattice::set_to_bottom): Same.
5432         (ipa_vr_operation_and_type_effects): Same.
5433         (propagate_vr_across_jump_function): Same.
5434         (ipcp_store_vr_results): Same.
5435         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
5436         (ipa_print_node_jump_functions_for_edge): Same.
5437         (ipa_get_value_range): Same.
5438         (ipa_compute_jump_functions_for_edge): Same.
5439         (ipa_write_jump_function): Same.
5440         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
5441         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
5442         Same.
5443         * vr-values.c (set_value_range_to_nonnegative): Same.
5444         (set_value_range_to_truthvalue): Same.
5445         (vr_values::get_value_range): Same.
5446         (vr_values::set_defs_to_varying): Same.
5447         (vr_values::update_value_range): Same.
5448         (symbolic_range_based_on_p): Same.
5449         (vr_values::op_with_boolean_value_range_p): Same.
5450         (vr_values::extract_range_for_var_from_comparison_expr): Same.
5451         (vr_values::extract_range_from_ssa_name): Same.
5452         (vr_values::extract_range_from_binary_expr): Same.
5453         (vr_values::extract_range_from_unary_expr): Same.
5454         (vr_values::extract_range_from_cond_expr): Same.
5455         (vr_values::extract_range_from_comparison): Same.
5456         (vr_values::check_for_binary_op_overflow): Same.
5457         (vr_values::extract_range_basic): Same.
5458         (vr_values::extract_range_from_assignment): Same.
5459         (compare_ranges): Same.
5460         (compare_range_with_value): Same.
5461         (vr_values::adjust_range_with_scev): Same.
5462         (vrp_valueize): Same.
5463         (vrp_valueize_1): Same.
5464         (vr_values::get_vr_for_comparison): Same.
5465         (vr_values::compare_name_with_value): Same.
5466         (vr_values::compare_names): Same.
5467         (vr_values::vrp_evaluate_conditional): Same.
5468         (find_case_label_ranges): Same.
5469         (vr_values::vrp_visit_switch_stmt): Same.
5470         (vr_values::extract_range_from_phi_node): Same.
5471         (vr_values::simplify_div_or_mod_using_ranges): Same.
5472         (vr_values::simplify_bit_ops_using_ranges): Same.
5473         (test_for_singularity): Same.
5474         (range_fits_type_p): Same.
5475         (vr_values::simplify_cond_using_ranges_1): Same.
5476         (vr_values::simplify_switch_using_ranges): Same.
5477         (vr_values::simplify_float_conversion_using_ranges): Same.
5478         (vr_values::two_valued_val_range_p): Same.
5479         (vr_values::add_equivalence): Move to value_range::equiv_add.
5480         * vr-values.h (vr_values::add_equivalence): Remove.
5481         (VR_INITIALIZER): Remove.
5482         * tree-vrp.c (value_range::set): New.
5483         (value_range::equiv_add): New.
5484         (value_range::value_range): New.
5485         (value_range::deep_copy): New.
5486         (value_range::check): New.
5487         (value_range::equal_p): New.
5488         (value_range::ignore_equivs_equal_p): New.
5489         (value_range::operator==): New.
5490         (value_range::operator!=): New.
5491         (value_range::symbolic_p): New.
5492         (value_range::numeric_p): New.
5493         (value_range::set_undefined): New.
5494         (value_range::set_varying): New.
5495         (value_range::may_contain_p): New.
5496         (value_range::equiv_clear): New.
5497         (value_range::singleton_p): New.
5498         (value_range::intersect): New.
5499         (value_range::dump): New.
5500         (value_range::set_and_canonicalize): New.
5501         (set_value_range): Adjust for value_range API.
5502         (set_value_range_to_undefined): Same.
5503         (set_value_range_to_varying): Same.
5504         (set_and_canonicalize_value_range): Same.
5505         (set_value_range_to_nonnull): Same.
5506         (set_value_range_to_null): Same.
5507         (range_is_null): Same.
5508         (range_is_nonnull): Same.
5509         (range_int_cst_p): Same.
5510         (range_int_cst_singleton_p): Same.
5511         (symbolic_range_p): Same.
5512         (range_includes_zero_p): Same.
5513         (value_range_constant_singleton): Same.
5514         (vrp_set_zero_nonzero_bits): Same.
5515         (ranges_from_anti_range): Same.
5516         (extract_range_into_wide_ints): Same.
5517         (extract_range_from_multiplicative_op): Same.
5518         (set_value_range_with_overflow): Same.
5519         (extract_range_from_binary_expr_1): Same.
5520         (extract_range_from_unary_expr): Same.
5521         (dump_value_range): Same.
5522         (debug_value_range): Same.
5523         (vrp_prop::check_array_ref): Same.
5524         (vrp_prop::check_mem_ref): Same.
5525         (vrp_prop::vrp_initialize): Same.
5526         (vrp_prop::visit_stmt): Same.
5527         (intersect_ranges): Same.
5528         (vrp_prop::visit_phi): Same.
5529         (vrp_prop::vrp_finalize): Same.
5530         (determine_value_range_1): Same.
5531         (determine_value_range): Same.
5532         (vrp_intersect_ranges_1): Rename to...
5533         (vrp_intersect_1): this.
5534         (vrp_intersect_ranges): Rename to...
5535         (value_range::intersect_helper): ...this.
5536         (vrp_meet_1): Rename to...
5537         (value_range::union_helper): ...this.
5538         (vrp_meet): Rename to...
5539         (value_range::union_): ...this.
5540         (copy_value_range): Remove.
5541         * tree-vrp.h (struct value_range): Rewrite into a proper class.
5542         (value_range::vrtype): New.
5543         (value_range::type): New.
5544         (value_range::equiv): New.
5545         (value_range::min): New.
5546         (value_range::max): New.
5547         (value_range::varying_p): New.
5548         (value_range::undefined_p): New.
5549         (value_range::null_p): New.
5550         (value_range::equiv_add): New.
5551         (copy_value_range): Remove.
5553 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
5555         * Makefile.in (SELFTEST_TARGETS): New.
5556         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
5557         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
5558         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
5559         c/Make-lang.in.
5560         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
5561         (selftest-c++-gdb, selftest-c++-valgrind): Move to
5562         cp/Make-lang.in.
5563         * configure: Regenerate.
5564         * configure.ac (selftest_languages): New.
5566 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
5568         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
5569         overflow wraps argument.
5570         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
5571         wide_int_range_multiplicative_op.
5572         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
5573         overflow wraps argument.
5574         (wide_int_range_multiplicative_op): Same.
5575         (wide_int_range_lshift): Same.
5576         (wide_int_range_div): Same.
5577         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
5578         (wide_int_range_lshift): Same.
5579         (wide_int_range_div): Same.
5581 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
5583         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
5584         use sign as argument.
5585         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
5586         wide_int_range_shift_undefined_p.
5588 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
5590         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
5591         Rename to...
5592         (@despeculate_copy<ALLI_TI:mode>): ... This.
5593         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
5594         switch statement.
5596 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
5598         * config.gcc: Obsolete *-*-solaris2.10*.
5599         * doc/install.texi (Specific, *-*-solaris2*): Document it.
5601 2018-10-12  Jeff Law  <law@redhat.com>
5603         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
5604         reg + sym +- const_int addressing modes.
5606 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
5608         * common.opt (fdiagnostics-minimum-margin-width=): New option.
5609         * diagnostic-show-locus.c (layout::layout): Apply the minimum
5610         margin width.
5611         (layout::start_annotation_line): Only print up to 3 of the
5612         margin character, to avoid touching the left-hand side.
5613         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
5614         minimum margin width, as set by test_diagnostic_context's ctor.
5615         (selftest::test_fixit_insert_containing_newline): Likewise.
5616         (selftest::test_fixit_insert_containing_newline_2): Likewise.
5617         (selftest::test_line_numbers_multiline_range): Clear
5618         dc.min_margin_width.
5619         * diagnostic.c (diagnostic_initialize): Initialize
5620         min_margin_width.
5621         * diagnostic.h (struct diagnostic_context): Add field
5622         "min_margin_width".
5623         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
5624         * opts.c (common_handle_option): Handle
5625         OPT_fdiagnostics_minimum_margin_width_.
5626         * selftest-diagnostic.c
5627         (selftest::test_diagnostic_context::test_diagnostic_context):
5628         Initialize min_margin_width to 6.
5629         * toplev.c (general_init): Initialize global_dc->min_margin_width.
5631 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
5633         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
5634         Fix usage of "error_at_rich_loc" in the comment.
5636 2018-10-15  Renlin Li  <renlin.li@arm.com>
5638         PR target/87563
5639         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
5640         if-conversioned loop when it contains ifn with types not
5641         supported by backend.
5642         * internal-fn.c (expand_direct_optab_fn): Add an assert.
5643         (direct_internal_fn_supported_p): New helper function.
5644         * internal-fn.h (direct_internal_fn_supported_p): Declare.
5646 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
5648         PR target/87572
5649         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
5650         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
5651         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
5653 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
5655         PR tree-optimization/87022
5656         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
5657         bits in dist vector rather than the first one.
5659 2018-10-15  Richard Biener  <rguenther@suse.de>
5661         PR middle-end/87610
5662         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
5663         (visit_loadstore): When a used restrict tag escaped verify that
5664         the points-to solution of "other" pointers do not include
5665         escaped.
5666         (compute_dependence_clique): If a used restrict tag escaped
5667         communicated that down to visit_loadstore.
5669 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
5671         * config/s390/s390.c (s390_expand_vec_init): Force vector element
5672         into reg if it isn't a general operand.
5674 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
5676         PR target/87599
5677         * config/i386/sse.md (*vec_dupv2di): Add register source to
5678         movddup.
5680 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
5682         PR target/87572
5683         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
5684         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
5685         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
5686         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
5688 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
5690         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
5691         (notice_args_size): Set it in the current trace if no insn that can
5692         throw internally has been seen yet.
5693         (connect_traces): When connecting args_size between traces, allow the
5694         incoming values not to match if there is an insn setting it before the
5695         first insn that can throw internally; in that case, force the creation
5696         of a CFI note on this latter insn.
5698 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
5700         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
5701         the base class, because using the injected-class-name was not clearly
5702         specified until DR 176.
5704 2018-10-12  Paul Koning  <ni1d@arrl.net>
5706         * config/pdp11/pdp11.md (doloop_end): New expander.
5707         (doloop_end_insn): renamed from "doloop_end".
5708         (addqi3): New pattern.
5709         (subqi3): New pattern.
5710         * config/pdp11/predicates.md (incdec_operand): New predicate.
5712 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
5714         PR middle-end/81376
5715         * real.c (format_helper::can_represent_integral_type_p): New function
5716         * real.h (format_helper::can_represent_integral_type_p): Ditto.
5717         * match.pd: New pattern.
5719 2018-10-12  Alexandre Oliva <oliva@adacore.com>
5721         * configure.ac: Introduce --enable-large-address-aware
5722         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
5723         * doc/install.texi: Document it.
5724         * configure, config.in: Rebuilt.
5725         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
5726         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
5727         (LINK_SPEC): Insert it.
5728         * config/i386/mingw-w64.h: Likewise.
5730         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
5732 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
5734         PR rtl-optimization/87600
5735         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
5737 2018-10-12  Paul Koning  <ni1d@arrl.net>
5739         * doc/md.texi (doloop_end): Document that the pattern code may
5740         need to check operand mode.
5742 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
5744         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
5745         to zero-extend between int and floating-point registers.
5746         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
5747         ldp into floating-point registers.  Add type and arch attributes.
5748         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
5749         Use f_loads for type attribute.
5751 2018-10-11  Martin Sebor  <msebor@redhat.com>
5753         * doc/extend.texi (attribute packed): Correct typos.
5755 2018-10-11  Martin Sebor  <msebor@redhat.com>
5757         * doc/extend.texi (attribute flatten): Mention interaction with
5758         noinline.
5760 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
5762         PR target/87156
5763         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5764         Set new_decl virtual flag to zero.
5766 2018-10-11  Martin Sebor  <msebor@redhat.com>
5768         PR middle-end/87593
5769         * doc/extend.texi (attribute format_arg): Discuss using multiple
5770         attributes on a single function.
5772 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
5774         PR tree-optimization/86829
5775         * match.pd (sin (atan (x))): New simplification rules.
5776         (cos (atan (x))): Likewise.
5777         * real.c (build_sinatan_real): New function.
5778         * real.h (build_sinatan_real): Prototype.
5780 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
5782         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
5783         function.
5784         (fold_mergeeo_helper): New helper function.
5785         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
5786         intrinsics.  Correct some whitespace indentation issues.
5788 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
5790         PR target/87511
5791         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
5792         Use HOST_WIDE_INT_1U for shift.
5794 2018-10-11  Doug Rupp  <rupp@adacore.com>
5795             Olivier Hainque  <hainque@adacore.com>
5797         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
5798         Pass --relax to the linker for RTPs.
5799         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
5801 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
5802             Jan Hubicka  <jh@suse.cz>
5803             Martin Jambor  <mjambor@suse.cz>
5805         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
5806         the same elements are repeated rather than printing all of them.
5807         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
5808         "repeated" elements.
5809         * read-rtl-function.c (test_loading_repeat): New function.
5810         (read_rtl_function_c_tests): Call test_loading_repeat.
5811         * rtl-tests.c (test_dumping_repeat): New function.
5812         (rtl_tests_c_tests): Call test_dumping_repeat.
5814 2018-10-11  Richard Biener  <rguenther@suse.de>
5816         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
5817         bdver?_cost): Unify to ...
5818         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
5819         * config/i386/i386.c (processor_cost_table): Adjust.
5821 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
5823         PR middle-end/87574
5824         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
5825         the thunk when expanding to GIMPLE.
5827 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5829         * varasm.c (mergeable_string_section): Don't try to move zero-length
5830         strings to the merge section.
5832 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
5834         PR target/87573
5835         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
5837 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
5839         PR target/87550
5840         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
5841         to special_args set.
5843 2018-10-10  Richard Biener  <rguenther@suse.de>
5845         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
5846         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
5847         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
5848         and recursing and pattern terminating the recursion on SSE
5849         vector width using ix86_expand_reduc.
5850         (reduc_sminmax_scal_<mode>): Split into part reducing to half
5851         width and recursing and SSE2 vector variant doing the final
5852         reduction with ix86_expand_reduc.
5853         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
5854         with terminating the recursion at AVX level, splitting that
5855         to SSE there.
5857 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
5859         * genmatch.c (error_cb): Rename to...
5860         (diagnostic_cb): ...this, converting int params to enums.
5861         (fatal_at): Update for renaming.
5862         (warning_at): Likewise.
5863         (main): Likewise.
5864         * input.c (selftest::ebcdic_execution_charset::apply):
5865         Update for renaming of...
5866         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
5867         to...
5868         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
5869         converting level and reason to enums.
5870         (class selftest::lexer_error_sink): Rename to...
5871         (class selftest::lexer_test_options): ...this, renaming field
5872         "m_errors" to "m_diagnostics".
5873         (selftest::lexer_test_options::apply): Update for renaming of...
5874         (selftest::lexer_test_options::on_error): ...this, renaming to...
5875         (selftest::lexer_test_options::on_diagnostic): ...this
5876         converting level and reason to enums.
5877         (selftest::test_lexer_string_locations_raw_string_unterminated):
5878         Update for renamings.
5879         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
5880         "reason".
5882 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
5884         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
5885         * config/rs6000/pmmintrin.h: New file.
5887 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
5889         PR tree-optimization/86659
5890         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
5892 2018-10-09  Richard Biener  <rguenther@suse.de>
5894         PR tree-optimization/63155
5895         * tree-ssa-structalias.c: Include tree-ssa.h.
5896         (get_constraint_for_ssa_var): For undefs return nothing_id.
5897         (find_func_aliases): Cleanup PHI handling.
5899 2018-10-09  Richard Biener  <rguenther@suse.de>
5901         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
5902         replacements.
5904 2018-10-09  Martin Liska  <mliska@suse.cz>
5906         * asan.c (asan_emit_stack_protection): If a stack variable
5907         is located in a same file as current function, then emit
5908         line info into variable definition string.
5910 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
5912         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
5913         information.
5915 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
5917         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
5918         on the thunk.
5920 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
5922         PR tree-optimization/86659
5923         * gimple-match.h (struct gimple_match_op): Add reverse field.
5924         (gimple_match_op::set_op): New overloaded method.
5925         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
5926         the REF_REVERSE_STORAGE_ORDER flag on the value.
5927         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
5928         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
5930 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
5932         PR middle-end/63155
5933         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
5934         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
5936 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
5938         PR target/87517
5939         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
5940         Defined with __builtin_ia32_vfmaddsubpd512_mask.
5942 2018-10-08  Richard Biener  <rguenther@suse.de>
5944         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
5945         cost the same as AVX128 ones.
5947 2018-10-08  Paul Koning  <ni1d@arrl.net>
5949         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
5950         (expand_block_move): New function.
5951         * config/pdp11/pdp11.c (output_block_move): Remove.
5952         (expand_block_move): New function.
5953         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
5954         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
5955         (*movmemhi1): Remove.
5957 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
5959         * config/s390/2827.md: Increase latencies for some FP instructions.
5961 2018-10-08  Richard Biener  <rguenther@suse.de>
5963         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5964         Open a dump scope.
5965         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
5966         * tree-vectorizer.h (dump_stmt_cost): Adjust.
5967         (add_stmt_cost): Dump return value of the hook.
5969 2018-10-08  Richard Biener  <rguenther@suse.de>
5971         PR tree-optimization/63155
5972         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
5973         (ssa_propagation_engine::ssa_propagate): Remove redundant
5974         bitmap bit clearing.
5976 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
5978         PR rtl-optimization/86939
5979         PR rtl-optimization/87479
5980         * ira.h (non_conflicting_reg_copy_p): New prototype.
5981         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
5982         (make_hard_regno_dead): Don't add conflicts for register
5983         ignore_reg_for_conflicts.
5984         (make_object_dead): Likewise.
5985         (non_conflicting_reg_copy_p): New function.
5986         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
5987         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
5988         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
5989         (make_hard_regno_dead): Don't add conflicts for register
5990         ignore_reg_for_conflicts.  Remove special conflict handling of
5991         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
5992         check_pic_pseudo_p and update callers.
5993         (mark_pseudo_dead): Don't add conflicts for register
5994         ignore_reg_for_conflicts.
5995         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
5997 2018-10-05  Andrew Waterman  <andrew@sifive.com>
5998             Jim Wilson  <jimw@sifive.com>
6000         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
6001         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
6002         new pattern using HONOR_SNANS that emits one extra instruction.
6004 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
6006         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
6007         patterns): Merge SI and DI patterns to a GPR pattern.
6008         (unnamed define_insn and define_split for record form of that): Merge
6009         to a single define_insn_and_split pattern.
6011 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
6013         PR c++/56856
6014         * input.c (expand_location_to_spelling_point): Add param "aspect"
6015         and use rather than hardcoding LOCATION_ASPECT_CARET.
6016         (get_substring_ranges_for_loc): Handle the case of a single token
6017         within a macro expansion.
6018         * input.h (expand_location_to_spelling_point): Add "aspect" param,
6019         defaulting to LOCATION_ASPECT_CARET.
6021 2018-10-05  Paul Koning  <ni1d@arrl.net>
6023         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
6024         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
6025         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
6026         (pdp11_guard_type): New function.
6028 2018-10-05  Paul Koning  <ni1d@arrl.net>
6030         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
6031         * config/pdp11/pdp11.opt (mfloat32): Remove.
6032         (mfloat64): Remove.
6033         * doc/invoke.texi (pdp11 -mfloat32): Remove:
6034         (pdp11 -mfloat64): Remove.
6036 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
6038         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
6039         (*cmp<mode>_cc_i387): Ditto.
6040         (*cmpu<mode>_cc_i387): Ditto.
6041         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
6042         * config/i386/i386.c (ix86_expand_fp_compare): Remove
6043         "scratch" argument.
6044         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
6045         Emit x86_sahf_1 pattern.
6046         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
6047         (ix86_expand_carry_flag_compare): Ditto.
6049 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
6051         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
6052         to reg_or_0_operand.  Add "C" constraint.
6053         (*cmpxf_cc_i387): Ditto.
6054         (*cmp<mode>_i387): Change operand 2 predicate
6055         to nonimm_or_0_operand.  Add "C" constraint.
6056         (*cmp<mode>_cc_i387): Ditto.
6057         (*cmp<mode>_0_i387): Remove insn pattern.
6058         (*cmp<mode>_0_cc_i387): Ditto.
6060 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
6062         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
6063         * config/i386/predicates.md (nonimm_or_0_operand): Rename
6064         from vector_move_operand.  Update all uses.
6066 2018-10-05  Martin Sebor  <msebor@redhat.com>
6068         PR tree-optimization/87490
6069         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
6070         consistently.
6072 2018-10-05  Richard Biener  <rguenther@suse.de>
6074         PR tree-optimization/63155
6075         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
6076         vertical space in dumpfiles.
6077         * tree-ssa-propagate.h
6078         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
6079         * tree-ssa-propagate.c (cfg_blocks_back): New global.
6080         (ssa_edge_worklist_back): Likewise.
6081         (curr_order): Likewise.
6082         (cfg_blocks_get): Remove abstraction.
6083         (cfg_blocks_add): Likewise.
6084         (cfg_blocks_empty_p): Likewise.
6085         (add_ssa_edge): Add to current or next worklist based on
6086         RPO index.
6087         (add_control_edge): Likewise.
6088         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
6089         into ...
6090         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
6091         iteration from CFG and SSA edge worklist so we process
6092         everything in RPO order, prioritizing forward progress
6093         over iteration.
6094         (ssa_prop_init): Allocate new worklists, do not dump
6095         immediate uses.
6096         (ssa_prop_fini): Free new worklists.
6098 2018-10-05  Richard Biener  <rguenther@suse.de>
6100         * tree-core.h (tree_block::abstract_flag): Remove.
6101         (tree_block::block_num): Make full 32bits.
6102         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
6103         * tree.h (BLOCK_ABSTRACT): Remove.
6104         * dwarf2out.c (gen_lexical_block_die): Remove dead code
6105         resulting from BLOCK_ABSTRACT being always false.
6106         (gen_inlined_subroutine_die): Likewise.
6107         (gen_block_die): Likewise.
6108         * tree.c (block_ultimate_origin): Likewise.
6109         * tree-pretty-print.c (dump_block_node): Remove code dealing
6110         with BLOCK_ABSTRACT.
6111         * tree-ssa-live.c (dump_scope_block): Likewise.
6112         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
6113         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
6115 2018-10-05   Richard Biener  <rguenther@suse.de>
6117         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
6118         is asked for initialize mode to the component mode of the
6119         vector type.
6121 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
6123         PR target/87522
6124         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
6125         assembler for -mavx.
6126         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
6128 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
6130         PR target/87509
6131         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
6132         RS6000_BTM_DFP.
6133         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
6134         to be DImode.  When using mffscrn, force the operand to a register.
6136 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
6138         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
6139         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
6140         X87MODEF mode iterator.
6141         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
6142         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
6143         X87MODEF mode iterator.
6145 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
6147         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
6148         -Wno-prio-ctor-dtor.
6150 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
6152         * Makefile.in (OBJS): Add opt-problem.o.
6153         * dump-context.h: Include "selftest.h.
6154         (selftest::temp_dump_context): New forward decl.
6155         (class dump_context): Make friend of class
6156         selftest::temp_dump_context.
6157         (dump_context::dump_loc_immediate): New decl.
6158         (class dump_pretty_printer): Move here from dumpfile.c.
6159         (class temp_dump_context): Move to namespace selftest.
6160         (temp_dump_context::temp_dump_context): Add param
6161         "forcibly_enable_dumping".
6162         (selftest::verify_dumped_text):
6163         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
6164         (selftest::verify_item):
6165         (ASSERT_IS_TEXT): Move here from dumpfile.c.
6166         (ASSERT_IS_TREE): Likewise.
6167         (ASSERT_IS_GIMPLE): Likewise.
6168         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
6169         to...
6170         (dump_context::dump_loc_immediate): ...this new function.
6171         (class dump_pretty_printer): Move to dump-context.h.
6172         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
6173         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
6174         (temp_dump_context::temp_dump_context): Move to "selftest"
6175         namespace.  Add param "forcibly_enable_dumping", and use it to
6176         conditionalize the use of m_pp;
6177         (selftest::verify_dumped_text): Make non-static.
6178         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
6179         (selftest::verify_item): Make non-static.
6180         (ASSERT_IS_TEXT): Move to dump-context.h.
6181         (ASSERT_IS_TREE): Likewise.
6182         (ASSERT_IS_GIMPLE): Likewise.
6183         (selftest::test_capture_of_dump_calls): Pass "true" for new
6184         param of temp_dump_context.
6185         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
6186         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
6187         TDF_COMPARE_DEBUG.
6188         * opt-problem.cc: New file.
6189         * opt-problem.h: New file.
6190         * optinfo-emit-json.cc
6191         (selftest::test_building_json_from_dump_calls): Pass "true" for
6192         new param of temp_dump_context.
6193         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
6194         (optinfo::emit_for_opt_problem): New function.
6195         (optinfo::emit): Clarity which emit_item is used.
6196         * optinfo.h (optinfo::get_dump_location): New accessor.
6197         (optinfo::emit_for_opt_problem): New decl.
6198         (optinfo::emit): Make const.
6199         * selftest-run-tests.c (selftest::run_tests): Call
6200         selftest::opt_problem_cc_tests.
6201         * selftest.h (selftest::opt_problem_cc_tests): New decl.
6202         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
6203         bool to opt_result, converting fprintf messages to
6204         opt_result::failure_at calls.  Add "stmt" param for use by the
6205         failure_at calls.
6206         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
6207         (runtime_alias_check_p): Convert return type from bool to
6208         opt_result, converting dump_printf calls to
6209         opt_result::failure_at, using the statement DDR_A for their
6210         location.
6211         (find_data_references_in_stmt): Convert return type from bool to
6212         opt_result, converting "return false" to opt_result::failure_at
6213         with a new message.
6214         * tree-data-ref.h: Include "opt-problem.h".
6215         (dr_analyze_innermost): Convert return type from bool to opt_result,
6216         and add a const gimple * param.
6217         (find_data_references_in_stmt): Convert return type from bool to
6218         opt_result.
6219         (runtime_alias_check_p): Likewise.
6220         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
6221         dr_analyze_innermost.
6222         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
6223         Convert return type from bool to opt_result, adding a message for
6224         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
6225         (vect_analyze_data_ref_dependence): Convert return type from bool
6226         to opt_result.  Change sense of return type from "false"
6227         effectively meaning "no problems" to "false" meaning a problem,
6228         so that "return false" becomes "return opt_result::success".
6229         Convert "return true" calls to opt_result::failure_at, using
6230         the location of statement A rather than vect_location.
6231         (vect_analyze_data_ref_dependences): Convert return type from bool
6232         to opt_result.
6233         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
6234         calls to opt_result::failure_at, using the stmt location rather
6235         than vect_location.
6236         (vect_verify_datarefs_alignment): Convert return type from bool
6237         to opt_result.
6238         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
6239         into multiple more-tightly-scoped copies.
6240         (vect_analyze_data_refs_alignment): Convert return type from bool
6241         to opt_result.
6242         (vect_analyze_data_ref_accesses): Likewise, converting a
6243         "return false" to a "return opt_result::failure_at", adding a
6244         new message.
6245         (vect_prune_runtime_alias_test_list): Convert return type from
6246         bool to opt_result, converting dump_printf_loc to
6247         opt_result::failure_at.  Add a %G to show the pertinent statement,
6248         and use the stmt's location rather than vect_location.
6249         (vect_find_stmt_data_reference): Convert return type from
6250         bool to opt_result, converting dump_printf_loc to
6251         opt_result::failure_at, using stmt's location.
6252         (vect_analyze_data_refs):  Convert return type from bool to
6253         opt_result.  Convert "return false" to "return
6254         opt_result::failure_at", adding messages as needed.
6255         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
6256         type from bool to opt_result.
6257         (vect_determine_vf_for_stmt): Likewise.
6258         (vect_determine_vectorization_factor): Likewise, converting
6259         dump_printf_loc to opt_result::failure_at, using location of phi
6260         rather than vect_location.
6261         (vect_analyze_loop_form_1): Convert return type from bool to
6262         opt_result, converting dump_printf_loc calls, retaining the use of
6263         vect_location.
6264         (vect_analyze_loop_form): Convert return type from loop_vec_info
6265         to opt_loop_vec_info.
6266         (vect_analyze_loop_operations): Convert return type from bool to
6267         opt_result, converting dump_printf_loc calls, using the location
6268         of phi/stmt rather than vect_location where available.  Convert
6269         various "return false" to "return opt_result::failure_at" with
6270         "unsupported phi" messages.
6271         (vect_get_datarefs_in_loop): Convert return type from bool to
6272         opt_result.  Add a message for the
6273         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
6274         (vect_analyze_loop_2): Convert return type from bool to
6275         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
6276         each "goto again;", adding new messages where needed.
6277         Add "unsupported grouped {store|load}" messages.
6278         (vect_analyze_loop): Convert return type from loop_vec_info to
6279         opt_loop_vec_info.
6280         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
6281         bool to opt_result.
6282         * tree-vect-stmts.c (process_use): Likewise, converting
6283         dump_printf_loc call and using stmt location, rather than
6284         vect_location.
6285         (vect_mark_stmts_to_be_vectorized): Likeise.
6286         (vect_analyze_stmt): Likewise, adding a %G.
6287         (vect_get_vector_types_for_stmt): Convert return type from bool to
6288         opt_result, converting dump_printf_loc calls and using stmt
6289         location, rather than vect_location.
6290         (vect_get_mask_type_for_stmt): Convert return type from tree to
6291         opt_tree, converting dump_printf_loc calls and using stmt location.
6292         * tree-vectorizer.c: Include "opt-problem.h.
6293         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
6294         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
6295         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
6296         enabled, use it to report at the top level "couldn't vectorize
6297         loop" followed by the problem.
6298         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
6299         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
6300         to opt_result.
6301         (vect_analyze_stmt): Likewise.
6302         (vect_get_vector_types_for_stmt): Likewise.
6303         (tree vect_get_mask_type_for_stmt): Likewise.
6304         (vect_analyze_data_ref_dependences): Likewise.
6305         (vect_enhance_data_refs_alignment): Likewise.
6306         (vect_analyze_data_refs_alignment): Likewise.
6307         (vect_verify_datarefs_alignment): Likewise.
6308         (vect_analyze_data_ref_accesses): Likewise.
6309         (vect_prune_runtime_alias_test_list): Likewise.
6310         (vect_find_stmt_data_reference): Likewise.
6311         (vect_analyze_data_refs): Likewise.
6312         (vect_analyze_loop): Convert return type from loop_vec_info to
6313         opt_loop_vec_info.
6314         (vect_analyze_loop_form): Likewise.
6315         (vect_analyze_slp): Convert return type from bool to opt_result.
6317 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
6319         * doc/invoke.texi (-fopt-info): Document new "internals"
6320         sub-option.
6321         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
6322         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
6323         MSG_ALL_KINDS.
6324         (optinfo_verbosity_options): Add "internals".
6325         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
6326         (dump_context::apply_dump_filter_p): New member function.
6327         (dump_context::dump_loc): Use apply_dump_filter_p rather than
6328         explicitly masking the dump_kind.
6329         (dump_context::begin_scope): Increment the scope depth first.  Use
6330         apply_dump_filter_p rather than explicitly masking the dump_kind.
6331         (dump_context::emit_item): Use apply_dump_filter_p rather than
6332         explicitly masking the dump_kind.
6333         (dump_dec): Likewise.
6334         (dump_hex): Likewise.
6335         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
6336         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
6337         (opt_info_switch_p): Update handling of default
6338         MSG_OPTIMIZED_LOCATIONS to cope with default of
6339         MSG_PRIORITY_USER_FACING.
6340         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
6341         masking the dump_kind.
6342         (selftest::test_capture_of_dump_calls): Update test_dump_context
6343         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
6344         than MSG_ALL.  Generalize scope test to be run at all four
6345         combinations of with/without MSG_PRIORITY_USER_FACING and
6346         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
6347         for each of the two values.
6348         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
6349         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
6350         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
6351         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
6352         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
6353         with MSG_PRIORITY_*.
6354         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
6355         dump messages as MSG_PRIORITY_USER_FACING.
6356         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
6357         about the interaction with MSG_PRIORITY_*.
6359 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6361         * varasm.c (output_constant): Add new parameter merge_strings.
6362         Make strings properly zero terminated in merge string sections.
6363         (mergeable_string_section): Don't fail if the last char is non-zero.
6364         (assemble_variable_contents): Handle merge string sections.
6365         (assemble_variable): Likewise.
6366         (assemble_constant_contents): Likewise.
6367         (output_constant_def_contents): Likewise.
6368         (output_constructor_array_range,
6369         output_constructor_regular_field): Adjust call to output_constant.
6370         (output_object_block): Adjust call to assemble_constant_contents
6371         and assemble_variable_contents.
6373 2018-10-04  Martin Liska  <mliska@suse.cz>
6375         PR c/87483
6376         * cgraphunit.c (process_function_and_variable_attributes):
6377         Warn about a function with alias attribute and a body.
6379 2018-10-04  Martin Liska  <mliska@suse.cz>
6381         PR ipa/82625
6382         * multiple_target.c (redirect_to_specific_clone): New function.
6383         (ipa_target_clone): Use it.
6384         * tree-inline.c: Fix comment.
6386 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
6388         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
6389         fields.
6390         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
6391         (gcc::dump_manager::register_pass): New member function, adapted
6392         from loop body in gcc::pass_manager::register_pass, adding a
6393         call to update_dfi_for_opt_info.
6394         (gcc::dump_manager::opt_info_enable_passes): Store the
6395         -fopt-info options into the new fields.  Move the loop
6396         bodies into...
6397         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
6398         function.
6399         * dumpfile.h (struct opt_pass): New forward decl.
6400         (gcc::dump_manager::register_pass): New decl.
6401         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
6402         (class gcc::dump_manager): Add fields "m_optgroup_flags",
6403         "m_optinfo_flags", and "m_optinfo_filename".
6404         * passes.c (gcc::pass_manager::register_pass): Move all of the
6405         dump-handling code to gcc::dump_manager::register_pass.
6407 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
6409         PR rtl-optimization/87466
6410         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
6411         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
6412         * doc/tm.texi: Regenerate.
6413         * ira-lives.c (process_bb_node_lives): Use the new target hook.
6414         * lra-lives.c (process_bb_lives): Likewise.
6415         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
6416         Define.
6418 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
6420         * params.c (add_params): Fix initialization.
6422 2018-10-04  Martin Liska  <mliska@suse.cz>
6424         PR gcov-profile/84107
6425         * tree-profile.c (init_ic_make_global_vars):
6426         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
6427         Come up with new ic_tuple* variables.  Emit
6428         __gcov_indirect_call{,_topn} variables.
6429         (gimple_gen_ic_profiler): Access the variable
6430         and emit gimple.
6431         (gimple_gen_ic_func_profiler): Access
6432         __gcov_indirect_call.callee field.
6433         (gimple_init_gcov_profiler): Use ptr_type_node.
6434         * value-prof.c (gimple_ic): Use ptr_type_node.
6436 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
6438         PR tree-optimization/85787
6439         * ipa-pure-const.c (malloc_candidate_p_1): Move most of
6440         malloc_candidate_p into this function and add support for
6441         detecting multiple phis.
6442         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
6444 2018-10-04  Martin Liska  <mliska@suse.cz>
6446         PR ipa/87491
6447         * ipa-inline.c (inline_to_all_callers_1):
6448         Call ultimate_alias_target for node being inlined.
6450 2018-10-03  Jeff Law  <law@redhat.com>
6452         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
6453         target's wchar_t.
6454         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
6455         * tree.h (get_typenode_from_name): Prototype.
6457 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
6459         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
6460         Change operand 2 predicate to nonimmediate_operand.
6461         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
6463 2018-10-03  Martin Sebor  <msebor@redhat.com>
6464             Jeff Law  <law@redhat.com>
6466         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
6467         initialize it.
6468         (get_string_length): Detect unterminated arrays.
6469         (format_string): Same.
6470         (format_directive): Warn about unterminated arrays.
6471         (handle_gimple_call): Mark statements with no_warning as needed.
6473 2018-10-03  Jim Wilson  <jimw@sifive.com>
6475         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
6476         also define __riscv_abi_rve.  Delete trailing white space.
6478 2018-10-03  Paul Koning  <ni1d@arrl.net>
6480         Enable LRA register allocator for PDP11.
6481         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
6482         (R): Likewise.
6483         (D): Likewise.
6484         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
6485         * config/pdp11/pdp11.opt (-mlra): New option.
6486         * doc/invoke.texi (PDP-11 Options): Document -mlra.
6488 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
6490         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
6491         (*<absneg:code>extend<mode>xf2): Ditto.
6493 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
6495         PR tree-optimization/87415
6496         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
6497         precision fields.
6499 2018-10-02  Jeff Law  <law@redhat.com>
6501         * gimple-fold.c (get_range_strlen): Only set *nonstr when
6502         an unterminated string is discovered.  Bubble up range
6503         even for unterminated strings.
6504         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
6505         indicates the string was not terminated via NONSTR.
6507 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
6509         * tree-vrp.c (extract_range_from_unary_expr): Special case all
6510         pointer conversions.
6511         Do not do anything special for anti-ranges.
6513 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
6515         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
6516         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
6518 2018-10-03  Martin Liska  <mliska@suse.cz>
6520         PR gcov-profile/86109
6521         * coverage.c (coverage_begin_function): Do not
6522         mark lambdas as artificial.
6523         * tree-core.h (struct GTY): Remove tm_clone_flag
6524         and introduce new lambda_function.
6525         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
6527 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
6529         PR target/87474
6530         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
6531         P8_VECTOR and VSX are enabled.
6533 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
6535         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
6536         0x3907 as CPU model number.
6538 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
6540         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
6541         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
6542         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
6543         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
6544         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
6545         * config/s390/s390.md: Likewise. Rename also the cpu attribute
6546         value from arch12 to z14.
6548 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
6550         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
6551         (isinfxf2): Ditto.
6552         (isinf<mode>2): Ditto.
6554 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
6556         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
6557         before emitting fxam.  Perform calculations in XFmode.
6559 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
6561         * match.pd (((X /[ex] A) +- B) * A): New transformation.
6563 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
6565         PR middle-end/87319
6566         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
6567         * tree.c (signed_or_unsigned_type_for): Handle complex.
6569 2018-10-02  Jeff Law  <law@redhat.com>
6571         * gimple-fold.c (get_range_strlen): Remove dead code.
6573 2018-10-02  Martin Sebor  <msebor@redhat.com>
6574             Jeff Law  <law@redhat.com>
6576         * builtins.c (unterminated_array): Add new arguments.
6577         If argument is not terminated, bubble up size and exact
6578         state to callers.
6579         (expand_builtin_strnlen): Detect, avoid expanding
6580         and diagnose unterminated arrays.
6581         (c_strlen): Fill in offset of start of unterminated strings.
6582         * builtins.h (unterminated_array): Update prototype.
6584 2018-10-02  Richard Biener  <rguenther@suse.de>
6586         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
6587         of haddv4df, first reduce to SSE width and exploit the fact
6588         that we only need element zero with the reduction result.
6589         (reduc_plus_scal_v2df): Likewise.
6591 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
6593         * dojump.h (do_jump): Delete.
6594         (do_jump_1): Likewise.
6595         (split_comparison): Move around.
6596         * dojump.c (do_jump): Make static.
6597         (do_jump_1): Likewise.
6598         (jumpifnot): Move around.
6599         (jumpifnot_1): Likewise.
6600         (jumpif): Likewise.
6601         (jumpif_1): Likewise.
6602         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
6604 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
6606         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
6607         insns in the delay slot and add_insn_after for the jump insn.
6609 2018-10-02  Richard Biener  <rguenther@suse.de>
6611         * tree-inline.c (expand_call_inline): Use the location of
6612         the callee declaration for the inline-entry marker.
6613         * final.c (notice_source_line): Remove special-casing of
6614         NOTE_INSN_INLINE_ENTRY.
6616 2018-10-01  Carl Love  <cel@us.ibm.com>
6618         PR 69431
6619         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
6620         (__builtin_mtfsb0): New.
6621         (__builtin_mtfsb1): New.
6622         ( __builtin_set_fpscr_rn): New.
6623         (__builtin_set_fpscr_drn): New.
6624         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
6625         (rs6000_expand_set_fpscr_rn_builtin): Add.
6626         (rs6000_expand_set_fpscr_drn_builtin): Add.
6627         (rs6000_expand_builtin): Add case statement entries for
6628         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
6629         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
6630         RS6000_BUILTIN_MFFSL.
6631         (rs6000_init_builtins): Add ftype initialization and def_builtin
6632         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
6633         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
6634         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
6635         rs6000_mffscdrn): Add define_insn.
6636         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
6637         * doc/extend.texi: Add documentation for the builtins.
6639 2018-10-01  Richard Biener  <rguenther@suse.de>
6641         PR tree-optimization/87465
6642         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
6643         causing branch miscounts.
6645 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6647         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
6648         aarch64_option_default_param):  New.
6649         (params.h): Include.
6650         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
6651         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
6652         stack-clash protection validation code.
6654 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6656         * params.c (validate_param): New.
6657         (add_params): Use it.
6658         (set_param_value): Refactor param validation into validate_param.
6659         (diagnostic.h): Include.
6660         * diagnostic.h (diagnostic_ready_p): New.
6662 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6664         * params.c (set_param_value):
6665         Add index of parameter being validated.
6666         * common/common-target.def (option_validate_param): New.
6667         * common/common-targhooks.h (default_option_validate_param): New.
6668         * common/common-targhooks.c (default_option_validate_param): New.
6669         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
6670         * doc/tm.texi: Regenerate.
6672 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6674         PR target/86486
6675         * config/aarch64/aarch64.c (aarch64_override_options_internal):
6676         Add validation for stack-clash parameters and set defaults.
6678 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6680         PR target/86486
6681         * configure.ac: Add stack-clash-protection-guard-size.
6682         * doc/install.texi: Document it.
6683         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
6684         * params.def: Update comment for guard-size.
6685         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
6686         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
6687         * configure: Regenerate.
6689 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6691         PR target/86486
6692         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
6693         STACK_DYNAMIC_OFFSET): New.
6694         * config/aarch64/aarch64.c (aarch64_layout_frame):
6695         Update outgoing args size.
6696         (aarch64_stack_clash_protection_alloca_probe_range,
6697         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
6699 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6701         PR target/86486
6702         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
6703         probe ranges.
6704         * target.def (stack_clash_protection_alloca_probe_range): New.
6705         (stack_clash_protection_final_dynamic_probe): Remove.
6706         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
6707         (default_stack_clash_protection_final_dynamic_probe): Remove.
6708         * targhooks.c: Likewise.
6709         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
6710         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
6711         * doc/tm.texi: Regenerate.
6713 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6715         PR target/86486
6716         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
6717         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
6718         aarch64_clamp_to_uimm12_shift): New.
6719         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
6720         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
6722 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
6724         PR target/86486
6725         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
6727 2018-10-01  Jeff Law  <law@redhat.com>
6728             Richard Sandiford <richard.sandiford@linaro.org>
6729             Tamar Christina  <tamar.christina@arm.com>
6731         PR target/86486
6732         * config/aarch64/aarch64.md
6733         (probe_stack_range): Add k (SP) constraint.
6734         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
6735         STACK_CLASH_MAX_UNROLL_PAGES): New.
6736         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
6737         stack probes for stack clash.
6738         (aarch64_allocate_and_probe_stack_space): New.
6739         (aarch64_expand_prologue): Use it.
6740         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
6741         (aarch64_sub_sp): Add emit_move_imm optional param.
6743 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
6745         PR tree-optimization/87261
6746         * match.pd: Remove trailing whitespace.
6747         Add (x & y) | ~(x | y) -> ~(x ^ y),
6748         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
6750 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
6752         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
6753         constraints.
6754         (ashlsi3_insn): Update instruction constraints.
6755         (ashrsi3_insn): Likewise.
6756         (rotrsi3): Likewise.
6757         (add_shift): Likewise.
6758         * config/arc/constraints.md (Csz): New 32 bit constraint. It
6759         avoids placing in the limm field small constants which, otherwise,
6760         could end into a small instruction.
6762 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
6764         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
6765         destination register is not odd-even.
6766         (umaddsidi4_split): Likewise.
6768 2018-10-01  Richard Biener  <rguenther@suse.de>
6770         * tree-inline.c (expand_call_inline): Store origin of fn
6771         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
6772         * tree.c (block_ultimate_origin): Simplify and do some
6773         checking.
6775 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
6777         * config/i386/mmx.md (EMMS): New int iterator.
6778         (emms): New int attribute.
6779         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
6780         EMMS int iterator.  Explicitly declare clobbers.
6781         (mmx_emms): Remove expander.
6782         (mmx_femms): Ditto.
6783         * config/i386/predicates.md (emms_operation): Remove predicate.
6784         (vzeroall_pattern): New predicate.
6785         (vzeroupper_pattern): Rename from vzeroupper_operation.
6786         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
6787         vzeroupper_pattern and vzeroall_pattern predicates.
6789 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
6791         PR rtl-optimization/86939
6792         * ira-lives.c (make_hard_regno_born): Rename from this...
6793         (make_hard_regno_live): ... to this.  Remove update to conflict
6794         information.  Update function comment.
6795         (make_hard_regno_dead): Add conflict information update.  Update
6796         function comment.
6797         (make_object_born): Rename from this...
6798         (make_object_live): ... to this.  Remove update to conflict information.
6799         Update function comment.
6800         (make_object_dead):  Add conflict information update.  Update function
6801         comment.
6802         (mark_pseudo_regno_live): Call make_object_live.
6803         (mark_pseudo_regno_subword_live): Likewise.
6804         (mark_hard_reg_dead): Update function comment.
6805         (mark_hard_reg_live): Call make_hard_regno_live.
6806         (process_bb_node_lives): Likewise.
6807         * lra-lives.c (make_hard_regno_born): Rename from this...
6808         (make_hard_regno_live): ... to this.  Remove update to conflict
6809         information.  Remove now uneeded check_pic_pseudo_p argument.
6810         Update function comment.
6811         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
6812         to conflict information.  Update function comment.
6813         (mark_pseudo_live): Remove update to conflict information.  Update
6814         function comment.
6815         (mark_pseudo_dead): Add conflict information update.
6816         (mark_regno_live): Call make_hard_regno_live.
6817         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
6818         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
6820 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
6822         PR target/87370
6823         * config/i386/i386.c (construct_container): Use TImode for
6824         BLKmode values in 2 integer registers.
6826 2018-09-29  Jeff Law  <law@redhat.com>
6828         * builtins.c (unterminated_array): Pass in c_strlen_data * to
6829         c_strlen rather than just a tree *.
6830         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
6831         Update recursive calls appropriately.  If caller did not provide a
6832         suitable data pointer, create a local one.  When a non-terminated
6833         string is discovered, bubble up information about the string via the
6834         c_strlen_data object.
6835         * builtins.h (c_strlen): Update prototype.
6836         (c_strlen_data): New structure.
6837         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
6838         For a type 2 call, if c_strlen indicates a non-terminated string
6839         use the length of the non-terminated string.
6840         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
6842 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
6844         PR target/87467
6845         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
6846         __m512d type for __A argument rather than __m512.
6848 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
6850         * match.pd (simple_comparison): Don't optimize if either operand is
6851         a function pointer when target needs function pointer canonicalization.
6853 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
6855         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
6856         power5 .. power9 to remove indirection.
6857         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
6858         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
6859         ASM_CPU_476_SPEC): Delete.
6860         (ASM_CPU_SPEC): Adjust.
6861         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
6862         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
6864 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
6866         * config.in: Delete HAVE_AS_DCI.
6867         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
6868         * config/rs6000/rs6000.h: Ditto.
6869         * configure.ac: Delete HAVE_AS_DCI.
6870         * configure: Regenerate.
6872 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
6874         * config.in (HAVE_AS_LWSYNC): Delete.
6875         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
6876         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
6877         do it as a .long .
6878         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
6879         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
6880         as a .long .
6881         * configure.ac: Delete HAVE_AS_LWSYNC.
6882         * configure: Regenerate.
6884 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
6885             Pierre-Marie de Rodat  <derodat@adacore.com>
6887         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
6888         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
6889         (cgraph_node::create_thunk): Add indirect_offset parameter.
6890         (thunk_adjust): Likewise.
6891         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
6892         and initialize the corresponding field with it.
6893         (cgraph_node::dump): Dump indirect_offset field.
6894         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
6895         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
6896         (thunk_adjust): Add indirect_offset parameter and deal with it.
6897         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
6898         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
6899         or if the thunk is external or local.  Fix formatting.  Do not chain
6900         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
6901         if any, in the GIMPLE representation.
6902         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
6903         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
6904         (input_node): Read indirect_offset field.
6905         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
6906         call to thunk_adjust.
6907         * tree-nested.c (struct nesting_info): Add thunk_p field.
6908         (create_nesting_tree): Set it.
6909         (convert_all_function_calls): Copy static chain from targets to thunks.
6910         (finalize_nesting_tree_1): Return early for thunks.
6911         (unnest_nesting_tree_1): Do not finalize thunks.
6912         (gimplify_all_functions): Do not gimplify thunks.
6914 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
6916         * opt-suggestions.c (option_proposer::build_option_suggestions):
6917         Release "option_values".
6919 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
6921         * coverage.c (get_coverage_counts): Convert problem-reporting dump
6922         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
6923         * dumpfile.c (kind_as_string): New function.
6924         (dump_loc): Rather than a hardcoded prefix of "note: ", use
6925         kind_as_string to vary the prefix based on dump_kind.
6926         (selftest::test_capture_of_dump_calls): Update for above.
6928 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
6930         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
6931         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
6933 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
6935         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
6936         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
6937         INVALID_REGNUM instead of FPSR_REG.
6938         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
6939         * config/i386/i386.md: Update comment of FP compares.
6940         (fldenv): Do not clobber FPSR_REG.
6942 2018-09-28  Richard Biener  <rguenther@suse.de>
6944         * tree.h (BLOCK_ORIGIN): New.
6945         * omp-expand.c (grid_expand_target_grid_body): Assign
6946         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
6947         * tree-inline.c (remap_block): Likewise.
6948         * auto-profile.c (get_function_decl_from_block): Simplify
6949         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
6950         * langhooks.c (lhd_print_error_function): Likewise.
6951         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
6952         Likewise.
6953         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6954         * tree.c (block_nonartificial_location): Likewise.
6955         (block_ultimate_origin): Likewise.
6956         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
6957         no longer needed LTO case.
6959 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
6960             Jan Hubicka  <jh@suse.cz>
6961             Martin Jambor  <mjambor@suse.cz>
6963         * simplify-rtx.c (simplify_merge_mask): New function.
6964         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
6965         same masks are used in op1 or op2.
6966         (test_vec_merge): New function.
6967         (test_vector_ops): Call test_vec_merge.
6969 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
6971         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
6972         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
6973         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
6974         (cypress_costs): Set it.
6975         (supersparc_costs): Likewise.
6976         (hypersparc_costs): Likewise.
6977         (leon_cost): Likewise.
6978         (leon3_costs): Likewise.
6979         (sparclet_costs): Likewise.
6980         (ultrasparc_costs): Likewise.
6981         (ultrasparc_costs): Likewise.
6982         (niagara_costs): Likewise.
6983         (niagara2_costs): Likewise.
6984         (niagara3_costs): Likewise.
6985         (niagara4_costs): Likewise.
6986         (niagara7_costs): Likewise.
6987         (m8_costs): Likewise.
6988         (TARGET_CAN_FOLLOW_JUMP): Define.
6989         (pass_work_around_errata::gate): Minor tweak.
6990         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
6991         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
6992         Automaitcally clear MASK_FSMULD mask for V7 processors.
6993         (sparc_can_follow_jump): New static function.
6994         (output_ubranch): Deal with CROSSING_JUMP_P.
6995         (sparc_use_sched_lookahead): Rewrite using switch statement.
6996         (sparc_issue_rate): Reorder.
6997         (sparc_branch_cost): New function.
6999 2018-09-27  Martin Sebor  <msebor@redhat.com>
7001         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
7002         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
7003         (int_fits_type_p): Same.
7005 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
7007         * config/i386/i386.md (FPCR_REG): Remove.
7008         (UNSPEC_FLDCW): Remove.
7009         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
7010         (x86_fldcw_1): Remove insn pattern.
7011         (fnstenv): Do not clobber FPCR_REG.
7012         (fldenv): Ditto.
7013         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
7014         (CALL_USED_REGISTERS): Ditto.
7015         (REG_ALLOC_ORDER): Ditto.
7016         (REG_CLASS_CONTENTS): Ditto.
7017         (HI_REGISTER_NAMES): Ditto.
7018         (ADDITIONAL_REGISTER_NAMES): Use defines instead
7019         of numerical constants.
7020         * config/i386/i386.c (regclass_map): Remove fpsr register.
7021         (dbx_register_map): Ditto.
7022         (dbx64_register_map): Ditto.
7023         (svr4_dbx_register_map): Ditto.
7024         (print_reg): Do not handle FPCR_REG.
7026 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
7028         PR target/87149
7029         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
7030         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
7031         Delete, always treat as true.
7032         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
7033         Ditto.  Simplify remaining code.
7034         * config/powerpcspe/powerpcspe.h: Ditto.
7035         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
7036         Simplify remaining code.
7037         (rs6000_expand_builtin): Ditto.
7038         * config/rs6000/rs6000.h: Ditto.
7039         * configure.ac: Ditto.
7040         * configure: Regenerate.
7042 2018-09-27  Martin Liska  <mliska@suse.cz>
7044         * coverage.c (get_coverage_counts): Revert the formatting
7045         of missing profile opt info.
7047 2018-09-27  Richard Biener  <rguenther@suse.de>
7049         PR debug/37801
7050         PR debug/87440
7051         * dwarf2out.c (set_block_origin_self): Do not mark outermost
7052         block as we do not output that.
7053         (gen_inlined_subroutine_die): Elide the originally outermost
7054         block, matching what we do for concrete instances.
7055         (decls_for_scope): Add parameter specifying whether to recurse
7056         to subblocks.
7058 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
7059             Tom de Vries  <tom@codesourcery.com>
7061         PR 82089
7063         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
7064         STORE_FLAG_VALUE == 1.
7066 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
7068         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
7069         constant definitions.
7070         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
7071         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
7072         ("speculation_barrier"): New expander definition.
7074 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
7076         PR gcov-profile/86957
7077         * common.opt: New warning option -Wmissing-profile.
7078         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
7079         * doc/invoke.texi: Document -Wmissing-profile.
7081 2018-09-26  Jim Wilson  <jimw@sifive.com>
7083         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
7084         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
7086 2018-09-26  Martin Sebor  <msebor@redhat.com>
7088         * tree.c (zerop): Change return type to bool.
7089         (integer_zerop, integer_onep, integer_each_onep): Same.
7090         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
7091         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
7092         (real_onep, real_minus_onep, chain_index): Same.
7093         (print_type_hash_statistics, type_list_equal): Same.
7094         * tree.h (zerop): Same.
7095         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
7096         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
7097         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
7098         (real_onep, real_minus_onep, chain_index): Same.
7099         (print_type_hash_statistics, type_list_equal): Same.
7101 2018-09-26  Jim Wilson  <jimw@sifive.com>
7103         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
7105 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
7107         PR target/87414
7108         * config/i386/i386.c: Include debug.h and dwarf2out.h.
7109         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
7110         call.
7112 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
7114         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
7116 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
7118         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
7119         and FP_SECOND_SSE_REGS.
7120         (REG_CLASS_NAMES): Ditto.
7121         (REG_CLASS_CONTENTS): Ditto.
7122         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
7123         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
7124         (ix86_preferred_output_reload_class): Ditto.
7125         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
7126         clobber constraint to "=&f".
7127         (fix_truncdi_i387): Ditto.
7128         (lrintxfdi2): Ditto.
7129         (fistdi2_<rounding>): Ditto.
7130         (fpremxf4_i387): Change "=u" constraint to "=f".
7131         (fprem1xf4_i387): Ditto.
7132         (sincosxf3): Ditto.
7133         (fptanxf4_i387): Ditto.
7134         (fxtractxf3_i387): Ditto.
7135         (fscalexf4_i387): Ditto.
7136         (atan2xf3): Change "u" constraint to "f".
7137         (fyl2xxf3_i387): Ditto.
7138         (fyl2xp1xf3_i387): Ditto.
7140 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
7142         PR target/87439
7143         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
7144         for removed I387_MASK_PM entity.
7147 2018-09-26  Jeff Law  <law@redhat.com>
7148         Revert
7149         2018-09-26  Alexey Neyman  <stilor@att.net>
7151         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
7152         headers are no longer pulled in by <isl/val.h>.
7154 2018-09-26  Richard Biener  <rguenther@suse.de>
7156         PR debug/87443
7157         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
7158         or concrete instance DIE to the tree.  Create abstract origin
7159         attributes also for concrete instances.
7161 2018-09-26  Alexey Neyman  <stilor@att.net>
7163         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
7164         headers are no longer pulled in by <isl/val.h>.
7166 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
7168         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
7169         Use new helper functions.
7170         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
7171         Use new helper functions.
7172         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
7173         aarch_mm_needs_release): New declarations.
7174         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
7175         aarch_mm_needs_release): New.
7177 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
7179         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
7180         (arm32_output_mi_thunk): Deal with long calls.
7182 2018-09-26  Richard Biener  <rguenther@suse.de>
7184         PR debug/87428
7185         PR debug/87362
7186         * tree-inline.c (expand_call_inline): When the location
7187         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
7188         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
7189         the inserted BLOCK to make inlined_function_outer_scope_p
7190         recognize it.
7191         * dwarf2out.c (add_call_src_coords_attributes): Do not add
7192         coords for reserved locations.
7194 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
7196         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
7197         (*call_indirect_nonlocal_sysv<mode>): Ditto.
7198         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
7199         (*sibcall_nonlocal_sysv<mode>): Ditto.
7200         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
7201         (<bd>_<mode>): Ditto.
7202         (<bd>tf_<mode>): Ditto.
7204 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
7206         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
7207         control string as a list of templates instead of as C code.
7208         (*altivec_movti): Ditto.
7209         * config/rs6000/darwin.md (movdf_low_di): Ditto.
7211 2018-09-25  Jim Wilson  <jimw@sifive.com>
7213         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
7214         when target symbol is weak.
7216 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7218         PR c/87387
7219         * builtins.c (unterminated_array): Simplify.
7220         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
7221         where pointer arithmetic is safe.
7223 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
7225         PR target/86987
7226         * config/rs6000/altivec.md (altivec_vspltb): Use
7227         const_0_to_15_operand instead of u5bit_cint_operand.
7228         (*altivec_vspltb_internal): Ditto.
7229         (altivec_vspltb_direct): Ditto.
7230         (altivec_vsplth): Use const_0_to_7_operand instead of
7231         u5bit_cint_operand.
7232         (*altivec_vsplth_internal): Ditto.
7233         (altivec_vsplth_direct): Ditto.
7234         (altivec_vspltw): Use const_0_to_3_operand instead of
7235         u5bit_cint_operand.
7236         (*altivec_vspltw_internal): Ditto.
7237         (altivec_vspltw_direct): Ditto.
7238         (altivec_vspltsf): Ditto.
7239         (*altivec_vspltsf_internal): Ditto.
7240         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
7241         various splats with the proper size immediate.  Reorder the various
7242         cases by ascending size of immediate, and put all such together.
7244 2018-09-25  Richard Biener  <rguenther@suse.de>
7246         PR debug/83941
7247         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
7248         GC-ification.
7249         (maybe_create_die_with_external_ref): Do not create
7250         DW_TAG_imported_unit here.
7251         (add_abstract_origin_attribute): Handle external BLOCK refs.
7252         (dwarf2out_abstract_function): Simplify LTO case.
7253         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
7254         rather than using maybe_create_die_with_external_ref.
7256 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
7258         PR target/71278
7259         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
7260         (frndintxf2_mask_pm_i387): Ditto.
7261         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
7262         Enable for !flag_trapping_math.
7263         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
7264         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
7265         Change operand 1 predicate to nonimmediate_operand.
7266         (attr "i387_cw"): Remove mask_pm.
7267         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
7268         (enum ix86_entity): Remove I387_MASK_PM.
7269         * config/i386/i386.c (ix86_i387_mode_needed): Do not
7270         handle I387_MASK_PM.
7271         (ix86_mode_needed): Ditto.
7272         (ix86_mode_after): Ditto.
7273         (ix86_mode_entry): Ditto.
7274         (ix86_mode_exit): Ditto.
7275         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
7277 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
7279         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
7280         to_update_switch_stmts to vNULL instead of calling create on them
7281         immediately.
7283 2018-09-25  Richard Biener  <rguenther@suse.de>
7285         PR tree-optimization/87402
7286         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
7287         (visit_phi): Re-instantiate handling of supposed to be VARYING
7288         but non-VARYING backedge value.
7290 2018-09-25  Richard Biener  <rguenther@suse.de>
7292         PR debug/83941
7293         * dwarf2out.c (struct sym_off_pair): New.
7294         (external_die_map): New global.
7295         (lookup_decl_die): When in LTO create DIEs lazily from the
7296         external_die_map.
7297         (lookup_block_die): New function, create DIEs lazily in LTO.
7298         (equate_block_to_die): New function.
7299         (dwarf2out_die_ref_for_decl): During WPA get the association
7300         from the external DIE map.
7301         (dwarf2out_register_external_die): Record mapping into the
7302         external DIE map.
7303         (maybe_create_die_with_external_ref): New function split out from
7304         DIE generation part of old dwarf2out_register_external_die.
7305         (add_abstract_origin_attribute): Do not return the DIE.  When
7306         in LTO reference externals directly.
7307         (dwarf2out_abstract_function): When in LTO ignore calls for
7308         decls with external DIEs (already present abstract instances).
7309         (gen_call_site_die): Adjust.
7310         (add_high_low_attributes): Likewise.
7311         (gen_lexical_block_die): Likewise.
7312         (gen_inlined_subroutine_die): Likewie.
7313         (gen_block_die): Likewise.
7314         (dwarf2out_inline_entry): Likewise.
7315         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
7316         DIEs.
7318 2018-09-25  Martin Liska  <mliska@suse.cz>
7320         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
7321         integers and not by a float value.
7323 2018-09-25  Martin Liska  <mliska@suse.cz>
7325         PR fortran/87394
7326         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
7327         instead of NULL.
7328         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
7329         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
7330         * godump.c (go_format_type): Remove extra parenthesis.
7332 2018-09-25  Martin Liska  <mliska@suse.cz>
7334         * alias.c (set_dest_equal_p): Remove unused function.
7335         * config/i386/i386.c (def_builtin_pure2): Likewise.
7336         * diagnostic-show-locus.c (class layout): Remove
7337         unused field.
7338         (layout::layout): Likewise here.
7339         * dump-context.h (class temp_dump_context): Likewise.
7340         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
7341         (add_AT_loclistsptr): Likewise.
7342         (add_AT_offset): Likewise.
7343         (get_AT_hi_pc): Likewise.
7344         (is_comdat_die): Likewise.
7345         (type_is_enum): Likewise.
7346         (ceiling): Likewise.
7347         (add_AT_vms_delta): Likewise.
7348         (is_class_die): Likewise.
7349         * edit-context.c (class line_event): Remove unused field.
7350         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
7351         unused function.
7352         * ipa-cp.c (ipa_get_vr_lat): Likewise.
7353         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
7354         (ok_for_base_p_nonstrict): Likewise.
7355         * tree-chrec.c (is_not_constant_evolution): Likewise.
7356         (chrec_fold_poly_cst): Likewise.
7357         * tree-if-conv.c (has_pred_critical_p): Likewise.
7358         * tree-ssa-coalesce.c (print_exprs): Likewise.
7359         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
7360         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
7361         * tree-vrp.c (value_ranges_intersect_p): Likewise.
7362         (value_range_nonnegative_p): Likewise.
7364 2018-09-25  Martin Liska  <mliska@suse.cz>
7366         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
7367         Do not handle "GNU Pascal".
7368         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
7369         Likewise.
7370         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
7371         from documentation. Likewise.
7372         * dbxout.c (dbxout_range_type): Likewise.
7373         * doc/cpp.texi: Likewise.
7374         * doc/extend.texi: Likewise.
7375         * doc/frontends.texi: Likewise.
7376         * doc/invoke.texi: Remove Pascal entry.
7377         * tree.def (CLEANUP_POINT_EXPR): Likewise.
7378         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
7380 2018-09-25  Martin Liska  <mliska@suse.cz>
7382         PR middle-end/86078
7383         * doc/invoke.texi: Document all parameters and remove default
7384         of the parameters.
7386 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
7388         PR bootstrap/87417
7389         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
7390         contains HOST_WIDE_INTs when computing its size.
7392 2018-09-24  Jim Wilson  <jimw@sifive.com>
7394         PR target/87391
7395         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
7396         not TARGET_RVE.
7397         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
7399 2018-09-24  Andrew Pinski  <apinski@marvell.com>
7401         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
7402         access prev before checking it for NULLness in the
7403         AARCH64_FUSE_CMP_BRANCH case.
7405 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
7407         PR target/82699
7408         * config/i386/i386.c (rest_of_insert_endbranch): Set
7409         endbr_queued_at_entrance to true and don't insert ENDBR if
7410         x86_function_profiler will be called.
7411         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
7412         is true.
7413         * config/i386/i386.h (machine_function): Add
7414         endbr_queued_at_entrance.
7416 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
7418         * genattrtab.c (mk_attr_alt): Use alternative_mask.
7419         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
7420         types.
7421         (check_attr_test): Use alternative_mask.
7422         (get_attr_value): Likewise.
7423         (compute_alternative_mask): Use alternative_mask and XWINT.
7424         (make_alternative_compare): Use alternative_mask.
7425         (attr_alt_subset_p): Use XWINT.
7426         (attr_alt_subset_of_compl_p): Likewise.
7427         (attr_alt_intersection): Use alternative_mask and XWINT.
7428         (attr_alt_union): Likewise.
7429         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
7430         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
7431         (simplify_test_exp): Use alternative_mask and XWINT.
7432         (write_test_expr): Use alternative_mask and XWINT, adjust bit
7433         number calculation to support 64 bits.  Generate code that
7434         checks 64-bit masks.
7435         (main): Use alternative_mask.
7436         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
7438 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
7440         PR target/80080
7441         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
7442         RETURN+USE when returning via %r14.
7444 2018-09-24  Martin Liska  <mliska@suse.cz>
7446         * gcov.c (output_lines): Print colorization legend
7447         for both flag_use_colors and flag_use_hotness_colors.
7448         Reword the help.
7450 2018-09-24  Martin Liska  <mliska@suse.cz>
7452         * coverage.c (get_coverage_counts): Use warning_at
7453         with current_function_decl location. Use %qD in warning
7454         message.
7456 2018-09-24  Martin Liska  <mliska@suse.cz>
7458         * memory-block.h (memory_block_pool::release): Annotate with
7459         valgrind that the memory is not accessible.
7461 2018-09-24  Martin Liska  <mliska@suse.cz>
7463         PR sanitizer/85774
7464         * asan.c: Make asan_handled_variables extern.
7465         * asan.h: Likewise.
7466         * cfgexpand.c (expand_stack_vars): Make sure
7467         a representative is unpoison if another
7468         variable in the partition is handled by
7469         use-after-scope sanitization.
7471 2018-09-24  Richard Biener  <rguenther@suse.de>
7473         PR tree-optimization/63155
7474         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
7475         the worklist when the edge of the respective argument isn't
7476         executable.
7478 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
7480         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
7481         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
7482         (MASK_CLASS_P): Update for rename.
7483         (MAYBE_MASK_CLASS_P): Ditto.
7484         (REG_CLASS_NAMES): Update.
7485         (REG_CLASS_CONTENT): Update.
7486         * config/i386/i386.c (regclass_map): Update for MASK_REG
7487         and ALL_MASK_REGS rename.
7488         * config/i386/constraints.md (Yk): Update for rename.
7489         (k): Ditto.
7491 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
7493         * config/i386/i386.h (enum reg_class): Remove
7494         EVEX_SSE_REGS and MOD4_SSE_REGS.
7495         (REG_CLASS_NAMES): Update.
7496         (REG_CLASS_CONTENT): Update.
7497         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
7498         registers as ALL_SSE_REGS.
7499         (ix86_additional_allocno_class_p): Remove.
7500         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
7501         (ix86_register_priority): Lower priority of EVEX SSE registers.
7502         Use IN_RANGE macro where appropriate.
7503         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
7504         AVX-5124VNNIW checks.
7505         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
7506         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
7507         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
7508         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
7509         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
7510         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
7511         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
7512         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
7513         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
7514         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
7515         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
7516         * config/i386/constraints.md (Yh): Remove.
7518 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
7520         * config/i386/i386.c (regclass_map): Declare integer REX registers
7521         as GENERAL_REGS.
7523 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
7525         * doc/service.texi (Service): Switch the fsf.org link to https.
7527 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
7529         PR target/86798
7530         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7531         Define to speculation_safe_value_not_needed.
7533 2018-09-21  Florian Weimer  <fweimer@redhat.com>
7535         PR middle-end/81035
7536         * doc/extend.texi (Common Function Attributes): Mention that
7537         noreturn suppresses tail call optimization.
7539 2018-09-21  Jeff Law  <law@redhat.com>
7541         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
7542         vr_values::cleanup_edges_and_switches.
7543         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
7544         vr_values class.
7545         (identify_jump_threads): Remove EDGE_IGNORE handling.
7546         (execute_vrp): Move handling of to_remove_edges and
7547         to_update_switch_stmts into vr_values class member functions.
7548         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
7549         (to_update_switch_stmts): Likewise.
7550         * vr-values.c: Include cfghooks.h.
7551         (vr_values::vr_values): Initialize to_remove_edges and
7552         to_update_switch_stmts.
7553         (vr_values::~vr_values): Verify to_remove_edges and
7554         to_update_switch_stmts are empty.
7555         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
7556         (vr_values::cleanup_edges_and_switches): New member function.
7557         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
7558         function.  Add new data members.
7560 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
7562         PR tree-optimization/87309
7563         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
7564         calls with pflags and alt_flags.
7565         (selftest::test_capture_of_dump_calls): Add test of interaction of
7566         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
7568 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7570         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
7572 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7574         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
7576 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7578         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
7579         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
7580         (VXWORKS_LIBS_RTP): Minor reordering.
7582 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7584         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
7585         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
7587 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7589         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
7590         (PTRDIFF_TYPE): Likewise.
7592 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7594         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
7595         triplet, similar to support for VxWorks7.
7596         * config/vxworks-dummy.h: Provide a default definition
7597         of TARGET_VXWORKS64 to 0.
7599 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7601         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
7602         * config/vxworks-dummy.h: here.
7604 2018-09-21  Olivier Hainque  <hainque@adacore.com>
7606         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
7608 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
7609             Bo Zhou  <zbo.zhou@hisilicon.com>
7611         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
7612         * config/aarch64/aarch64-tune.md: Regenerated.
7613         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
7614         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
7615         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
7617 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
7618             Julian Brown  <julian@codesourcery.com>
7620         * builtins.c (get_builtin_sync_mem): Handle address spaces.
7622 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
7624         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
7625         if the call takes a static chain.
7627 2018-09-21  Martin Liska  <mliska@suse.cz>
7629         * auto-profile.c (autofdo_source_profile::read): Do not
7630         set sum_all.
7631         (read_profile): Do not add working sets.
7632         (read_autofdo_file): Remove sum_all.
7633         (afdo_callsite_hot_enough_for_early_inline): Remove const
7634         qualifier.
7635         * coverage.c (struct counts_entry): Remove gcov_summary.
7636         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
7637         do not support GCOV_TAG_PROGRAM_SUMMARY.
7638         (get_coverage_counts): Remove summary and expected
7639         arguments.
7640         * coverage.h (get_coverage_counts): Likewise.
7641         * doc/gcov-dump.texi: Remove -w option.
7642         * gcov-dump.c (dump_working_sets): Remove.
7643         (main): Do not support '-w' option.
7644         (print_usage): Likewise.
7645         (tag_summary): Likewise.
7646         * gcov-io.c (gcov_write_summary): Do not dump
7647         histogram.
7648         (gcov_read_summary): Likewise.
7649         (gcov_histo_index): Remove.
7650         (gcov_histogram_merge): Likewise.
7651         (compute_working_sets): Likewise.
7652         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
7653         it not obsolete.
7654         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
7655         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
7656         (GCOV_HISTOGRAM_SIZE): Remove.
7657         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
7658         (struct gcov_summary): Simplify rapidly just
7659         to runs and sum_max fields.
7660         (gcov_histo_index): Remove.
7661         (NUM_GCOV_WORKING_SETS): Likewise.
7662         (compute_working_sets): Likewise.
7663         * gcov-tool.c (print_overlap_usage_message): Remove
7664         trailing empty line.
7665         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
7666         (output_lines): Remove program related line.
7667         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
7668         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
7669         histogram.
7670         (input_profile_summary): Do not read it.
7671         (merge_profile_summaries): And do not merge it.
7672         (input_symtab): Do not call removed function.
7673         * modulo-sched.c (sms_schedule): Do not print sum_max.
7674         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
7675         removed when histogram method was invented.
7676         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
7677         mode.
7678         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
7679         GCOV coding style.
7680         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
7681         and dump selected value.
7682         * profile.c (add_working_set): Remove.
7683         (get_working_sets): Likewise.
7684         (find_working_set): Likewise.
7685         (get_exec_counts): Do not work with working sets.
7686         (read_profile_edge_counts): Do not inform as sum_max is removed.
7687         (compute_branch_probabilities): Likewise.
7688         (compute_value_histograms): Remove argument for call of
7689         get_coverage_counts.
7690         * profile.h: Do not make gcov_summary const.
7692 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
7694         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
7696 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
7698         PR tree-optimization/86990
7699         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
7700         Check that the entire merged store group is made of constants only for
7701         overlapping stores.
7703 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
7705         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
7706         (VTABLE_VERIFICATION_SPEC): Likewise.
7707         (SANITIZER_EARLY_SPEC): Likewise.
7708         (SANITIZER_SPEC): Likewise.
7709         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
7710         * doc/invoke.texi (Link Options): Document -r.
7712 2018-09-20  Richard Biener <rguenther@suse.de>
7714         PR middle-end/87054
7715         * gimplify.c (gimplify_expr): Retain alignment of
7716         addressable lvalue in dereference.
7718 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
7720         PR bootstrap/87013
7721         * configure.ac: Check for .loc is_stmt support.
7722         * configure, config.in: Rebuilt.
7723         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
7724         if not supported.
7726 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
7728         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
7729         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
7730         -misel=no.
7732 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
7734         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
7735         VECTOR_OTHER.
7736         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
7737         case VECTOR_OTHER.
7739 2018-09-20  Marek Polacek  <polacek@redhat.com>
7741         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
7743 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
7745         PR tree-optimization/87288
7746         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
7747         into account when determining PEELING_FOR_NITERS.
7749 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
7751         PR tree-optimization/86877
7752         * tree-vect-loop.c (vect_analyze_loop_2): Call
7753         vect_verify_datarefs_alignment.
7755 2018-09-19  Marek Polacek  <polacek@redhat.com>
7757         * doc/invoke.texi: Document -Wclass-conversion.
7759 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
7761         * config/pa/pa.c (pa_adjust_priority): Delete.
7762         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
7764         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
7765         (atomic_storehi): Likewise.
7766         (atomic_storesi): Likewise.
7767         (atomic_loaddi): Restore compare and swap exchange loop code.
7769 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
7771         PR rtl-optimization/86902
7772         * combine.c (try_combine): When changing the CC mode used, don't change
7773         an unrelated mode in other_insn to that new CC mode.
7775 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
7777         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
7778         with %T in place of calls to dump_generic_expr.
7779         (prune_runtime_alias_test_list): Likewise.
7780         (create_runtime_alias_checks): Likewise.
7781         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
7782         (vect_analyze_data_ref_dependence): Likewise.
7783         (vect_slp_analyze_data_ref_dependence): Likewise.
7784         (vect_record_base_alignment): Likewise.  Use %G in place of call
7785         to dump_gimple_stmt.
7786         (vect_compute_data_ref_alignment): Likewise.
7787         (verify_data_ref_alignment): Likewise.
7788         (vect_find_same_alignment_drs): Likewise.
7789         (vect_analyze_group_access_1): Likewise.
7790         (vect_analyze_data_ref_accesses): Likewise.
7791         (dependence_distance_ge_vf): Likewise.
7792         (dump_lower_bound): Likewise.
7793         (vect_prune_runtime_alias_test_list): Likewise.
7794         (vect_find_stmt_data_reference): Likewise.
7795         (vect_analyze_data_refs): Likewise.
7796         (vect_create_addr_base_for_vector_ref): Likewise.
7797         (vect_create_data_ref_ptr): Likewise.
7798         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
7799         (vect_can_advance_ivs_p): Likewise.
7800         (vect_update_ivs_after_vectorizer): Likewise.
7801         (vect_gen_prolog_loop_niters): Likewise.
7802         (vect_prepare_for_masked_peels): Likewise.
7803         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
7804         (vect_determine_vectorization_factor): Likewise.
7805         (vect_is_simple_iv_evolution): Likewise.
7806         (vect_analyze_scalar_cycles_1): Likewise.
7807         (vect_analyze_loop_operations): Likewise.
7808         (report_vect_op): Likewise.
7809         (vect_is_slp_reduction): Likewise.
7810         (check_reduction_path): Likewise.
7811         (vect_is_simple_reduction): Likewise.
7812         (vect_create_epilog_for_reduction): Likewise.
7813         (vect_finalize_reduction:): Likewise.
7814         (vectorizable_induction): Likewise.
7815         (vect_transform_loop_stmt): Likewise.
7816         (vect_transform_loop): Likewise.
7817         (optimize_mask_stores): Likewise.
7818         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
7819         (vect_split_statement): Likewise.
7820         (vect_recog_over_widening_pattern): Likewise.
7821         (vect_recog_average_pattern): Likewise.
7822         (vect_determine_min_output_precision_1): Likewise.
7823         (vect_determine_precisions_from_range): Likewise.
7824         (vect_determine_precisions_from_users): Likewise.
7825         (vect_mark_pattern_stmts): Likewise.
7826         (vect_pattern_recog_1): Likewise.
7827         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7828         (vect_record_max_nunits): Likewise.
7829         (vect_build_slp_tree_1): Likewise.
7830         (vect_build_slp_tree_2): Likewise.
7831         (vect_print_slp_tree): Likewise.
7832         (vect_analyze_slp_instance): Likewise.
7833         (vect_detect_hybrid_slp_stmts): Likewise.
7834         (vect_detect_hybrid_slp_1): Likewise.
7835         (vect_slp_analyze_operations): Likewise.
7836         (vect_slp_analyze_bb_1): Likewise.
7837         (vect_transform_slp_perm_load): Likewise.
7838         (vect_schedule_slp_instance): Likewise.
7839         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
7840         (vect_mark_stmts_to_be_vectorized): Likewise.
7841         (vect_init_vector_1): Likewise.
7842         (vect_get_vec_def_for_operand): Likewise.
7843         (vect_finish_stmt_generation_1): Likewise.
7844         (vect_check_load_store_mask): Likewise.
7845         (vectorizable_call): Likewise.
7846         (vectorizable_conversion): Likewise.
7847         (vectorizable_operation): Likewise.
7848         (vectorizable_load): Likewise.
7849         (vect_analyze_stmt): Likewise.
7850         (vect_is_simple_use): Likewise.
7851         (vect_get_vector_types_for_stmt): Likewise.
7852         (vect_get_mask_type_for_stmt): Likewise.
7853         * tree-vectorizer.c (increase_alignment): Likewise.
7855 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
7857         * doc/rtl.texi: Adjust vec_select description.
7858         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
7859         non-constant selectors.
7861 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
7863         * config/aarch64/aarch64-protos.h
7864         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
7865         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
7866         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
7867         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
7868         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
7869         (AARCH64_FL_PROFILE): Move index so flags are ordered.
7870         (AARCH64_ISA_RCPC8_4): New flag.
7871         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
7872         to aarch64_offset_9bit_signed_unscaled_p.
7873         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
7874         and use stlur.
7875         * config/aarch64/constraints.md (Ust): New constraint.
7876         * config/aarch64/predicates.md.
7877         (aarch64_9bit_offset_memory_operand): New predicate.
7878         (aarch64_rcpc_memory_operand): New predicate.
7880 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
7882         PR rtl-optimization/87361
7883         * rtlanal.c (nonzero_bits1): Revert accidental change.
7885 2018-09-19  Richard Biener  <rguenther@suse.de>
7887         PR tree-optimization/87349
7888         PR tree-optimization/87342
7889         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
7891 2018-09-18  Marek Polacek  <polacek@redhat.com>
7893         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
7894         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
7896 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
7898         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
7900 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
7902         PR rtl-optimization/86882
7903         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
7905 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
7907         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
7908         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
7910 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
7912         PR other/87353
7913         * doc/invoke.texi (Link Options): Fix formatting and grammar.
7915 2018-09-18  Richard Biener  <rguenther@suse.de>
7917         PR middle-end/63155
7918         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
7919         (compute_samebase_partition_bases): Likewise.
7920         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
7921         (gimple_can_coalesce_p): Simplify.
7923 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
7925         Handle a library implementation of ffs calling __builtin_ffs.
7926         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
7927         (mmix_init_libfuncs): New function: make __builtin_ffs expand
7928         to __ffsdi2.
7930 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
7932         * diagnostic-show-locus.c (class layout_range): Add field
7933         "m_original_idx".
7934         (layout_range::layout_range): Add "original_idx" param and use it
7935         to initialize new field.
7936         (make_range): Use 0 for original_idx.
7937         (layout::layout): Pass in index to calls to
7938         maybe_add_location_range.
7939         (layout::maybe_add_location_range): Add param "original_idx" and
7940         pass it on to layout_range.
7941         (layout::print_any_labels): Pass on range->m_original_idx to
7942         get_text call.
7943         (gcc_rich_location::add_location_if_nearby): Use 0 for
7944         original_idx.
7945         * gcc-rich-location.h (text_range_label::get_text): Update for new
7946         param.
7947         (range_label_for_type_mismatch::get_text): Likewise.
7949 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
7951         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
7953 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
7955         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
7956         format_string_diagnostic_t.
7957         (fmtwarn_n): Likewise.
7958         * substring-locations.c
7959         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
7960         (format_warning_n_va): Convert to...
7961         (format_string_diagnostic_t::emit_warning_n_va): ...this.
7962         (format_warning_va): Convert to...
7963         (format_string_diagnostic_t::emit_warning_va): ...this.
7964         (format_warning_at_substring): Convert to...
7965         (format_string_diagnostic_t::emit_warning): ...this.
7966         (format_warning_at_substring_n): Convert to...
7967         (format_string_diagnostic_t::emit_warning_n): ...this.
7968         * substring-locations.h (class format_string_diagnostic_t): New
7969         class.
7970         (format_warning_va): Convert to
7971         format_string_diagnostic_t::emit_warning_va.
7972         (format_warning_n_va): Convert to
7973         format_string_diagnostic_t::emit_warning_n_va.
7974         (format_warning_at_substring): Convert to
7975         format_string_diagnostic_t::emit_warning.
7976         (format_warning_at_substring_n): Convert to
7977         format_string_diagnostic_t::emit_warning_n.
7979 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
7980             Bernd Schmidt <bernds_cb1@t-online.de>
7982         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
7983         SImode args.
7985 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
7987         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
7988         operand 0 predicate to nonimmediate operand.
7989         (rint<mode>2_frndint): Remove insn pattern.
7990         (rint<mode>2): Change operand 1 predicate to general_operand.
7991         Extend operand 1 to XFmode and generate rintxf2 insn.
7992         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
7993         Do not use X87MODEF mode macro.
7994         (frndintxf2_<rounding>_i387): Rename from
7995         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
7996         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
7997         to XFmode and generate significandxf3 insn.
7999 2018-09-17  Richard Biener  <rguenther@suse.de>
8001         PR tree-optimization/87328
8002         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
8003         visiting unexecutable backedges when not iterating.
8004         (do_rpo_vn): Mark all edges not executable even when not
8005         iterating.
8007 2018-09-17  Martin Jambor  <mjambor@suse.cz>
8009         PR c/63886
8010         * doc/invoke.texi (Warning Options): Likewise.
8012 2018-09-17  Richard Biener  <rguenther@suse.de>
8014         PR tree-optimization/87301
8015         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
8016         clean EH info from leftover copy assignments.
8018 2018-09-17  Martin Liska  <mliska@suse.cz>
8020         PR gcov-profile/85871
8021         * gcov.c (output_intermediate_file): Fix out of bounds
8022         access.
8024 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
8026         * config/arc/arc.c: Object attributes for core4 not reflected
8027         correctly.
8028         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
8029         core3).
8031 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
8033         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
8035 2018-09-17  Martin Liska  <mliska@suse.cz>
8037         * doc/gcov.texi: Document new option --use-hotness-colors.
8038         * gcov.c (struct source_info): Declare new field.
8039         (source_info::source_info): Set default for maximum_count.
8040         (print_usage): Add new -q option.
8041         (process_args): Process it.
8042         (accumulate_line_info): Save src->maximum_count.
8043         (output_line_beginning): Make color line number if
8044         flag_use_hotness_colors is set.
8045         (output_line_details): Pass default argument value.
8046         (output_lines): Pass src->maximum_count.
8048 2018-09-17  Martin Liska  <mliska@suse.cz>
8050         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
8051         Use processor_names table.
8052         * config/i386/i386.c (ix86_default_align): Use
8053         processor_cost_table for alignment values.
8054         (ix86_option_override_internal): Use processor_names.
8055         (ix86_function_specific_print): Likewise.
8056         * config/i386/i386.h (struct processor_costs):
8057         Add alignment values.
8058         (struct ptt): Remove and replace with const char *.
8059         * config/i386/x86-tune-costs.h (struct processor_costs):
8060         Declare default alignments for all costs.
8062 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
8064         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
8065         symbolics or VR_VARYING ranges for ABS_EXPR.
8066         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
8067         when range will wrap.
8069 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
8071         PR middle-end/86864
8072         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
8073         before and after a JUMP_TABLE_DATA.
8075 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
8077         PR middle-end/87188
8078         * dojump.c (do_compare_and_jump): Canonicalize function pointers
8079         when one operand is a function pointer.  Use POINTER_TYPE_P and
8080         FUNC_OR_METHOD_TYPE_P.
8081         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
8082         * fold-const.c (build_range_check): Likewise.
8083         * match.pd (simple_comparison): Likewise.
8085 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
8087         PR c/82967
8088         * spellcheck.c (get_edit_distance_cutoff): New function.
8089         (selftest::test_edit_distance_unit_test_oneway): Rename to...
8090         (selftest::test_get_edit_distance_one_way): ...this.
8091         (selftest::test_get_edit_distance_unit): Rename to...
8092         (selftest::test_get_edit_distance_both_ways): ...this.
8093         (selftest::test_edit_distances): Move tests to this new function,
8094         and test some more pairs of strings.  Update for above renaming.
8095         (selftest::get_old_cutoff): New function.
8096         (selftest::test_get_edit_distance_cutoff): New function.
8097         (selftest::assert_suggested_for): New function.
8098         (ASSERT_SUGGESTED_FOR): New macro.
8099         (selftest::assert_not_suggested_for): New function.
8100         (ASSERT_NOT_SUGGESTED_FOR): New macro.
8101         (selftest::test_suggestions): New function.
8102         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
8103         tests to selftest::test_edit_distances and call it.  Add calls to
8104         selftest::test_get_edit_distance_cutoff and
8105         selftest::test_suggestions.
8106         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
8107         (best_match::consider): Replace hard-coded cutoff calculation with
8108         a call to...
8109         (best_match::get_cutoff): New declaration.
8110         (best_match::get_best_meaningful_candidate): Likewise.
8112 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8114         * builtins.c (fold_builtin_strlen): Remove TODO comment.
8116 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8118         revert:
8119         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8121         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
8122         terminated string literal.
8124 2018-09-14  Martin Sebor  <msebor@redhat.com>
8126         * builtins.c (unterminated_array): Handle ARRAY_REF.
8127         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
8128         * builtins.h (unterminated_array): Declare extern.
8129         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
8130         arrays.
8131         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
8132         calls.
8134 2018-09-14  Martin Sebor  <msebor@redhat.com>
8135             Jeff Law  <law@redhat.com>
8137         * builtins.c (unterminated_array): New.
8138         (expand_builtin_strcpy): Adjust.
8139         (expand_builtin_strcpy_args): Detect unterminated arrays.
8140         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
8141         unterminated arrays.
8142         * gimple-fold.h (get_maxval_strlen): Add argument.
8143         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
8145         * gimple-fold.c (get_range_strlen): Add argument.
8146         (get_maxval_strlen): Adjust.
8147         * gimple-fold.h (get_range_strlen): Add argument.
8149 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
8151         * config/i386/movdirintrin.h: Fix copyright year.
8153 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
8155         * reg-stack.c: Include regs.h.
8156         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
8157         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
8158         FIRST_STACK_REG, not DFmode.
8159         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
8160         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
8161         (change stack): Default register mode to the reg_raw_mode of
8162         FIRST_STACK_REG, not DFmode.
8163         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
8164         (*swapxf): Rename from swapxf.
8166 2018-09-14  Carl Love  <cel@us.ibm.com>
8168         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
8169         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
8171 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
8173         PR target/87224
8174         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
8175         alternatives.
8177 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
8179         PR target/85628
8180         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
8182 2018-09-14  Jason Merrill  <jason@redhat.com>
8184         Fix --enable-gather-detailed-mem-stats.
8185         * hash-table.c (hash_table_usage): Change from variable to function.
8186         * hash-table.h: Adjust.
8187         * Makefile.in: Add missing dependencies on hash-table.h.
8189 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8191         PR tree-optimization/87259
8192         PR lto/87283
8193         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
8194         execute_cse_reciprocals_1 has tried transforming.
8196 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
8198         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
8199         VR_VARYING for PLUS/MINUS_EXPR.
8201 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
8203         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
8204         formatting.
8206 2018-09-14  Richard Biener  <rguenther@suse.de>
8208         PR middle-end/63155
8209         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
8210         bits for the merged partition.
8212 2018-09-13  Martin Sebor  <msebor@redhat.com>
8213             Bernd Edlinger  <bernd.edlinger@hotmail.de>
8215         * builtins.h (c_srlen): Add argument.
8216         * builtins.c (warn_string_no_nul): New function.
8217         (c_strlen): Add argument and use it.  Update recursive calls.
8218         Pass DECL argument to string_constant to get info on non
8219         terminated strings.  Update *NONSTR as needed.
8220         (fold_builtin_strlen): Add argument to calls to c_strlen.
8221         Warn for unterminated arrays.
8222         (warn_string_no_null): Add prototype.
8223         * expr.c (string_constant): Update arguments.  Update recursive
8224         calls appropriately.  Detect missing NUL terminator and outermost
8225         declaration its missing in.
8226         Improve checks for arrays with nonzero lower bound or elements
8227         that are not a single byte.  Simplify offset computation.
8228         Simplify checks for non-NUL terminated strings.
8229         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
8230         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
8232 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8234         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
8235         correctly.
8236         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
8237         argument.  Fix range checks.
8238         * fold-const.h (c_getstr): Adjust protoype.
8239         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
8240         string is constant but contains no NUL byte.
8242         * expr.c (string_constant): Adjust function comment.
8243         Remove bogus check for zero termination.
8245         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
8247         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
8248         (get_constant_size): Don't make STRING_CSTs larger than they are.
8249         (check_string_literal): New check function for STRING_CSTs.
8250         (output_constant): Use it.
8252 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
8254         PR target/86812
8255         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
8257 2018-09-13  Richard Biener  <rguenther@suse.de>
8259         PR tree-optimization/87263
8260         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
8261         (struct unwind_state): Add max_rpo field.
8262         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
8263         Compute max_rpo, the max RPO number a block can be backwards reached
8264         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
8265         separating it from the iterating mode.
8267 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
8269         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
8270         (rfs_decision): New scheduling decision.
8272 2018-09-13  Richard Biener  <rguenther@suse.de>
8274         PR bootstrap/87134
8275         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
8276         (vn_nary_op_insert_pieces_predicated): Do not write useless
8277         valid_dominated_by_p entry outside of the allocated storage.
8279 2018-09-13  Omar Sandoval  <osandov@osandov.com>
8280             Tom de Vries  <tdevries@suse.de>
8282         PR debug/86985
8283         * dwarf2out.c (is_c): New function.
8284         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
8286 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
8288         PR target/85628
8289         * config/aarch64/aarch64.md (*aarch64_bfxil):
8290         Define.
8291         * config/aarch64/constraints.md (Ulc): Define.
8292         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
8293         Define.
8294         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
8295         New function.
8297 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
8299         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
8300         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
8301         aarch64_layout_frame call.
8302         (aarch64_expand_epilogue): Likewise.
8303         (aarch64_initial_elimination_offset): Likewise.
8304         (aarch64_get_separate_components): Likewise.
8305         (aarch64_use_return_insn_p): Likewise.
8306         (aarch64_layout_frame): Remove unneeded check.
8308 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
8310         * configure.ac: Only append
8311         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
8312         gcc_config_arguments if it was never reconfigured or last reconfigure
8313         was with different arguments.
8314         * configure: Regenerated.
8316 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
8317             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8319         PR middle-end/87290
8320         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
8321         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
8323 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
8325         PR tree-optimization/87287
8326         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
8327         X % C == 0 to X % (unsigned) C == 0 optimization to ...
8328         * match.pd (X % C == 0): ... here.  New optimization.
8330 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
8332         PR middle-end/82853
8333         * expr.h (maybe_optimize_mod_cmp): Declare.
8334         * expr.c (mod_inv): New function.
8335         (maybe_optimize_mod_cmp): New function.
8336         (do_store_flag): Use it.
8337         * cfgexpand.c (expand_gimple_cond): Likewise.
8339 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
8340             Julian Brown  <julian@codesourcery.com>
8342         PR middle-end/86336
8343         * gimplify.c (gimplify_scan_omp_clauses): Set
8344         target_firstprivatize_array_bases in OpenACC parallel and kernels
8345         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
8346         OpenACC data regions.
8348 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
8350         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
8351         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
8353 2018-09-12  Richard Biener  <rguenther@suse.de>
8355         PR tree-optimization/87280
8356         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
8357         edge but unreachable target.
8358         (do_rpo_vn): For conservatively handling a PHI only mark
8359         the backedge executable but not the block reachable.
8361 2018-09-12  Richard Biener  <rguenther@suse.de>
8363         PR tree-optimization/87266
8364         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
8365         visited blocks.
8367 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
8369         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
8370         constants.
8371         ("trunc<BFP:mode><DFP_ALL:mode>2")
8372         ("trunc<DFP_ALL:mode><BFP:mode>2")
8373         ("extend<BFP:mode><DFP_ALL:mode>2")
8374         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
8375         according to the target operand type.
8377 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
8378             Andreas Krebbel  <krebbel@linux.ibm.com>
8380         PR tree-optimization/86844
8381         * gimple-ssa-store-merging.c
8382         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
8383         there are any overlapping stores in between them, make sure they are
8384         also coalesced or we give up completely.
8386 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
8388         PR middle-end/87248
8389         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
8390         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
8392 2018-09-12  Tom de Vries  <tdevries@suse.de>
8394         * common.opt (gdescribe-dies): Add option.
8395         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
8396         attribute for artifical and nameless decls.
8397         (dwarf2out_register_external_die): Add description attribute to
8398         external reference die.
8399         (add_desc_attribute): New functions.
8400         (gen_subprogram_die): Add description attribute to
8401         DW_TAG_call_site_parameter.
8402         * tree-pretty-print.c (print_generic_expr_to_str): New function.
8403         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
8404         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
8405         -gno-describe-dies.
8406         (@item -gdescribe-dies): Add.
8408 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
8410         * tree-vrp.c (vrp_shift_undefined_p): Remove.
8411         (extract_range_from_binary_expr_1: Call
8412         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
8413         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
8414         depend on sign.
8416 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
8418         * gimple-ssa-warn-alloca.c
8419         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
8420         field for ALLOCA_BOUND_*_LARGE.
8422 2018-09-11  Nathan Sidwell  <nathan@acm.org>
8424         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
8426 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
8428         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
8429         for clobbers.  Remove obsolete comment.
8431 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
8433         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
8434         mpxchk, mpxld and mpxst types.
8435         (define_attr length_immediate): Remove all processing of mpx types.
8436         (define_attr prefix_0f): Ditto.
8437         (define_attr memory): Ditto.
8439 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
8441         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
8442         (log<mode>2): Change operand 1 predicate to general_operand.
8443         Extend operand 1 to XFmode and generate logxf3 insn.
8444         (log10<mode>2): Change operand 1 predicate to general_operand.
8445         Extend operand 1 to XFmode and generate log10xf3 insn.
8446         (log2<mode>2): Change operand 1 predicate to general_operand.
8447         Extend operand 1 to XFmode and generate log2xf3 insn.
8448         (fyl2xp1_extend<mode>xf3_i387): Remove.
8449         (log1p<mode>2): Change operand 1 predicate to general_operand.
8450         Extend operand 1 to XFmode and generate log1pxf3 insn.
8451         (fxtract_extend<mode>xf3_i387): Remove.
8452         (logb<mode>2): Change operand 1 predicate to general_operand.
8453         Extend operand 1 to XFmode and generate logbxf3 insn.
8454         (ilogb<mode>2): Change operand 1 predicate to general_operand.
8455         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
8456         (significand<mode>2): Change operand 1 predicate to general_operand.
8457         Extend operand 1 to XFmode and generate significandxf3 insn.
8459 2018-09-11  Nathan Sidwell  <nathan@acm.org>
8461         * gcc.c (perror_with_name, pfatal_with_name): Delete.
8462         (load_specs): Use fatal_error.
8463         (DELETE_IF_ORDINARY, process_command): Use error.
8464         (execute, run_attempt): Use fatal_error.
8466 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
8468         * diagnostic-core.h (sorry_at): New prototype.
8469         * diagnostic.c (sorry_at): New function.
8471 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
8473         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
8474         by zero as VR_UNDEFINED.
8476 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
8478         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
8479         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
8480         (<sincos>mode2): New expander.
8481         (sincos_extend<mode>xf3_i387): Remove insn pattern.
8482         (sincos -> sin, cos splitters): Remove splitter patterns.
8483         (sincos<mode>3): Change operand 2 predicate to general_operand.
8484         Extend operand 2 to XFmode and generate sincosxf3 insn.
8485         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
8486         Change operand 3 predicate to const1_operand.
8487         (fptan_extend<mode>xf4_i387): Remove insn pattern.
8488         (tanxf2): Update operands in the call to fptanxf4_i387.
8489         (tan<mode>2): Change operand 1 predicate to general_operand.
8490         Extend operand 1 to XFmode and generate tanxf3 insn.
8491         (atan2xf3): Rename from *fpatanxf3_i387.
8492         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
8493         (atan2xf3): Remove expander.
8494         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
8495         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
8496         (atan<mode>2): Change operand 1 predicate to general_operand.
8497         Extend operand 1 to XFmode and generate atanxf3 insn.
8499 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
8501         * config/i386/i386.md (x87/SSE constant load splitter): Use
8502         memory_operand instead of nonimmediate_operand for input operand
8503         predicate.
8505 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
8507         * config/i386/i386.md (float partial SSE register stall splitter): Move
8508         splitter near its instruction pattern.
8509         (float_extend partial SSE register stall splitter): Ditto.
8510         (float_truncate partial SSE register stall splitter): Ditto.
8512 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
8514         PR target/86794
8515         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8516         to speculation_safe_value_not_needed.
8518         PR target/85666
8519         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
8520         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
8521         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
8522         leaf_function_p, instead use has_hard_reg_initial_val.
8524 2018-09-09  Nathan Sidwell  <nathan@acm.org>
8526         * gcc.h (pfatal_with_name): Don't declare here.
8527         * gcc.c (pfatal_with_name): Make static.
8529 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
8531         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
8532         earlyclobber.
8534 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
8536         PR rtl-optimization/85458
8537         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
8538         priority hook to reduce the priority of EXPR.
8540 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
8542         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
8543         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
8544         with X87_ENABLE_FLOAT.  Remove preparation code.
8545         (*float<SWI48:mode><MODEF:mode>2): Rename from
8546         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
8547         math using "enabled" attribute.
8548         (*floatdi<MODEF:mode>2_i387): Rename from
8549         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
8550         enable for 32bit targets only.
8551         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
8552         splitter.
8553         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
8554         as operand 1 predicate.  Rewrite as define_insn_and_split.
8555         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
8557 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
8559         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
8560         to fallthru to FLOAT case.
8562 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
8564         PR target/86731
8565         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
8566         around folding of vec_sl to handle out of range shift values.
8568 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
8570         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
8571         Update callers to gen_fix_trunc<mode>_i387_fisttp
8572         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
8573         nonimmediate_operand.
8574         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
8575         and corresponding splitters.
8576         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
8577         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
8578         (fix_truncdi_i387_with_temp): Remove insn pattern
8579         and corresponding splitters.
8580         (fix_trunc<mode>_i387): Change operand 0 predicate to
8581         nonimmediate_operand.
8582         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
8583         and corresponding splitters.
8584         (*fistdi2_1): Remove.
8585         (fistdi2): Ditto.
8586         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
8587         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
8588         (*fist<mode>2_1): Remove.
8589         (fist<mode>2): Ditto.
8590         (fist<mode>2_with_temp): Remove insn pattern and corresponding
8591         splitters.
8592         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
8593         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
8594         (fistdi2_<rounding>): Change operand 0 predicate to
8595         nonimmediate_operand.
8596         (fistdi2_<rounding>_with_temp): Remove insn pattern
8597         and corresponding splitters.
8598         (fist<mode>2_<rounding>): Change operand 0 predicate to
8599         nonimmediate_operand.
8600         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
8601         and corresponding splitters.
8603         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
8605 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8607         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
8608         the init value.
8610 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
8612         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
8613         early gimple folding of vec_splat().
8614         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
8615         * gimple-fold.h: Add an extern define for tree_vec_extract().
8617 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
8619         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
8620         wrappers around TREE_TYPE comparisons.
8622 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
8624         PR target/80080
8625         * config/s390/predicates.md: Add nonsym_memory_operand.
8626         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
8627         contains a SYMBOL_REF, load it into an intermediate pseudo.
8628         (s390_emit_compare_and_swap): Legitimize operand.
8629         * config/s390/s390.md: Use the new nonsym_memory_operand
8630         with UNSPECV_CAS patterns.
8632 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
8634         PR target/80080
8635         * config/s390/s390-passes.def: New file.
8636         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
8637         declaration.
8638         (make_pass_s390_early_mach): Add declaration.
8639         * config/s390/s390.c (make_pass_s390_early_mach):
8640         (s390_option_override): Remove dynamic registration.
8641         * config/s390/t-s390: Add s390-passes.def.
8643 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
8645         * config/s390/s390.c (s390_decompose_constant_pool_ref):
8646         Remove UNSPEC_LTREL_BASE check.
8647         (annotate_constant_pool_refs): Likewise.
8648         (find_constant_pool_ref): Likewise.
8649         (find_ltrel_base): Removed.
8650         (replace_ltrel_base): Removed.
8651         (s390_mainpool_finish): Remove replace_ltrel_base call.
8652         (s390_chunkify_start): Remove pending LTREL_BASE logic.
8653         (s390_chunkify_finish): Remove replace_ltrel_base call.
8654         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
8656 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
8658         PR target/86779
8659         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8660         to speculation_safe_value_not_needed.
8662 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
8663             Bernd Schmidt  <bernds_cb1@t-online.de>
8665         * config/nvptx/nvptx-opts.h: New file.
8666         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
8667         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
8668         (ASM_SPEC): Define.
8669         (TARGET_SM35): New macro.
8670         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
8671         correct predicate.
8672         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
8673         values.
8674         (misa=): New option.
8675         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
8677 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
8679         * config/i386/i386.md (truncdfsf2): Remove expander.
8680         (truncdfsf2_with_temp): Ditto.
8681         (truncxf<mode>2): Ditto.
8682         (*truncdfsf_fast_mixed): Remove insn pattern.
8683         (*truncdfsf_fast_i387): Ditto.
8684         (*truncdfsf_mixed): Ditto.
8685         (*truncdfsf_i387): Ditto.
8686         (*truncdfsf2_i387_1): Ditto.
8687         (*truncxfsf2_mixed): Ditto.
8688         (*truncxfdf2_mixed): Ditto.
8689         (*truncxf<mode>2_i387_noop): Ditto. Update callers
8690         to call gen_truncxf<mode>2 instead.
8691         (*truncxf<mode>2_i387): Remove.
8692         (reg->reg splitters): Remove splitter pattern.
8693         (reg->mem splitters): Ditto.
8695         (truncdfsf2): New insn pattern.
8696         (truncxf<mode>2): Ditto.
8698 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8700         * tree-ssa-math-opts.c (is_mult_by): New function.
8701         (is_square_of): Use the above.
8702         (optimize_recip_sqrt): New function.
8703         (pass_cse_reciprocals::execute): Use the above.
8705 2018-09-05  Richard Biener  <rguenther@suse.de>
8707         PR bootstrap/87134
8708         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
8709         to zero-init the emplaced vec.
8711 2018-09-05  Martin Liska  <mliska@suse.cz>
8713         PR tree-optimization/87205
8714         * tree-switch-conversion.c (pass_lower_switch::execute):
8715         Group cases for switch statements.
8717 2018-09-05  Richard Biener  <rguenther@suse.de>
8719         PR tree-optimization/87217
8720         * tree-ssa-sccvn.c (vuse_valueize): New.
8721         (vn_reference_lookup_pieces): Use it.
8722         (vn_reference_lookup): Likewise.
8724 2018-09-05  Nathan Sidwell  <nathan@acm.org>
8726         PR c++/87137
8727         * stor-layout.c (place_field): Scan forwards to check last
8728         bitfield when ms_bitfield_placement is in effect.
8730 2018-09-05  Richard Biener  <rguenther@suse.de>
8732         PR bootstrap/87225
8733         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
8734         return.
8736 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
8737             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
8739         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
8740         * config.gcc (extra_objs): Build it.
8741         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
8742         Likewise.
8743         * config/aarch64/aarch64-passes.def
8744         (pass_tag_collision_avoidance): New pass.
8745         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
8746         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
8747         (aarch64_classify_address): Remove static qualifier.
8748         (aarch64_address_info, aarch64_address_type): Move to...
8749         * config/aarch64/aarch64-protos.h: ... here.
8750         (make_pass_tag_collision_avoidance): New function.
8751         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
8752         New tuning flag.
8754 2018-09-05  Martin Liska  <mliska@suse.cz>
8756         * doc/gcov.texi: Update documentation of humar
8757         readable mode.
8758         * gcov.c (format_count): Print one decimal place, it provides
8759         more fine number of situations like '1G' vs. '1.4G'.
8761 2018-09-05  Martin Liska  <mliska@suse.cz>
8763         PR target/87164
8764         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
8765         * optc-gen.awk: Allow 'Var' for Deprecated options in order
8766         to generate a MASK value.
8768 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
8770         PR debug/86593
8771         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
8772         if frame pointer isn't used.
8773         (compute_frame_pointer_to_fb_displacement): Likewise.
8775 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
8777         PR target/87198
8778         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
8779         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
8780         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
8781         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
8782         and OPTION_MASK_ISA_XSAVEC_UNSET.
8784 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
8786         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
8787         XOR operations in NAND case.
8789 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
8791         * wide-int-range.cc (wide_int_range_convert): New.
8792         * wide-int-range.h (wide_int_range_convert): New.
8793         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
8794         code into wide_int_range_convert.
8795         (extract_range_into_wide_ints): Do not munge anti range constants
8796         into the entire domain.  Just return the range back.
8798 2018-09-04  Martin Liska  <mliska@suse.cz>
8800         * genmatch.c (output_line_directive): Add new argument
8801         fnargs.
8802         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
8804 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
8806         * doc/invoke.texi (Option Summary): Add whitespace.
8808         * doc/invoke.texi (Option Summary): Add -Waligned-new.
8810 2018-09-04  Richard Biener  <rguenther@suse.de>
8812         PR tree-optimization/87211
8813         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
8814         backedge value we're supposed to treat as VARYING also number
8815         the PHI to VARYING in case it got a different value-number already.
8817 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
8819         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
8820         (extract_range_from_binary_expr_1): Do not call
8821         vrp_can_optimize_bit_op.
8822         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
8823         static.
8824         (wide_int_range_get_mask_and_bounds): New.
8825         (wide_int_range_optimize_bit_op): New.
8826         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
8827         (wide_int_range_bit_and): Same.
8828         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
8829         (wide_int_range_optimize_bit_op): New.
8830         (wide_int_range_get_mask_and_bounds): New.
8832 2018-09-04  Richard Biener  <rguenther@suse.de>
8834         PR tree-optimization/87176
8835         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
8836         variable.  When value-numbering a virtual PHI node make sure
8837         to not value-number to the backedge value.
8839 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
8841         * doc/extend.texi (Long Long, Hex Floats): Document support for
8842         long long and hex floats in more recent versions of ISO C++.
8844 2018-09-03  Richard Biener  <rguenther@suse.de>
8846         PR tree-optimization/87177
8847         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
8848         cleanup.
8850 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
8852         * bb-reorder.c (edge_order): Convert to C-qsort-style
8853         tri-state comparator.
8854         (reorder_basic_blocks_simple): Change std::stable_sort to
8855         gcc_stablesort.
8857 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
8859         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
8860         tri-state comparator.
8861         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
8863 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
8865         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
8866         (mergesort): ... here as maximum count for using netsort.
8867         (gcc_qsort): Set nlim to 3 if stable sort is requested.
8868         (gcc_stablesort): New.
8869         * system.h (gcc_stablesort): Declare.
8871 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
8873         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
8874         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
8875         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
8877 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
8879         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
8880         lxsdx and stxsdx alternatives.
8881         (*mov<mode>_hardfloat64): Ditto.
8882         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
8884 2018-09-03  Richard Biener  <rguenther@suse.de>
8886         PR tree-optimization/87200
8887         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
8888         simplify result.
8890 2018-09-03  Martin Liska  <mliska@suse.cz>
8892         PR tree-optimization/87201
8893         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
8894         Fix parenthesis in an expression.
8896 2018-09-03  Richard Biener  <rguenther@suse.de>
8898         PR tree-optimization/87197
8899         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
8900         visited.  CSE the VN_INFO hashtable lookup.
8902         PR tree-optimization/87169
8903         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
8904         iterating make sure there's no extra backedges from irreducible
8905         regions feeding the header.  Mark the destination block
8906         executable.
8908 2018-09-03  Martin Liska  <mliska@suse.cz>
8910         PR driver/83193
8911         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
8912         * common/common-targhooks.c (default_get_valid_option_values):
8913         New function.
8914         * common/common-targhooks.h (default_get_valid_option_values):
8915         Likewise.
8916         * common/config/i386/i386-common.c: Move processor_target_table
8917         from i386.c.
8918         (ix86_get_valid_option_values): New function.
8919         (TARGET_GET_VALID_OPTION_VALUES): New macro.
8920         * config/i386/i386.c (struct ptt): Move to i386-common.c.
8921         (PTA_*): Move all defined masks into i386-common.c.
8922         (ix86_function_specific_restore): Use new processor_cost_table.
8923         * config/i386/i386.h (struct ptt): Moved from i386.c.
8924         (struct pta): Likewise.
8925         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
8926         * doc/tm.texi.in: Likewise.
8927         * opt-suggestions.c (option_proposer::suggest_option):
8928         Pass prefix to build_option_suggestions.
8929         (option_proposer::get_completions): Likewise.
8930         (option_proposer::build_option_suggestions): Use the new target
8931         hook.
8932         * opts.c (struct option_help_tuple): New struct.
8933         (print_filtered_help): Use the new target hook.
8935 2018-09-03  Martin Liska  <mliska@suse.cz>
8937         PR middle-end/59521
8938         * predict.c (set_even_probabilities): Add likely_edges
8939         argument and handle cases where we have precisely one
8940         likely edge.
8941         (combine_predictions_for_bb): Catch also likely_edges.
8942         (tree_predict_by_opcode): Handle gswitch statements.
8943         * tree-cfg.h (find_case_label_for_value): New declaration.
8944         (find_taken_edge_switch_expr): Likewise.
8945         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
8946         Find pivot in decision tree based on probabily, not by number of
8947         nodes.
8949 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
8951         * doc/standards.texi (Standards): Update Objective-C reference.
8953 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
8955         * doc/install.texi (Prerequisites): Update link for MPC.
8957 2018-09-01  Michael Matz  <matz@suse.de>
8959         PR tree-optimization/87074
8960         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
8961         PHIs for outer-loop uses.
8963 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
8965         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
8966         * doc/invoke.texi (C Dialect Options): Ditto.
8968 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
8970         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
8972 2018-08-31  Richard Biener  <rguenther@suse.de>
8974         PR tree-optimization/87168
8975         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
8976         (rpo_elim::eliminate_avail): When OP was not visited it must
8977         be available.
8979 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
8981         * tree-vrp.c (copy_value_range): Convert param "from" from
8982         "value_range *" to "const value_range *".
8983         (range_is_null): Likewise for param "vr".
8984         (range_int_cst_p): Likewise.
8985         (range_int_cst_singleton_p): Likewise.
8986         (symbolic_range_p): Likewise.
8987         (value_ranges_intersect_p): Likewise for both params.
8988         (value_range_nonnegative_p): Likewise for param "vr".
8989         (value_range_constant_singleton): Likewise.
8990         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
8991         (extract_range_into_wide_ints): Likewise for param "vr".
8992         (extract_range_from_multiplicative_op): Likewise for params "vr0"
8993         and "vr1".
8994         (vrp_can_optimize_bit_op): Likewise.
8995         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
8996         "vr1_".
8997         (extract_range_from_unary_expr): Likewise.
8998         (debug_value_range): Likewise for param "vr".
8999         (value_range::dump): Add "const" qualifier.
9000         (vrp_prop::check_array_ref): Convert local "vr" from
9001         "value_range *" to "const value_range *".
9002         (vrp_prop::check_mem_ref): Likewise.
9003         (vrp_prop::visit_stmt): Likewise for local "old_vr".
9004         (vrp_intersect_ranges_1): Likewise for param "vr_1".
9005         (vrp_intersect_ranges): Likewise.
9006         (simplify_stmt_for_jump_threading): Likewise for local "vr".
9007         (vrp_prop::vrp_finalize): Likewise.
9008         * tree-vrp.h (value_range::dump): Add "const" qualifier.
9009         (vrp_intersect_ranges): Add "const" qualifier to params as above.
9010         (extract_range_from_unary_expr): Likewise.
9011         (value_range_constant_singleton): Likewise.
9012         (symbolic_range_p): Likewise.
9013         (copy_value_range): Likewise.
9014         (extract_range_from_binary_expr_1): Likewise.
9015         (range_int_cst_p): Likewise.
9016         (vrp_set_zero_nonzero_bits): Likewise.
9017         (range_int_cst_singleton_p): Likewise.
9019 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
9021         * config/aarch64/arm_neon.h (vabsd_s64): New.
9022         (vnegd_s64): Likewise.
9024 2018-08-31  Martin Jambor  <mjambor@suse.cz>
9026         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
9028 2018-08-31  Martin Liska  <mliska@suse.cz>
9030         * ipa-icf.c (sem_item::add_type): Use
9031         sem_item::m_type_hash_cache.
9032         * ipa-icf.h: Move the cache from sem_item_optimizer
9033         to sem_item.
9035 2018-08-31  Nathan Sidwell  <nathan@acm.org>
9037         * doc/extend.texi (Backwards Compatibility): Remove implicit
9038         extern C leeway of () being (...).
9040 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9042         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
9044 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
9046         PR target/86684
9047         PR target/87149
9048         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
9050 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
9052         PR middle-end/87138
9053         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
9054         gen_int_mode.  Formatting fixes.
9056 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
9058         * target.def (custom_function_descriptors): Improve documentation.
9059         * doc/tm.texi.in (Trampolines): Expand discussion of function
9060         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
9061         beginning of the section.
9062         * doc/tm.texi: Regenerated.
9064 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
9066         * cfg.h (class auto_edge_flag): Spell out the template-id of the
9067         base class in the initializer list.  This is a workaround for
9068         building with older GCC.
9069         (class auto_bb_flag): Likewise.
9071 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
9073         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
9074         (altivec_vcmpequ<VI_char>_p): Remove star.
9075         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
9076         vector load modes.
9077         (expand_strncmp_vec_sequence): New function.
9078         (emit_final_str_compare_vec): New function.
9079         (expand_strn_compare): Add support for vector strncmp.
9080         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
9081         length specification to bytes.
9082         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
9083         (vcmpnezb_p): New pattern.
9084         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
9085         for option -mstring-compare-inline-limit.
9087 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
9089         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
9090         (PTA_SKYLAKE): Add PTA_AES.
9091         (PTA_GOLDMONT): Likewise.
9093 2018-08-29  Jan Hubicka  <jh@suse.cz>
9095         PR lto/86517
9096         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
9097         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
9099 2018-08-29  Jan Hubicka  <jh@suse.cz>
9101         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
9102         TYPE_STUB_DECL.
9103         (hash_tree): Do not visit TYPE_STUB_DECL.
9104         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
9105         stream TYPE_STUB_DECL.
9106         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
9107         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
9108         after free_lang_data.
9109         (type_in_anonymous_namespace_p): Likewise.
9111 2018-08-29  Jan Hubicka  <jh@suse.cz>
9113         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
9114         comment that it has to be even number.
9115         (class sreal): Change m_sig type to int32_t.
9116         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
9117         int64_t for temporary calculations.
9118         (sreal_verify_basics): Drop one bit from minimum and maximum.
9120 2018-08-30  Richard Biener  <rguenther@suse.de>
9122         PR tree-optimization/87147
9123         * tree-ssa-sccvn.c (SSA_VISITED): New function.
9124         (visit_phi): When the degenerate result is from the backedge and
9125         we didn't visit its definition yet drop to VARYING.
9126         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
9128 2018-08-29  Jan Hubicka  <jh@suse.cz>
9130         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
9131         DECL_VINDEX.
9132         (hash_tree): Likewise.
9134 2018-08-29  Jan Hubicka  <jh@suse.cz>
9136         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
9137         and TYPE_NEXT_REF_TO.
9139 2018-08-29  Jan Hubicka  <jh@suse.cz>
9141         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
9142         comment that it has to be even number.
9143         (class sreal): Change m_sig type to int32_t.
9144         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
9145         int64_t for temporary calculations.
9146         (sreal_verify_basics): Drop one bit from minimum and maximum.
9148 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
9150         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
9152 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
9154         PR middle-end/86995
9155         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
9156         if to_add is negative.
9158 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9160         PR middle-end/87053
9161         * builtins.c (c_strlen): Improve range checks.
9163 2018-08-29  Martin Sebor  <msebor@redhat.com>
9164             Jeff Law  <law@redhat.com>
9166         PR tree-optimization/86714
9167         PR tree-optimization/86711
9168         * builtins.c (c_strlen): Add arguments to call to string_constant.
9169         * expr.c (string_constant): Add argument.  Detect missing nul
9170         terminator and outermost declaration it's missing in.
9171         * expr.h (string_constant): Add argument.
9172         * fold-const.c (read_from_constant_string): Add arguments to call to
9173         string_constant.
9174         (c_getstr): Likewise.
9175         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
9176         to string_constant.
9177         * tree-ssa-strlen.c (get_stridx): Likewise.
9179 2018-08-29  Jan Hubicka  <jh@suse.cz>
9181         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
9182         Do not stream DECL_VINDEX.
9183         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
9184         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
9185         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
9187 2018-08-29  Richard Biener  <rguenther@suse.de>
9189         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
9190         virtual operands that are not default defs to honor region
9191         boundaries.
9192         (rpo_vn_valueize): Remove ineffective code here.
9194 2018-08-29  Richard Biener  <rguenther@suse.de>
9196         PR tree-optimization/87132
9197         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
9198         when skipping defs reachable over backedges.
9200 2018-08-29  Richard Biener  <rguenther@suse.de>
9202         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
9203         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
9204         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
9205         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
9206         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
9207         (vn_reference_lookup_3): Remove use of const_parms.
9208         (free_rpo_vn): Do not free const_parms.
9209         (do_rpo_vn): Do not call init_const_parms.
9210         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
9211         SSA_NAME_POINTS_TO_READONLY_MEMORY.
9212         (call_may_clobber_ref_p_1): Likewise.
9214 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
9216         PR other/86726
9217         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
9218         (-O): Ditto.
9219         (-ftree-scev-cprop): Document.
9221 2018-08-29  Jan Hubicka  <jh@suse.cz>
9223         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
9224         parameters.
9225         (sreal constructor): Update.
9226         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
9227         sreal:operator/): Update.
9229 2018-08-29  Martin Liska  <mliska@suse.cz>
9231         * tree-switch-conversion.c (switch_conversion::expand):
9232         Strenghten assumption about gswitch statements.
9234 2018-08-29  Richard Biener  <rguenther@suse.de>
9236         PR tree-optimization/87117
9237         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
9238         re-value-number released SSA VDEFs.
9240 2018-08-29  Richard Biener  <rguenther@suse.de>
9242         PR tree-optimization/87126
9243         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
9245 2018-08-28  Jim Wilson  <jimw@sifive.com>
9247         * config/riscv/pic.md: Rewrite.
9248         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
9249         invalid address.
9250         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
9251         (SOFTF, default_load, softload, softstore): New.
9253 2018-08-28  Jeff Law  <law@redhat.com>
9255         * fold-const.c (fold_binary_loc): Remove recently added assert.
9257 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
9259         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
9260         to OP parmeter of generated function.
9262 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
9264         PR tree-optimization/87009
9265         * match.pd: Add boolean optimizations.
9267 2018-08-28  Martin Sebor  <msebor@redhat.com>
9269         PR middle-end/86631
9270         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
9271         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
9272         (pass_walloca::gate): Use it.
9273         (alloca_call_type): Same.
9274         (pass_walloca::execute): Same.
9275         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
9277 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
9279         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
9280         GCC_VERSION for usage of "__gcc_dump_printf__" format from
9281         >= 3005 to >= 9000.
9283 2018-08-28  Richard Biener  <rguenther@suse.de>
9285         PR tree-optimization/87124
9286         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
9287         constants before looking up avail.
9289 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
9291         PR middle-end/87099
9292         * calls.c (maybe_warn_nonstring_arg): Punt early if
9293         warn_stringop_overflow is zero.  Don't call get_range_strlen
9294         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
9295         Swap comparison operands to have constants on rhs.  Only use
9296         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
9297         increment lenrng[0].
9299 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
9301         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
9302         use of tree_to_shwi.  Remove duplicated test for the size being
9303         a whole number of bytes.
9305 2018-08-28  Richard Biener  <rguenther@suse.de>
9307         PR tree-optimization/87117
9308         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
9309         Handle removed stmt without LHS (GIMPLE_NOP).
9311 2018-08-28  Richard Biener  <rguenther@suse.de>
9313         PR tree-optimization/87117
9314         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
9315         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
9317 2018-08-28  Richard Biener  <rguenther@suse.de>
9319         PR tree-optimization/87117
9320         * tree-ssa-pre.c (compute_avail): Do not make expressions
9321         with predicated values available.
9322         (get_expr_value_id): Assert we do not run into predicated value
9323         expressions.
9325 2018-08-28  Richard Biener  <rguenther@suse.de>
9327         PR tree-optimization/87117
9328         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
9329         get virtual operands.
9330         (get_expr_operands): Handle STRING_CST like other decls.
9332 2018-08-28  Martin Liska  <mliska@suse.cz>
9334         * tree.h: Update documentation of fndecl_built_in_p
9335         functions.
9338 2018-08-27  Jeff Law  <law@redhat.com>
9339         PR tree-optimization/87110
9340         * tree-ssa-dse.c (compute_trims): Handle non-constant
9341         TYPE_SIZE_UNIT.
9343 2018-08-27  Martin Sebor  <msebor@redhat.com>
9345         PR tree-optimization/86914
9346         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
9348 2018-08-27  Martin Sebor  <msebor@redhat.com>
9350         PR tree-optimization/87112
9351         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
9352         the type of the bound argument.
9354 2018-08-27  Jeff Law  <law@redhat.com>
9356         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
9357         type does not have a TYPE_SIZE_UNIT.
9359 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
9361         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
9362         with include of backend.h.
9364 2018-08-27  Richard Biener  <rguenther@suse.de>
9366         PR tree-optimization/86927
9367         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
9368         use const cond reduction code.
9370 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
9372         PR tree-optimization/85758
9373         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
9375 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
9377         PR c++/87091
9378         * diagnostic-show-locus.c (class layout_range): Update for
9379         conversion of show_caret_p to a tri-state.
9380         (layout_range::layout_range): Likewise.
9381         (make_range): Likewise.
9382         (layout::maybe_add_location_range): Likewise.
9383         (layout::should_print_annotation_line_p): Don't show annotation
9384         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
9385         (layout::get_state_at_point): Update for conversion of
9386         show_caret_p to a tri-state.  Bail out early for
9387         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
9388         underlining or source colorization.
9389         (gcc_rich_location::add_location_if_nearby): Update for conversion
9390         of show_caret_p to a tri-state.
9391         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
9392         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
9393         Likewise.
9394         (selftest::test_one_liner_labels): Likewise.
9395         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
9396         conversion of show_caret_p to a tri-state.
9397         * pretty-print.c (text_info::set_location): Likewise.
9398         * pretty-print.h (text_info::set_location): Likewise.
9399         * substring-locations.c (format_warning_n_va): Likewise.
9400         * tree-diagnostic.c (default_tree_printer): Likewise.
9401         * tree-pretty-print.c (newline_and_indent): Likewise.
9403 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
9405         PR c++/87091
9406         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
9407         line above for line-insertion fix-it hints.
9408         (selftest::test_fixit_insert_containing_newline): Update the
9409         expected results, and add a test with line-numbering enabled.
9411 2018-08-27  Martin Liska  <mliska@suse.cz>
9413         PR sanitizer/86962
9414         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
9415         params with DECL_HAS_VALUE_EXPR_P.
9417 2018-08-27  Martin Liska  <mliska@suse.cz>
9419         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
9420         selected expansion strategy.
9422 2018-08-27  Martin Liska  <mliska@suse.cz>
9424         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
9425         * builtins.c (is_builtin_fn): Likewise.
9426         * attribs.c (diag_attr_exclusions): Use new function
9427         fndecl_built_in_p and remove check for FUNCTION_DECL if
9428         possible.
9429         (builtin_mathfn_code): Likewise.
9430         (fold_builtin_expect): Likewise.
9431         (fold_call_expr): Likewise.
9432         (fold_builtin_call_array): Likewise.
9433         (fold_call_stmt): Likewise.
9434         (set_builtin_user_assembler_name): Likewise.
9435         (is_simple_builtin): Likewise.
9436         * calls.c (gimple_alloca_call_p): Likewise.
9437         (maybe_warn_nonstring_arg): Likewise.
9438         * cfgexpand.c (expand_call_stmt): Likewise.
9439         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
9440         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
9441         (cgraph_node::verify_node): Likewise.
9442         * cgraphclones.c (build_function_decl_skip_args): Likewise.
9443         (cgraph_node::create_clone): Likewise.
9444         * config/arm/arm.c (arm_insert_attributes): Likewise.
9445         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
9446         * dse.c (scan_insn): Likewise.
9447         * expr.c (expand_expr_real_1): Likewise.
9448         * fold-const.c (operand_equal_p): Likewise.
9449         (fold_binary_loc): Likewise.
9450         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
9451         * gimple-low.c (lower_stmt): Likewise.
9452         * gimple-pretty-print.c (dump_gimple_call): Likewise.
9453         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9454         Likewise.
9455         * gimple.c (gimple_build_call_from_tree): Likewise.
9456         (gimple_call_builtin_p): Likewise.
9457         (gimple_call_combined_fn): Likewise.
9458         * gimplify.c (gimplify_call_expr): Likewise.
9459         (gimple_boolify): Likewise.
9460         (gimplify_modify_expr): Likewise.
9461         (gimplify_addr_expr): Likewise.
9462         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
9463         * ipa-cp.c (determine_versionability): Likewise.
9464         * ipa-fnsummary.c (compute_fn_summary): Likewise.
9465         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
9466         * ipa-split.c (visit_bb): Likewise.
9467         (split_function): Likewise.
9468         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
9469         * lto-cgraph.c (input_node): Likewise.
9470         * lto-streamer-out.c (write_symbol): Likewise.
9471         * omp-low.c (setjmp_or_longjmp_p): Likewise.
9472         (lower_omp_1): Likewise.
9473         * predict.c (strip_predict_hints): Likewise.
9474         * print-tree.c (print_node): Likewise.
9475         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
9476         * trans-mem.c (is_tm_irrevocable): Likewise.
9477         (is_tm_load): Likewise.
9478         (is_tm_simple_load): Likewise.
9479         (is_tm_store): Likewise.
9480         (is_tm_simple_store): Likewise.
9481         (is_tm_abort): Likewise.
9482         (tm_region_init_1): Likewise.
9483         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
9484         * tree-cfg.c (verify_gimple_call): Likewise.
9485         (move_stmt_r): Likewise.
9486         (stmt_can_terminate_bb_p): Likewise.
9487         * tree-eh.c (lower_eh_constructs_2): Likewise.
9488         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
9489         * tree-inline.c (remap_gimple_stmt): Likewise.
9490         (copy_bb): Likewise.
9491         (estimate_num_insns): Likewise.
9492         (fold_marked_statements): Likewise.
9493         * tree-sra.c (scan_function): Likewise.
9494         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
9495         (optimize_stack_restore): Likewise.
9496         (pass_fold_builtins::execute): Likewise.
9497         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
9498         (mark_all_reaching_defs_necessary_1): Likewise.
9499         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
9500         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
9501         (pass_forwprop::execute): Likewise.
9502         * tree-ssa-loop-im.c (stmt_cost): Likewise.
9503         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
9504         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
9505         * tree-ssa-strlen.c (get_string_length): Likewise.
9506         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
9507         (find_func_aliases_for_call): Likewise.
9508         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
9509         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
9510         * tree-tailcall.c (find_tail_calls): Likewise.
9511         * tree.c (need_assembler_name_p): Likewise.
9512         (free_lang_data_in_decl): Likewise.
9513         (get_call_combined_fn): Likewise.
9514         * ubsan.c (is_ubsan_builtin_p): Likewise.
9515         * varasm.c (incorporeal_function_p): Likewise.
9516         * tree.h (DECL_BUILT_IN): Remove and replace with
9517         fndecl_built_in_p.
9518         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
9519         (fndecl_built_in_p): New.
9521 2018-08-27  Martin Liska  <mliska@suse.cz>
9523         PR tree-optimization/86847
9524         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
9525         Dump also subtree probability.
9526         (switch_decision_tree::do_jump_if_equal): New function.
9527         (switch_decision_tree::emit_case_nodes): Handle special
9528         situations in balanced tree that can be emitted much simpler.
9529         Fix calculation of probabilities that happen in tree expansion.
9530         * tree-switch-conversion.h (struct cluster): Add
9531         is_single_value_p.
9532         (struct simple_cluster): Likewise.
9533         (struct case_tree_node): Add new function has_child.
9534         (do_jump_if_equal): New.
9536 2018-08-27  Martin Liska  <mliska@suse.cz>
9538         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
9539         Add new argument to bit_test_cluster constructor.
9540         (bit_test_cluster::emit): Set bits really number of values
9541         handlel by a test.
9542         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
9543         probability argument.
9544         * tree-switch-conversion.h (struct bit_test_cluster):
9545         Add m_handles_entire_switch.
9547 2018-08-27  Martin Liska  <mliska@suse.cz>
9549         PR tree-optimization/86702
9550         * tree-switch-conversion.c (jump_table_cluster::emit):
9551         Make probabilities even for values in jump table
9552         according to number of cases handled.
9553         (switch_decision_tree::compute_cases_per_edge): Pass
9554         argument to reset_out_edges_aux function.
9555         (switch_decision_tree::analyze_switch_statement): Likewise.
9556         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
9557         Make it static.
9559 2018-08-27  Martin Liska  <mliska@suse.cz>
9561         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
9562         cfun argument explicitly.
9563         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
9564         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
9565         function gimple_switch_default_bb.
9566         (convert_switch_statements):
9567         (expand_builtins):
9568         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
9569         * stmt.c (label_to_block_fn): Use label_to_block and pass
9570         cfun argument explicitly and use gimple_switch_label_bb.
9571         (expand_case): Likewise.
9572         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
9573         cfun argument explicitly. Likewise.
9574         (make_edges_bb): Likewise.
9575         (make_cond_expr_edges): Likewise.
9576         (get_cases_for_edge): Likewise.
9577         (make_gimple_switch_edges): Likewise.
9578         (label_to_block_fn): Likewise.
9579         (label_to_block): Likewise.
9580         (make_goto_expr_edges): Likewise.
9581         (make_gimple_asm_edges): Likewise.
9582         (main_block_label): Likewise.
9583         (group_case_labels_stmt): Likewise.
9584         (find_taken_edge_computed_goto): Likewise.
9585         (find_taken_edge_switch_expr): Likewise.
9586         (gimple_verify_flow_info): Likewise.
9587         (gimple_redirect_edge_and_branch): Likewise.
9588         (gimple_switch_label_bb): New function.
9589         (gimple_switch_default_bb): Likewise.
9590         (gimple_switch_edge): Likewise.
9591         (gimple_switch_default_edge): Likewise.
9592         * tree-cfg.h (label_to_block_fn): Remove and replace ...
9593         (label_to_block): ... with this.
9594         (gimple_switch_label_bb): New.
9595         (gimple_switch_default_bb): Likewise.
9596         (gimple_switch_edge): Likewise.
9597         (gimple_switch_default_edge): Likewise.
9598         * tree-cfgcleanup.c (convert_single_case_switch): Use
9599         new gimple functions and pass new argument to label_to_block.
9600         (cleanup_control_flow_bb):
9601         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
9602         cfun argument explicitly.
9603         (make_eh_edges): Likewise.
9604         (redirect_eh_dispatch_edge): Likewise.
9605         (lower_resx): Likewise.
9606         (lower_eh_dispatch): Likewise.
9607         (maybe_remove_unreachable_handlers): Likewise.
9608         (unsplit_eh): Likewise.
9609         (cleanup_empty_eh): Likewise.
9610         (verify_eh_edges): Likewise.
9611         (verify_eh_dispatch_edge): Likewise.
9612         * tree-ssa-dom.c (record_edge_info): Likewise.
9613         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
9614         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
9615         (thread_through_normal_block): Likewise.
9616         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
9617         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
9618         * tree-switch-conversion.c (switch_conversion::collect): Use new
9619         gimple functions.
9620         (switch_conversion::check_final_bb): Likewise.
9621         (switch_conversion::gather_default_values): Pass new argument
9622         to label_to_block.
9623         (switch_conversion::build_constructors): Likewise.
9624         (switch_decision_tree::compute_cases_per_edge): Use new
9625         gimple_switch_edge function.
9626         (switch_decision_tree::analyze_switch_statement): Pass new argument
9627         to label_to_block.
9628         (switch_decision_tree::try_switch_expansion): Use
9629         gimple_switch_default_edge.
9630         * tree-vrp.c (find_switch_asserts): Pass new argument
9631         to label_to_block.
9632         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
9633         (vr_values::simplify_switch_using_ranges): Likewise.
9635 2018-08-27  Richard Biener  <rguenther@suse.de>
9637         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
9638         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
9640         * tree-ssa-sccvn.h (struct vn_pval): New structure.
9641         (struct vn_nary_op_s): Add unwind_to member.  Add
9642         predicated_values flag and put result into a union together
9643         with a linked list of vn_pval.
9644         (struct vn_ssa_aux): Add name member to make maintaining
9645         a map of SSA name to vn_ssa_aux possible.  Remove no longer
9646         needed info, dfsnum, low, visited, on_sccstack, use_processed
9647         and range_info_anti_range_p members.
9648         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
9649         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
9650         New functions.
9651         (vn_valueize): New global.
9652         (vn_context_bb): Likewise.
9653         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
9654         VN_INFO_PTR_INFO): Remove.
9655         * tree-ssa-sccvn.c: ... (rewrite)
9656         (pass_fre::execute): For -O2+ initialize loops and run
9657         RPO VN in optimistic mode (iterating).  For -O1 and -Og
9658         run RPO VN in non-optimistic mode.
9659         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
9660         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
9661         * doc/invoke.texi (sccvn-max-scc-size): Remove.
9662         (rpo-vn-max-loop-depth): Document.
9663         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
9664         when valuezing the VUSE signals we walked out of the region.
9665         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
9666         (phi_translate): Set VN context block to use for availability
9667         lookup.
9668         (compute_avail): Likewise.
9669         (pre_valueize): New function.
9670         (pass_pre::execute): Adjust to the RPO VN API.
9672         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
9673         (propagate_constants_for_unrolling): Remove.
9674         (tree_unroll_loops_completely): Perform value-numbering
9675         on the unrolled bodies loop parent.
9677 2018-08-27  Richard Biener  <rguenther@suse.de>
9679         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
9680         for partial antic compute.
9682 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
9684         PR rtl-optimization/87065
9685         * combine.c (simplify_if_then_else): Formatting fix.
9686         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
9687         check.
9688         (known_cond): Don't return const_true_rtx for vector modes.  Use
9689         CONST0_RTX instead of const0_rtx.  Formatting fixes.
9691 2018-08-27  Martin Liska  <mliska@suse.cz>
9693         PR gcov-profile/87069
9694         * gcov.c (process_file): Record files already processed
9695         and warn about a file being processed multiple times.
9697 2018-08-27  Martin Liska  <mliska@suse.cz>
9699         PR driver/83193
9700         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9701         Set default values for x_aarch64_*_string strings.
9702         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
9703         prefix.  For -mabi do not print '=ABI' in help and use
9704         <option_value> format for -msve-vector-bits and -moverride
9705         options.
9707 2018-08-26  Jeff Law  <law@redhat.com>
9709         * config/mips/frame-header-opt.c: Include "backend.h" rather than
9710         "cfg.h"
9712 2018-08-26  Marek Polacek  <polacek@redhat.com>
9714         PR c++/87029, Implement -Wredundant-move.
9715         * doc/invoke.texi: Document -Wredundant-move.
9717 2018-08-25  Martin Sebor  <msebor@redhat.com>
9719         PR tree-optimization/87059
9720         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
9721         to the same type as the other.
9722         * fold-const.c (fold_binary_loc): Assert expectation.
9724 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
9726         * config/darwin.c (machopic_legitimize_pic_address): Clean up
9727         extraneous parentheses, dead code section and formatting.
9729 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
9731         PR c++/87091
9732         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
9733         wide enough for jumps in the line-numbering to be visible.
9734         (layout::print_gap_in_line_numbering): New member function.
9735         (layout::calculate_line_spans): When using line numbering, merge
9736         line spans that are only 1 line apart.
9737         (diagnostic_show_locus): When printing line numbers, show gaps in
9738         line numbering directly, rather than printing headers.
9739         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
9740         line-numbering with multiple line spans.
9741         (selftest::test_fixit_insert_containing_newline_2): Add test of
9742         line-numbering, in which the spans are close enough to be merged.
9744 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
9746         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
9747         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
9748         * tree-vrp.c (range_is_nonnull): Remove.
9749         (range_includes_zero_p): Accept value_range instead of min/max.
9750         (extract_range_from_binary_expr_1): Do not early bail on
9751         POINTER_PLUS_EXPR.
9752         Use range_includes_zero_p instead of range_is_nonnull.
9753         (extract_range_from_unary_expr): Use range_includes_zero_p instead
9754         of range_is_nonnull.
9755         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
9756         special case VR_ANTI_RANGE.
9757         (vrp_finalize): Same.
9758         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
9759         instead of min/max.
9760         (range_is_nonnull): Remove.
9761         * vr-values.c (vrp_stmt_computes_nonzero): Use
9762         range_includes_zero_p instead of range_is_nonnull.
9763         (extract_range_basic): Pass value_range to range_includes_zero_p
9764         instead of range_is_nonnull.
9766 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
9768         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
9769         * emit-rtl.h (rtl_data): Remove return_bnd.
9770         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
9771         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
9772         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
9773         (POINTER_BOUNDS_MODE): Remove definition.
9774         (make_pointer_bounds_mode): Remove.
9775         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
9776         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
9777         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
9778         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
9779         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
9780         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
9781         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
9783         * config/i386/i386-modes.def (BND32, BND64): Remove.
9784         * config/i386/i386.c (dbx_register_map): Remove bound registers.
9785         (dbx64_register_map): Ditto.
9786         (svr4_dbx_register_map): Ditto.
9787         (indirect_thunk_bnd_needed): Remove.
9788         (indirect_thunks_bnd_used): Ditto.
9789         (indirect_return_bnd_needed): Ditto.
9790         (indirect_return_via_cx_bnd): Ditto.
9791         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
9792         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
9793         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
9794         (output_indirect_thunk_function): Remove handling of
9795         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
9796         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
9797         (ix86_save_reg): Remove handling of crtl->return_bnd.
9798         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
9799         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
9800         and UNSPEC_BNDLX_ADDR.
9801         (ix86_output_indirect_branch_via_reg): Remove handling of
9802         indirect_thunk_prefix_bnd.
9803         (ix86_output_indirect_branch_via_push): Ditto.
9804         (ix86_output_function_return): Ditto.
9805         (ix86_output_indirect_function_return): Ditto.
9806         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
9807         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
9808         (CALL_USED_REGISTERS): Ditto.
9809         (REG_ALLOC_ORDER): Update for removal of bound registers.
9810         (HI_REGISTER_NAMES): Ditto.
9811         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
9812         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
9813         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
9814         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
9815         (FIRST_PSEUDO_REG): Update.
9816         (BND): Remove mode iterator.
9817         * config/i386/predicates.md (bnd_mem_operator): Remove.
9819 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
9821         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
9822         vectors.
9824 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
9826         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
9827         the case in which the permute needs only a single element and
9828         repeats for every vector of the result.  Extend that case to
9829         handle variable-length vectors.
9830         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
9832 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
9834         PR debug/79342
9835         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
9836         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
9838 2018-08-24  Richard Biener  <rguenther@suse.de>
9840         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
9841         bb_flags_allocated members.
9842         (auto_flag): New RAII class for allocating flags.
9843         (auto_edge_flag): New RAII class for allocating edge flags.
9844         (auto_bb_flag): New RAII class for allocating bb flags.
9845         * cfgloop.c (verify_loop_structure): Allocate temporary edge
9846         flag dynamically.
9847         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
9848         in favor of temporarily allocated BB flag.
9849         * hsa-brig.c: Re-order includes.
9850         * hsa-dump.c: Likewise.
9851         * hsa-regalloc.c: Likewise.
9852         * print-rtl.c: Likewise.
9853         * profile-count.c: Likewise.
9855 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
9857         PR target/86989
9858         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
9859         the TOC register.
9861 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
9863         PR 87073/bootstrap
9864         * wide-int-range.cc (wide_int_range_div): Do not ignore result
9865         from wide_int_range_multiplicative_op.
9867 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
9869         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
9870         "permutaion".
9872 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
9874         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
9875         to 'expanded'.
9877 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
9879         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
9880         full GENERIC expression used for replacement.
9882 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
9884         * tree-vrp.c (abs_extent_range): Remove.
9885         (extract_range_into_wide_ints): Pass wide ints by reference.
9886         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
9887         Pass wide ints by reference in all calls to
9888         extract_range_into_wide_ints.
9889         * wide-int-range.cc (wide_int_range_div): New.
9890         * wide-int-range.h (wide_int_range_div): New.
9891         (wide_int_range_includes_zero_p): New.
9892         (wide_int_range_zero_p): New.
9894 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
9896         * config/aarch64/aarch64.md (arches): New enum.
9897         (arch): New enum attr.
9898         (arch_enabled): New attr.
9899         (enabled): Now uses arch_enabled only.
9900         (simd, sve, fp16): Removed attribute.
9901         (fp): Attr now defined in terms of 'arch'.
9902         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
9903         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
9904         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
9905         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
9906         attributes into 'arch'.
9907         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
9908         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
9909         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
9910         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
9911         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
9912         'simd' attribute into 'arch'.
9913         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
9914         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
9915         Convert use of 'fp' attribute to 'arch'.
9916         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
9917         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
9918         into 'arch'.
9919         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
9920         (different modes) Merge 'fp' and 'simd' into 'arch'.
9921         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
9922         'simd' into 'arch'.
9924 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
9926         PR rtl-optimization/87026
9927         * expmed.c (canonicalize_comparison): If we can no longer create
9928         pseudoregisters, don't.
9930 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
9932         PR target/86951
9933         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
9934         prototype.
9935         * config/arm/arm.c (speculation_barrier_libfunc): New static
9936         variable.
9937         (arm_init_libfuncs): Initialize it.
9938         (arm_emit_speculation_barrier): New function.
9939         * config/arm/arm.md (speculation_barrier): Call
9940         arm_emit_speculation_barrier for architectures that do not have
9941         DSB or ISB.
9942         (speculation_barrier_insn): Only match on Armv7 or later.
9944 2018-08-23  Richard Biener  <rguenther@suse.de>
9946         PR middle-end/87024
9947         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
9948         calls.
9950 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
9952         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
9953         of single-vector TBLs.
9954         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
9955         one input is given.
9957 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
9959         PR target/85910
9960         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
9961         aarch64_evpc_tbl guard.
9963 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9965         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
9966         behaviour.
9968 2018-08-22  Martin Sebor  <msebor@redhat.com>
9970         PR middle-end/87052
9971         * tree-pretty-print.c (pretty_print_string): Add argument.
9972         (dump_generic_node): Call to pretty_print_string with string size.
9974 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
9976         PR rtl-optimization/86771
9977         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
9978         of two SETs into those two SETs, one to be placed at i2, if that SETs
9979         destination is modified between i2 and i3.
9981 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
9983         PR tree-optimization/86725
9984         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
9985         function.
9986         (vect_analyze_scalar_cycles_1): Check it.
9988 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
9990         PR tree-optimization/86725
9991         * tree-vect-loop.c (vect_is_simple_reduction): When treating
9992         an outer loop phi as a double reduction, make sure that the
9993         single user of the phi result is an inner loop phi.
9995 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
9997         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
9998         grouped stores with gaps to a strided group.
10000 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
10002         * tree-vect-stmts.c (get_group_load_store_type)
10003         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
10004         first statement in a group.
10006 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
10008         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
10009         the sequence used in gcc/gcc.c.
10011 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
10013         PR other/704
10014         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
10015         building it.
10017 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
10019         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
10020         Darwin10-specific unwinder-shim.
10021         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
10022         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
10023         New to cater for Darwin10 Rosetta.
10025 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
10027         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
10028         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
10030 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
10032         PR bootstrap/81033
10033         PR target/81733
10034         PR target/52795
10035         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
10036         (dwarf2out_switch_text_section): Generate a local label for the
10037         second function sub-section and apply it as the second FDE start
10038         label.
10039         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
10040         second sub-section start.
10042 2018-08-22  Richard Biener  <rguenther@suse.de>
10044         PR tree-optimization/86988
10045         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
10047 2018-08-22  Richard Biener  <rguenther@suse.de>
10049         PR tree-optimization/86945
10050         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
10052 2018-08-22  Alexandre Oliva <oliva@adacore.com>
10054         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
10055         a comment about how uses of r2 for .sdata2 come about.
10057 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
10059         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
10061 2018-08-21  Marek Polacek  <polacek@redhat.com>
10063         PR c++/86981, Implement -Wpessimizing-move.
10064         * doc/invoke.texi: Document -Wpessimizing-move.
10066 2018-08-21  Jan Hubicka  <jh@suse.cz>
10068         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
10069         * tree.h (is_redundant_typedef): Remove.
10070         * dwarf2out.c (is_redundant_typedef): Turn into static function.
10072 2018-08-21  Jan Hubicka  <jh@suse.cz>
10074         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
10075         when possible.
10077 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
10079         * expmed.c (extract_low_bits): Reject invalid subregs early.
10081 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10083         PR middle-end/86121
10084         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
10085         behaviour.
10087 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
10089         * config/vxworks.h: Guard vxworks_asm_out_constructor and
10090         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
10091         * config/vxworks.c: Likewise.
10093 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
10095         * config/vxworks.c: Set targetm.have_ctors_dtors
10096         if HAVE_INITFINI_ARRAY_SUPPORT.
10097         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
10098         if HAVE_INITFINI_ARRAY_SUPPORT.
10100 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
10102         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
10103         default search path for VxWorks < 7.
10105 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10107         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
10108         (get_format_string): Refer to c_getstr.
10110 2018-08-21  Tom de Vries  <tdevries@suse.de>
10112         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
10113         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
10114         (debuginfo_early_stop): Declare.
10115         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
10116         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
10117         (debuginfo_early_stop): New function.
10118         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
10119         and debuginfo_early_stop.
10120         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
10121         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
10122         (general_init): Call debuginfo_early_init.
10123         (finalize): Call debuginfo_fini.
10124         (do_compile): Call debuginfo_init.
10125         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
10126         -fdump-early-debug.
10127         (@item -fdump-debug, @item -fdump-earlydebug): Add.
10129 2018-08-21  Tom de Vries  <tdevries@suse.de>
10131         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
10132         flag_dump_noaddr and flag_dump_unnumbered.
10134 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
10136         * wide-int-range.cc (wide_int_range_abs): New.
10137         (wide_int_range_order_set): Rename from wide_int_range_min_max.
10138         * wide-int-range.h (wide_int_range_abs): New.
10139         (wide_int_range_min_max): New.
10140         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
10141         case to call wide_int_range_abs.
10142         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
10143         (extract_range_from_abs_expr): Delete.
10145 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
10147         PR target/87033
10148         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
10149         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
10150         for indexed loads.
10152 2018-08-20  Nathan Sidwell  <nathan@acm.org>
10153             Jeff Law <law@redhat.com>
10155         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
10156         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
10158 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
10160         PR other/84889
10161         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
10162         (decl_attributes): Likewise.
10163         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
10164         instance.
10165         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
10166         * diagnostic-core.h (class auto_diagnostic_group): New class.
10167         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
10168         (diagnostic_report_diagnostic): Handle the first diagnostics within
10169         a group.
10170         (emit_diagnostic): Add auto_diagnostic_group instance.
10171         (inform): Likewise.
10172         (inform_n): Likewise.
10173         (warning): Likewise.
10174         (warning_at): Likewise.
10175         (warning_n): Likewise.
10176         (pedwarn): Likewise.
10177         (permerror): Likewise.
10178         (error): Likewise.
10179         (error_n): Likewise.
10180         (error_at): Likewise.
10181         (sorry): Likewise.
10182         (fatal_error): Likewise.
10183         (internal_error): Likewise.
10184         (internal_error_no_backtrace): Likewise.
10185         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
10186         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
10187         * diagnostic.h (struct diagnostic_context): Add fields
10188         "diagnostic_group_nesting_depth",
10189         "diagnostic_group_emission_count", "begin_group_cb",
10190         "end_group_cb".
10191         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
10192         Add auto_diagnostic_group instance(s).
10193         (find_explicit_erroneous_behavior): Likewise.
10194         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
10195         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
10196         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
10197         (gimplify_va_arg_expr): Likewise.
10198         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
10199         (HSA_SORRY_AT): Likewise.
10200         * ipa-devirt.c (compare_virtual_tables): Likewise.
10201         (warn_odr): Likewise.
10202         * multiple_target.c (expand_target_clones): Likewise.
10203         * opts-common.c (cmdline_handle_error): Likewise.
10204         * reginfo.c (globalize_reg): Likewise.
10205         * substring-locations.c (format_warning_n_va): Likewise.
10206         * tree-inline.c (expand_call_inline): Likewise.
10207         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
10208         * tree-ssa-loop-niter.c
10209         (do_warn_aggressive_loop_optimizations): Likewise.
10210         * tree-ssa-uninit.c (warn_uninit): Likewise.
10211         * tree.c (warn_deprecated_use): Likewise.
10213 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
10215         PR target/87014
10216         * config/i386/i386.md (eh_return): Always update EH return
10217         address in word_mode.
10219 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
10221         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
10222         TARGET_SPLIT_COMPLEX_ARG is defined.
10224 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10226         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
10228 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10230         PR target/86984
10231         * expr.c (expand_assignment): Assert that bitpos is positive.
10232         (store_field): Likewise
10233         (expand_expr_real_1): Make sure that bitpos is positive.
10234         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
10235         integer overflow.
10237 2018-08-20  Nathan Sidwell  <nathan@acm.org>
10239         * Makefile.in (CPP_ID_DATA_H): Delete.
10240         (CPP_INTERNAL_H): Don't add it.
10241         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
10242         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
10244 2018-08-20  Richard Biener  <rguenther@suse.de>
10246         PR tree-optimization/78655
10247         * tree-vrp.c (extract_range_from_binary_expr_1): Make
10248         pointer + offset nonnull if either operand is nonnull work.
10250 2018-08-20  Tom de Vries  <tdevries@suse.de>
10252         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
10253         unless the referenced die describes the added property using
10254         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
10255         Otherwise, add a DW_AT_location to the referenced die.
10257 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
10259         PR target/86994
10260         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
10261         register_operand when calling ix86_set_reg_reg_cost.
10262         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
10263         Set *total to 0 for operands that satisfy x86_64_immediate_operand
10264         predicate and to 1 otherwise.
10266 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
10268         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
10269         emit a diagnostic that it is not supported and reset the option.
10270         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
10271         supported and consume the option.  (ASM_FINAL_SPEC): New.
10273 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
10275         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
10276         a sentence.
10278 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
10280         C-SKY port: Documentation
10282         * doc/extend.texi (C-SKY Function Attributes): New section.
10283         * doc/invoke.texi (Option Summary): Add C-SKY options.
10284         (C-SKY Options): New section.
10285         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
10287 2018-08-17  Jojo  <jijie_rong@c-sky.com>
10288             Huibin Wang  <huibin_wang@c-sky.com>
10289             Sandra Loosemore  <sandra@codesourcery.com>
10290             Chung-Lin Tang  <cltang@codesourcery.com>
10292         C-SKY port: Backend implementation
10294         * config/csky/*: New.
10295         * common/config/csky/*: New.
10297 2018-08-17  Jojo  <jijie_rong@c-sky.com>
10298             Huibin Wang  <huibin_wang@c-sky.com>
10299             Sandra Loosemore  <sandra@codesourcery.com>
10300             Chung-Lin Tang  <cltang@codesourcery.com>
10301             Andrew Jenner  <andrew@codesourcery.com>
10303         C-SKY port: Configury
10305         * config.gcc (csky-*-*): New.
10306         * configure.ac: Add csky to targets for dwarf2 debug_line support.
10307         * configure: Regenerated.
10309 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
10311         * dump-context.h: Include "dumpfile.h".
10312         (dump_context::dump_printf_va): Convert final param from va_list
10313         to va_list *.  Convert from ATTRIBUTE_PRINTF to
10314         ATTRIBUTE_GCC_DUMP_PRINTF.
10315         (dump_context::dump_printf_loc_va): Likewise.
10316         * dumpfile.c: Include "stringpool.h".
10317         (make_item_for_dump_printf_va): Delete.
10318         (make_item_for_dump_printf): Delete.
10319         (class dump_pretty_printer): New class.
10320         (dump_pretty_printer::dump_pretty_printer): New ctor.
10321         (dump_pretty_printer::emit_items): New member function.
10322         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
10323         function.
10324         (dump_pretty_printer::emit_item): New member function.
10325         (dump_pretty_printer::stash_item): New member function.
10326         (dump_pretty_printer::format_decoder_cb): New member function.
10327         (dump_pretty_printer::decode_format): New member function.
10328         (dump_context::dump_printf_va): Reimplement in terms of
10329         dump_pretty_printer.
10330         (dump_context::dump_printf_loc_va): Convert final param from va_list
10331         to va_list *.
10332         (dump_context::begin_scope): Reimplement call to
10333         make_item_for_dump_printf.
10334         (dump_printf): Update for change to dump_printf_va.
10335         (dump_printf_loc): Likewise.
10336         (selftest::test_capture_of_dump_calls): Convert "stmt" from
10337         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
10338         with %T, %E, and %G.
10339         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
10340         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
10341         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
10342         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
10343         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
10344         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
10345         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
10346         within a dump_printf_loc call to "%wu".
10347         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
10348         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
10349         missing space after "=".
10350         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
10351         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
10352         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
10353         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
10354         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
10355         duplicate "vectorized" from message.
10357 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
10359         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
10360         polyNxK_t element's TYPE_STRING_FLAG.
10362 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
10364         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
10365         (they were unnamed before).  Fix comments.
10367 2018-08-17  Nathan Sidwell  <nathan@acm.org>
10369         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
10371 2018-08-17  Richard Biener  <rguenther@suse.de>
10373         PR tree-optimization/86841
10374         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
10376 2018-08-17  Martin Liska  <mliska@suse.cz>
10378         * common.opt: Remove Warn, Init and Report for options with
10379         Ignore/Deprecated flag. Warning is done automatically for
10380         Deprecated flags.
10381         * config/i386/i386.opt: Likewise.
10382         * config/ia64/ia64.opt: Likewise.
10383         * config/rs6000/rs6000.opt: Likewise.
10384         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
10385         Remove usage of flag_check_pointer_bounds.
10386         * lto-wrapper.c (merge_and_complain): Do not handle
10387         OPT_fcheck_pointer_bounds.
10388         (append_compiler_options): Likewise.
10389         * opt-functions.awk: Do not handle Deprecated.
10390         * optc-gen.awk: Check that Var, Report and Init are not
10391         used for an option with Ignore/Deprecated flag.
10392         * opts-common.c (decode_cmdline_option): Do not report
10393         CL_ERR_DEPRECATED.
10394         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
10395         options.
10396         * opts.h (struct cl_option): Remove cl_deprecated flag.
10397         (CL_ERR_DEPRECATED): Remove error enum value.
10399 2018-08-17  Richard Biener  <rguenther@suse.de>
10401         PR middle-end/86505
10402         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
10403         across a va-arg-pack using call adjust its return value accordingly.
10405 2018-08-16  Martin Sebor  <msebor@redhat.com>
10407         PR tree-optimization/86853
10408         * gimple-ssa-sprintf.c (struct format_result): Rename member.
10409         (struct fmtresult): Add member and initialize it in ctors.
10410         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
10411         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
10412         (format_directive): Set POSUNDER4K when MAYFAIL is set.
10413         (parse_directive): Handle %C same as %c.
10414         (sprintf_dom_walker::compute_format_length): Adjust.
10415         (is_call_safe): Adjust.
10417 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10419         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
10420         for determining how to count the elements.
10421         * builtins.h (c_strlen): Adjust prototype.
10422         * expr.c (string_constant): Add new parameter mem_size.
10423         Set *mem_size appropriately.
10424         * expr.h (string_constant): Adjust protoype.
10425         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
10426         * gimple-fold.h (get_range_strlen): Adjust prototype.
10427         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
10428         (format_string): Call get_string_length with eltsize.
10430 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
10432         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
10433         to emit the span, rather than setting it as the prefix.
10435 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
10437         * diagnostic-show-locus.c (layout::start_annotation_line): Add
10438         "margin_char" parameter, defaulting to space.  Use it in place
10439         of pp_space for the initial part of the margin.
10440         (layout::print_leading_fixits): Use '+' when filling the margin
10441         of line-insertion fix-it hints.
10443 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
10445         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
10446         Delete.
10448 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
10450         * config/rs6000/altivec.md: Don't set length attribute to the default
10451         value.
10452         * config/rs6000/darwin.md: Ditto.
10453         * config/rs6000/dfp.md: Ditto.
10454         * config/rs6000/htm.md: Ditto.
10455         * config/rs6000/rs6000.md: Ditto.
10456         * config/rs6000/sync.md: Ditto.
10457         * config/rs6000/vsx.md: Ditto.
10459 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
10461         * config/rs6000/altivec.md: Don't set length attribute to the default
10462         value, for branch instructions.
10463         * config/rs6000/darwin.md: Ditto.
10464         * config/rs6000/rs6000.md: Ditto.
10466 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
10468         * config/rs6000/rs6000.md (length): Always define as const_int 4.
10469         (unnamed conditional branch define_insn): Set length to 4 or 8
10470         depending on offset.
10471         (<bd>_<mode>): Similar, for alternative 0.
10472         (<bd>tf_<mode>): Ditto.
10474 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
10476         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
10478 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
10480         * doc/rtl.texi: Replace old RTX class names with new names.
10483 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
10485         * expmed.h (canonicalize_comparison): New declaration.
10486         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
10487         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
10488         * optabs.c (prepare_cmp_insn): Likewise.
10489         * rtl.h (unsigned_condition_p): New function which checks if a
10490         comparison operator is unsigned.
10492 2018-08-16  Nathan Sidwell  <nathan@acm.org>
10494         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
10495         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
10497 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
10499         PR target/84711
10500         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
10501         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
10502         (mov<mov>): ..this and enable unconditionally.
10504 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
10506         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
10508 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
10510         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
10511         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
10512         "Common" with "Target".
10514 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
10516         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
10517         * doc/invoke.texi (mmitigate-rop): Remove.
10518         * config/i386/i386.c: Do not include "regrename.h".
10519         (ix86_rop_should_change_byte_p, reg_encoded_number)
10520         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
10521         Remove.
10522         (ix86_reorg): Remove call to ix86_mitigate_rop.
10523         * config/i386/i386.md (attr "modrm_class"): Remove.
10524         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
10525         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
10526         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
10528 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
10530         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
10531         allow folding of mergeh() and mergel() for the float and double types.
10532         (fold_mergehl_helper): Rework to handle building a permute tree
10533         for float vectors.
10535 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
10537         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
10538         for TARGET_SSE.
10540 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
10542         * common.opt (fdiagnostics-show-labels): New option.
10543         * diagnostic-show-locus.c (class layout_range): Add field
10544         "m_label".
10545         (class layout): Add field "m_show_labels_p".
10546         (layout_range::layout_range): Add param "label" and use it to
10547         initialize m_label.
10548         (make_range): Pass in NULL for new "label" param of layout_range's
10549         ctor.
10550         (layout::layout): Initialize m_show_labels_p.
10551         (layout::maybe_add_location_range): Pass in loc_range->m_label
10552         when constructing layout_range instances.
10553         (struct line_label): New struct.
10554         (layout::print_any_labels): New member function.
10555         (layout::print_line): Call it if label-printing is enabled.
10556         (selftest::test_one_liner_labels): New test.
10557         (selftest::test_diagnostic_show_locus_one_liner): Call it.
10558         * diagnostic.c (diagnostic_initialize): Initialize
10559         context->show_labels_p.
10560         * diagnostic.h (struct diagnostic_context): Add field
10561         "show_labels_p".
10562         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
10563         -fno-diagnostics-show-labels.
10564         * dwarf2out.c (gen_producer_string): Add
10565         OPT_fdiagnostics_show_labels to the ignored options.
10566         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
10567         param.
10568         (gcc_rich_location::maybe_add_expr): Likewise.
10569         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
10570         label" param, defaulting to NULL.
10571         (gcc_rich_location::add_expr): Add "label" param.
10572         (gcc_rich_location::maybe_add_expr): Likewise.
10573         (class text_range_label): New class.
10574         (class range_label_for_type_mismatch): New class.
10575         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
10576         of format_warning_va.
10577         (fmtwarn_n): Likewise for new params of format_warning_n_va.
10578         * lto-wrapper.c (merge_and_complain): Add
10579         OPT_fdiagnostics_show_labels to the "pick one setting" options.
10580         (append_compiler_options): Likewise to the dropped options.
10581         (append_diag_options): Likewise to the passed-on options.
10582         * opts.c (common_handle_option): Handle the new option.
10583         * selftest-diagnostic.c
10584         (test_diagnostic_context::test_diagnostic_context): Enable
10585         show_labels_p.
10586         * substring-locations.c: Include "gcc-rich-location.h".
10587         (format_warning_n_va): Add "fmt_label" and "param_label" params
10588         and use them as appropriate.
10589         (format_warning_va): Add "fmt_label" and "param_label" params,
10590         passing them on to format_warning_n_va.
10591         (format_warning_at_substring): Likewise.
10592         (format_warning_at_substring_n): Likewise.
10593         * substring-locations.h (format_warning_va): Add "fmt_label" and
10594         "param_label" params.
10595         (format_warning_n_va): Likewise.
10596         (format_warning_at_substring): Likewise.
10597         (format_warning_at_substring_n): Likewise.
10598         * toplev.c (general_init): Initialize global_dc->show_labels_p.
10600 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
10602         PR testsuite/86519
10603         * builtins.c (expand_builtin_memcmp): Do not expand the call
10604         when overflow is detected.
10606 2018-08-15  Martin Sebor  <msebor@redhat.com>
10608         PR tree-optimization/71625
10609         * config/aarch64/aarch64-builtins.c
10610         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
10612 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
10614         * config/s390/s390.c (s390_reorg): Remove loop.
10616 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
10618         * config/darwin.c
10619          (darwin_function_switched_text_sections): Delete.
10620         * gcc/config/darwin.h
10621          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
10623 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
10625         PR target/81685
10626         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
10627         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
10628         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
10630 2018-08-15  Martin Liska  <mliska@suse.cz>
10632         PR tree-optimization/86925
10633         * predict.c (expr_expected_value_1): When taking
10634         later predictor, assign also probability.
10635         Use fold_build2_initializer_loc in order to fold
10636         the expression in -frounding-math.
10638 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
10640         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
10641         patterns.
10642         (expand_vec_perm_1): Try the new method.
10644 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
10646         PR target/86547
10647         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
10648         Check whether lra_live_max_point is 0 before dividing.
10650 2018-08-14  Martin Sebor  <msebor@redhat.com>
10652         PR tree-optimization/86650
10653         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
10654         (vrp_prop::check_mem_ref): Same.
10656 2018-08-13  Liu Hao <lh_mouse@126.com>
10658         * pretty-print.c (eat_esc_sequence): Swap the foreground and
10659         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
10660         and clear it thereafter, as it only works for DBCS.
10662 2018-08-13  Liu Hao <lh_mouse@126.com>
10664         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
10665         handle returned by _get_osf_handle().
10667 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
10669         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
10670         for folding vec_perm.
10672 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
10674         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
10675         Add support for gimple-folding of vec_pack() and vec_unpack()
10676         intrinsics.
10678 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
10680         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
10681         vec_xst variants to the list.
10682         (rs6000_gimple_fold_builtin): Add support for folding unaligned
10683         vector loads and stores.
10685 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
10687         * config.gcc (rs6000-ibm-aix4.x): Delete.
10688         (rs6000-ibm-aix5.1): Delete.
10689         (rs6000-ibm-aix5.2): Delete.
10690         (rs6000-ibm-aix5.3): Delete.
10691         * config/rs6000/aix43.h: Delete.
10692         * config/rs6000/aix51.h: Delete.
10693         * config/rs6000/aix52.h: Delete.
10694         * config/rs6000/t-aix43: Delete.
10696 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
10698         * config/s390/s390.c (s390_decompose_constant_pool_ref):
10699         New function.
10700         (s390_decompose_address): Factor out constant pool ref
10701         decomposition.
10703 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
10705         * config/nds32/nds32-predicates.c
10706         (nds32_can_use_bclr_p): Change return type as bool.
10707         (nds32_can_use_bset_p): Ditto.
10708         (nds32_can_use_btgl_p): Ditto.
10709         (nds32_can_use_bitci_p): Ditto.
10710         * config/nds32/nds32-protos.h
10711         (nds32_can_use_bclr_p): Change declaration.
10712         (nds32_can_use_bset_p): Ditto.
10713         (nds32_can_use_btgl_p): Ditto.
10714         (nds32_can_use_bitci_p): Ditto.
10716 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
10718         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
10719         Support -msched-prolog-epilog option.
10720         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
10722 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
10724         * common/config/nds32/nds32-common.c
10725         (nds32_option_optimization_table): Enalbe -malways-align.
10727 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
10729         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
10730         extra_headers.
10731         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
10732         OPT_misr_secure_ case.
10733         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
10734         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
10735         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
10736         secure attribute.
10737         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
10738         (nds32_isr_info): New field security_level.
10739         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
10740         * config/nds32/nds32.md (return_internal): Consider critical attribute.
10741         * config/nds32/nds32.opt (misr-secure): New option.
10742         * config/nds32/nds32_init.inc: New file.
10743         * config/nds32/nds32_isr.h: New file.
10745 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
10747         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
10748         Update comment for atomic instructions.
10749         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
10750         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
10751         Remove.
10752         (atomic_loaddi): Revise fence expansion to only emit fence prior to
10753         load for __ATOMIC_SEQ_CST model.
10754         (atomic_loaddi_1): Remove float register target.
10755         (atomic_storedi): Handle CONST_INT values.
10756         (atomic_storedi_1): Remove float register source.  Add special case
10757         for zero value.
10758         (memory_barrier): New expander and insn.
10760 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
10762         PR tree-optimization/86835
10763         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
10764         new_stmt after def_gsi, make sure to insert new_square_stmt after
10765         that stmt, not 2 stmts before it.
10767 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
10769         PR target/82418
10770         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
10771         instead of SWI48.
10773 2018-08-10  Martin Liska  <mliska@suse.cz>
10775         PR target/83610
10776         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
10777         function type.
10778         * builtins.c (expand_builtin_expect_with_probability):
10779         New function.
10780         (expand_builtin_expect_with_probability): New function.
10781         (build_builtin_expect_predicate): Add new argumnet probability
10782         for BUILT_IN_EXPECT_WITH_PROBABILITY.
10783         (fold_builtin_expect):
10784         (fold_builtin_2):
10785         (fold_builtin_3):
10786         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
10787         * builtins.h (fold_builtin_expect): Set new argument.
10788         * doc/extend.texi: Document __builtin_expect_with_probability.
10789         * doc/invoke.texi: Likewise.
10790         * gimple-fold.c (gimple_fold_call): Pass new argument.
10791         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
10792         also BUILT_IN_EXPECT_WITH_PROBABILITY.
10793         * predict.c (get_predictor_value): New function.
10794         (expr_expected_value): Add new argument probability. Assume
10795         that predictor and probability are always non-null.
10796         (expr_expected_value_1): Likewise.  For __builtin_expect and
10797         __builtin_expect_with_probability set probability.  Handle
10798         combination in binary expressions.
10799         (tree_predict_by_opcode): Simplify code by simply calling
10800         get_predictor_value.
10801         (pass_strip_predict_hints::execute): Add handling of
10802         BUILT_IN_EXPECT_WITH_PROBABILITY.
10803         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
10804         new predictor.
10805         * tree.h (DECL_BUILT_IN_P): New function.
10807 2018-08-10  Martin Liska  <mliska@suse.cz>
10809         PR tree-optimization/85799
10810         * passes.def: Add argument for pass_strip_predict_hints.
10811         * predict.c (class pass_strip_predict_hints): Add new argument
10812         early_p.
10813         (strip_predictor_early): New function.
10814         (pass_strip_predict_hints::execute): Call the function to
10815         strip predictors.
10816         (strip_predict_hints): New function.
10817         * predict.def: Fix comment.
10819 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
10821         * Makefile.in: Clarify which tm.texi to copy over to assert the
10822         right to grant a GFDL license for all.
10824 2018-08-09  Jeff Law  <law@redhat.com>
10826         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
10827         unused variable.
10829 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
10831         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
10832         prototype.
10834 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
10836         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
10837         reductions for variable-length vectors.
10839 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
10841         PR other/84889
10842         * common.opt (fdiagnostics-show-line-numbers): New option.
10843         * diagnostic-show-locus.c (class layout): Add fields
10844         "m_show_line_numbers_p" and "m_linenum_width";
10845         (num_digits): New function.
10846         (test_num_digits): New function.
10847         (layout::layout): Initialize new fields.  Update m_x_offset
10848         logic to handle any left margin.
10849         (layout::print_source_line): Print line number when requested.
10850         (layout::start_annotation_line): New member function.
10851         (layout::print_annotation_line): Call it.
10852         (layout::print_leading_fixits): Likewise.
10853         (layout::print_trailing_fixits): Likewise.  Update calls to
10854         move_to_column for new parameter.
10855         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
10856         it to potentially call start_annotation_line.
10857         (layout::show_ruler): Call start_annotation_line.
10858         (selftest::test_line_numbers_multiline_range): New selftest.
10859         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
10860         and selftest::test_line_numbers_multiline_range.
10861         * diagnostic.c (diagnostic_initialize): Initialize
10862         show_line_numbers_p.
10863         * diagnostic.h (struct diagnostic_context): Add field
10864         "show_line_numbers_p".
10865         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
10866         -fno-diagnostics-show-line-numbers.
10867         * dwarf2out.c (gen_producer_string): Add
10868         OPT_fdiagnostics_show_line_numbers to the ignored options.
10869         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
10870         one setting" options.
10871         (append_compiler_options): Likewise to the dropped options.
10872         (append_diag_options): Likewise to the passed-on options.
10873         * opts.c (common_handle_option): Handle the new option.
10874         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
10876 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10878         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
10879         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
10880         third argument of type "const signed char" to descriptions of
10881         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
10882         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
10883         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
10884         __builtin_bcdsub_ov functions.
10886 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
10888         PR tree-optimization/86858
10889         * tree-vect-loop.c (vect_is_simple_reduction): Restore
10890         flow_bb_inside_loop_p calls.
10892 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
10894         PR tree-optimization/86871
10895         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
10896         instead of gimple_assign_lhs.
10898 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
10900         PR target/86887
10901         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
10902         register constraint to operand 0.
10903         (add<mode>3_carryinC): Likewise.
10904         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
10906 2018-08-09  Martin Liska  <mliska@suse.cz>
10908         PR c/86895
10909         * common.opt: Remove extra line.
10911 2018-08-09  Martin Liska  <mliska@suse.cz>
10913         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
10914         at the end of a line, make first letter capital and end up
10915         a sentence with a dot.
10916         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
10917         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
10918         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
10919         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
10920         (PARAM_MAX_ISL_OPERATIONS): Likewise.
10921         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
10922         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
10923         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
10924         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
10925         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
10926         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
10927         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
10928         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
10929         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
10930         (PARAM_TREE_REASSOC_WIDTH): Likewise.
10931         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
10932         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
10933         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
10935 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
10937         PR target/84332
10938         * config/s390/s390.c (s390_option_override_internal): Reduce the
10939         stack-clash-protection-probe-interval param if it would be too big
10940         for z900.
10942 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
10944         PR target/46179
10945         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
10946         * config/m68k/m68k.c (handle_move_double): Don't call
10947         m68k_final_prescan_insn.
10948         (m68k_adjust_decorated_operand): Renamed from
10949         m68k_final_prescan_insn, remove first and third operand and
10950         simplify.
10951         (print_operand): Call it.
10952         (print_operand_address): Call it.
10954 2018-08-08  Nathan Sidwell  <nathan@acm.org>
10956         * diagnostic.c (diagnostic_report_current_module): Use
10957         linemap_included_from & linemap_included_from_linemap.
10959 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
10961         * config/aarch64/aarch64-cores.def: Add phecda core.
10962         * config/aarch64/aarch64-tune.md: Regenerate.
10963         * doc/invoke.texi: Add phecda core.
10965 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
10967         PR target/85295
10968         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
10969         definitions.
10970         * config/s390/s390.md ("movti"): Add more alternatives for
10971         constant to GPR copies.
10973 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
10975         * config/s390/s390.c: Fix whitespace damage throughout the file.
10976         * config/s390/s390.h: Likewise.
10977         * config/s390/tpf.h: Likewise.
10979 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
10981         * config/s390/s390.c (s390_loadrelative_operand_p):
10982         Remove TARGET_CPU_ZARCH usages.
10983         (s390_rtx_costs): Likewise.
10984         (s390_legitimate_constant_p): Likewise.
10985         (s390_cannot_force_const_mem): Likewise.
10986         (legitimate_reload_constant_p): Likewise.
10987         (s390_preferred_reload_class): Likewise.
10988         (legitimize_pic_address): Likewise.
10989         (legitimize_tls_address): Likewise.
10990         (s390_split_branches): Removed.
10991         (s390_add_execute): Removed.
10992         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
10993         (s390_mainpool_start): Likewise.
10994         (s390_mainpool_finish): Likewise.
10995         (s390_mainpool_cancel): Removed.
10996         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
10997         (s390_chunkify_cancel): Likewise.
10998         (s390_return_addr_rtx): Likewise.
10999         (s390_register_info): Remove split_branches_pending_p uages.
11000         (s390_optimize_register_info): Likewise.
11001         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
11002         split_branches_pending_p usages.
11003         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
11004         (s390_load_got): Likewise.
11005         (s390_expand_split_stack_prologue): Likewise.
11006         (output_asm_nops): Likewise.
11007         (s390_function_profiler): Likewise.
11008         (s390_emit_call): Likewise.
11009         (s390_conditional_register_usage): Likewise.
11010         (s390_optimize_prologue): Likewise.
11011         (s390_reorg): Remove TARGET_CPU_ZARCH and
11012         split_branches_pending_p usages.
11013         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
11014         usages.
11015         (s390_output_indirect_thunk_function): Likewise.
11016         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
11017         (TARGET_CPU_ZARCH_P): Removed.
11018         (struct machine_function): Remove split_branches_pending_p.
11019         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
11021 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
11023         * common/config/s390/s390-common.c (processor_flags_table):
11024         Remove flags.
11025         * config.gcc: Remove with_arch/with_tune support.
11026         * config/s390/2064.md: Remove cpu attribute comparisons.
11027         * config/s390/driver-native.c (s390_host_detect_local_cpu):
11028         Remove MTN.
11029         * config/s390/linux.h (ASM_SPEC):
11030         Remove -march support.
11031         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
11032         Use a table to get an arch level.
11033         * config/s390/s390-opts.h (enum processor_type):
11034         Remove enum values.
11035         * config/s390/s390.c
11036         (processor_table): Remove entries, add arch_level values.
11037         (s390_issue_rate): Remove cases.
11038         (s390_option_override): Adjust
11039         s390_option_override_internal() call.
11040         (s390_option_override_internal): Remove deprecation warning.
11041         (s390_valid_target_attribute_tree): Adjust
11042         s390_option_override_internal() call.
11043         * config/s390/s390.h (struct s390_processor):
11044         Share with s390-c.c, add arch_level field.
11045         * config/s390/s390.md:
11046         Remove occurrences in cpu attribute.
11047         * config/s390/s390.opt: Remove -march/-mtune support.
11048         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
11049         * doc/invoke.texi: Remove deprecation warning.
11051 2018-08-08  Luis Machado  <luis.machado@linaro.org>
11053         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
11054         global.
11055         (qdf24xx_tunings): Set vector cost structure to
11056         qdf24xx_vector_cost.
11058         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
11059         <register_sextend>: Set to 3.
11061 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
11063         PR target/86838
11064         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
11065         * config/aarch64/aarch64-simd.md
11066         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
11067         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
11068         (aarch64_frecpx<mode>): ...this new pattern.
11069         * config/aarch64/aarch64-simd-builtins.def: Remove comment
11070         about aarch64_frecp<FRECP:frecp_suffix><mode>.
11072 2018-08-07  Martin Liska  <mliska@suse.cz>
11074         PR middle-end/83023
11075         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
11076         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
11077         * predict.def (PRED_MALLOC_NONNULL): New predictor.
11078         * doc/extend.texi: Document that malloc attribute adds
11079         hit to compiler.
11081 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
11083         PR target/86785
11084         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11085         Define to speculation_safe_value_not_needed.
11087 2018-08-06  Jeff Law  <law@redhat.com>
11089         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
11090         the vr_values instance to cprop_into_stmt.
11091         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
11092         (cprop_operand): Also query EVRP to determine if OP is a constant.
11094 2018-08-06  Nathan Sidwell  <nathan@acm.org>
11096         * diagnostic.c (diagnostic_report_current_module): Reroll
11097         included-at loop.  Translate text.
11099 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
11101         * function-tests.c (selftest::test_expansion_to_rtl): Call
11102         free_after_compilation.
11104 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
11106         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
11108 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
11110         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
11111         loops with memory block operations from getting unrolled.
11113 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
11115         PR target/86807
11116         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11117         Define to speculation_safe_value_not_needed.
11119 2018-08-06  Jeff Law  <law@redhat.com>
11121         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
11122         assert.
11124 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11126         PR target/86662
11127         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
11128         with all enabled __intN types.
11130         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
11132 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
11134         * alias.c (record_set): Check for clobber high.
11135         * cfgexpand.c (expand_gimple_stmt): Likewise.
11136         * combine-stack-adj.c (single_set_for_csa): Likewise.
11137         * combine.c (find_single_use_1): Likewise.
11138         (set_nonzero_bits_and_sign_copies): Likewise.
11139         (get_combine_src_dest): Likewise.
11140         (is_parallel_of_n_reg_sets): Likewise.
11141         (try_combine): Likewise.
11142         (record_dead_and_set_regs_1): Likewise.
11143         (reg_dead_at_p_1): Likewise.
11144         (reg_dead_at_p): Likewise.
11145         * dce.c (deletable_insn_p): Likewise.
11146         (mark_nonreg_stores_1): Likewise.
11147         (mark_nonreg_stores_2): Likewise.
11148         * df-scan.c (df_find_hard_reg_defs): Likewise.
11149         (df_uses_record): Likewise.
11150         (df_get_call_refs): Likewise.
11151         * dwarf2out.c (mem_loc_descriptor): Likewise.
11152         * haifa-sched.c (haifa_classify_rtx): Likewise.
11153         * ira-build.c (create_insn_allocnos): Likewise.
11154         * ira-costs.c (scan_one_insn): Likewise.
11155         * ira.c (equiv_init_movable_p): Likewise.
11156         (rtx_moveable_p): Likewise.
11157         (interesting_dest_for_shprep): Likewise.
11158         * jump.c (mark_jump_label_1): Likewise.
11159         * postreload-gcse.c (record_opr_changes): Likewise.
11160         * postreload.c (reload_cse_simplify): Likewise.
11161         (struct reg_use): Add source expr.
11162         (reload_combine): Check for clobber high.
11163         (reload_combine_note_use): Likewise.
11164         (reload_cse_move2add): Likewise.
11165         (move2add_note_store): Likewise.
11166         * print-rtl.c (print_pattern): Likewise.
11167         * recog.c (decode_asm_operands): Likewise.
11168         (store_data_bypass_p): Likewise.
11169         (if_test_bypass_p): Likewise.
11170         * regcprop.c (kill_clobbered_value): Likewise.
11171         (kill_set_value): Likewise.
11172         * reginfo.c (reg_scan_mark_refs): Likewise.
11173         * reload1.c (maybe_fix_stack_asms): Likewise.
11174         (eliminate_regs_1): Likewise.
11175         (elimination_effects): Likewise.
11176         (mark_not_eliminable): Likewise.
11177         (scan_paradoxical_subregs): Likewise.
11178         (forget_old_reloads_1): Likewise.
11179         * reorg.c (find_end_label): Likewise.
11180         (try_merge_delay_insns): Likewise.
11181         (redundant_insn): Likewise.
11182         (own_thread_p): Likewise.
11183         (fill_simple_delay_slots): Likewise.
11184         (fill_slots_from_thread): Likewise.
11185         (dbr_schedule): Likewise.
11186         * resource.c (update_live_status): Likewise.
11187         (mark_referenced_resources): Likewise.
11188         (mark_set_resources): Likewise.
11189         * rtl.c (copy_rtx): Likewise.
11190         * rtlanal.c (reg_referenced_p): Likewise.
11191         (single_set_2): Likewise.
11192         (noop_move_p): Likewise.
11193         (note_stores): Likewise.
11194         * sched-deps.c (sched_analyze_reg): Likewise.
11195         (sched_analyze_insn): Likewise.
11197 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
11199         * cse.c (invalidate_reg): New function extracted from...
11200         (invalidate): ...here.
11201         (canonicalize_insn): Check for clobber high.
11202         (invalidate_from_clobbers): invalidate clobber highs.
11203         (invalidate_from_sets_and_clobbers): Likewise.
11204         (count_reg_usage): Check for clobber high.
11205         (insn_live_p): Likewise.
11206         * cselib.c (cselib_expand_value_rtx_1):Likewise.
11207         (cselib_invalidate_regno): Check for clobber in setter.
11208         (cselib_invalidate_rtx): Pass through setter.
11209         (cselib_invalidate_rtx_note_stores):
11210         (cselib_process_insn): Check for clobber high.
11211         * cselib.h (cselib_invalidate_rtx): Add operand.
11213 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
11215         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
11216         (mark_not_eliminable): Likewise.
11217         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
11218         * lra-lives.c (process_bb_lives): Check for clobber high.
11219         * lra.c (new_insn_reg): Remember clobber highs.
11220         (collect_non_operand_hard_regs): Check for clobber high.
11221         (lra_set_insn_recog_data): Likewise.
11222         (add_regs_to_insn_regno_info): Likewise.
11223         (lra_update_insn_regno_info): Likewise.
11225 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
11227         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
11228         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
11230 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
11232         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
11233         (copy_insn_1): Likewise.
11234         (gen_hard_reg_clobber_high): New gen function.
11235         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
11236         * genemit.c (gen_exp): Likewise.
11237         (gen_emit_seq): Pass through info.
11238         (gen_insn): Check for CLOBBER_HIGH.
11239         (gen_expand): Pass through info.
11240         (gen_split): Likewise.
11241         (output_add_clobbers): Likewise.
11242         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
11243         (remove_clobbers): Likewise.
11244         * rtl.h (gen_hard_reg_clobber_high): New declaration.
11246 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
11248         * doc/rtl.texi (clobber_high): Add.
11249         (parallel): Add in clobber high
11250         * rtl.c (rtl_check_failed_code3): Add function.
11251         * rtl.def (CLOBBER_HIGH): Add expression.
11252         * rtl.h (RTL_CHECKC3): Add macro.
11253         (rtl_check_failed_code3): Add declaration.
11254         (XC3EXP): Add macro.
11256 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
11258         PR target/86386
11259         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
11260         cfun->machine->max_used_stack_alignment if needed.
11262 2018-08-04  Martin Sebor  <msebor@redhat.com>
11264         PR tree-optimization/86571
11265         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
11266         NaN output to 4.
11268 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
11270         PR target/86799
11271         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11272         Define.
11274 2018-08-03  Jeff Law  <law@redhat.com>
11276         PR target/86795
11277         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11278         Define to speculation_safe_value_not_needed.
11280 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
11282         * doc/gcov.texi (-x): Remove duplicate "to".
11283         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
11284         (-Wif-not-aligned): Remove duplicate "is".
11285         (-flto): Remove duplicate "the".
11286         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
11287         duplicate "v5.00.b".
11288         (MSP430 Options): Remove duplicate "and" from the description
11289         of "-mgprel-sec=regexp".
11290         (x86 Options): Remove duplicate copies of "vmldLog102" and
11291         vmlsLog104 from description of "-mveclibabi=type".
11293 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
11295         * internal-fn.h (first_commutative_argument): Declare.
11296         * internal-fn.c (first_commutative_argument): New function.
11297         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
11298         restrictions for pattern statements.  Use first_commutative_argument
11299         to look for commutative operands in calls to internal functions.
11301 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
11303         * Makefile.in (wide-int-range.o): New.
11304         * tree-vrp.c: Move all the wide_int_* functions to...
11305         * wide-int-range.cc: ...here.
11306         * tree-vrp.h: Move all the wide_int_* prototypes to...
11307         * wide-int-range.h: ...here.
11309 2018-08-03  Tom de Vries  <tdevries@suse.de>
11311         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
11312         UI_NONE.
11313         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
11314         * except.c (output_function_exception_table): Do early exit if
11315         targetm_common.except_unwind_info (&global_options) == UI_NONE.
11317 2018-08-03  Martin Liska  <mliska@suse.cz>
11319         * predict.c (dump_prediction): Change to 2 digits
11320         in fraction part.
11322 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
11324         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
11325         neon_dup_q to...
11326         (falkor_am_1_gtov_gtov): ... a new insn reservation.
11328 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
11330         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
11331         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
11332         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
11333         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
11334         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
11336 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
11338         * diagnostic-show-locus.c (diagnostic_show_locus): Use
11339         pp_take_prefix when saving the existing prefix.
11340         * diagnostic.c (diagnostic_append_note): Likewise.
11341         * langhooks.c (lhd_print_error_function): Likewise.
11342         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
11343         param's type.  Free the existing prefix.
11344         (pp_take_prefix): New function.
11345         (pretty_printer::pretty_printer): Drop the prefix parameter.
11346         Rename the length parameter to match the comment.
11347         (pretty_printer::~pretty_printer): Free the prefix.
11348         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
11349         parameter.
11350         (struct pretty_printer): Drop the "const" from "prefix" field's
11351         type and clarify memory management.
11352         (pp_set_prefix): Drop the "const" from the 2nd param.
11353         (pp_take_prefix): New decl.
11355 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
11357         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
11358         for word_mode_ok here instead of passing as argument.
11359         (expand_block_compare): Change select_block_compare_mode() call.
11360         (expand_strncmp_gpr_sequence): New function.
11361         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
11363 2018-08-02  Jeff Law  <law@redhat.com>
11365         PR target/86790
11366         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11367         Define to speculation_safe_value_not_needed.
11369         PR target/86784
11370         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11371         Define to speculation_safe_value_not_needed.
11373 2018-08-02  Tom de Vries  <tdevries@suse.de>
11375         PR target/86660
11376         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
11377         function.  Return UI_TARGET unconditionally.
11378         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
11379         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
11381 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
11383         * genemit.c (print_overload_test): Fix typo.
11385 2018-08-02  Richard Biener  <rguenther@suse.de>
11387         PR tree-optimization/86816
11388         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
11389         which checks for value availability before querying it.
11390         (gvn_uses_equal): Use it.
11391         (same_succ_hash): Likewise.
11392         (gimple_equal_p): Likewise.
11394 2018-08-02  Nick Clifton  <nickc@redhat.com>
11396         PR target/86813
11397         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11398         Define to speculation_safe_value_not_needed.
11400         PR target/86810
11401         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11402         Define to speculation_safe_value_not_needed.
11404         PR target/86810
11405         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11406         Define to speculation_safe_value_not_needed.
11408         PR target/86803
11409         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11410         Define to speculation_safe_value_not_needed.
11412         PR target/86797
11413         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11414         Define to speculation_safe_value_not_needed.
11416         PR target/86791
11417         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11418         Define to speculation_safe_value_not_needed.
11420         PR target/86789
11421         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11422         Define to speculation_safe_value_not_needed.
11424         PR target/86787
11425         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11426         Define to speculation_safe_value_not_needed.
11428         PR target/86782
11429         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
11430         speculation_safe_value_not_needed.
11432         PR target/86781
11433         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
11434         to speculation_safe_value_not_needed.
11436 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
11438         * doc/md.texi: Expand the documentation of instruction names
11439         to mention port-local uses.  Document '@' in pattern names.
11440         * read-md.h (overloaded_instance, overloaded_name): New structs.
11441         (mapping): Declare.
11442         (md_reader::handle_overloaded_name): New member function.
11443         (md_reader::get_overloads): Likewise.
11444         (md_reader::m_first_overload): New member variable.
11445         (md_reader::m_next_overload_ptr): Likewise.
11446         (md_reader::m_overloads_htab): Likewise.
11447         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
11448         m_next_overload_ptr and m_overloads_htab.
11449         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
11450         (get_mode_token, get_code_token, get_int_token): New functions.
11451         (map_attr_string): Add an optional argument that passes back
11452         the associated iterator.
11453         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
11454         (md_reader::handle_overloaded_name, add_overload_instance): New
11455         functions.
11456         (apply_iterators): Handle '@' names.  Report an error if '@'
11457         is used without iterators.
11458         (initialize_iterators): Initialize the new iterator_group fields.
11459         * genopinit.c (handle_overloaded_code_for)
11460         (handle_overloaded_gen): New functions.
11461         (main): Use them to print declarations of maybe_code_for_* and
11462         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
11463         * genemit.c (print_overload_arguments, print_overload_test)
11464         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
11465         (main): Use it to print definitions of maybe_code_for_* and
11466         maybe_gen_* functions.
11467         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
11468         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
11469         instead of explicit mode checks.
11470         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
11471         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
11472         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
11473         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
11474         (aarch64_expand_compare_and_swap): Likewise
11475         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
11476         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
11477         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
11478         (aarch64_constant_pool_reload_icode): Delete.
11479         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
11480         instead of aarch64_constant_pool_reload_icode.  Use
11481         code_for_aarch64_reload_mov instead of explicit mode checks.
11482         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
11483         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
11484         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
11485         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
11486         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
11487         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
11488         (aarch64_atomic_load_op_code): Delete.
11489         (aarch64_emit_atomic_load_op): Likewise.
11490         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
11491         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
11492         instead of aarch64_emit_atomic_load_op.
11493         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
11494         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
11495         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
11496         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
11497         character before the pattern name.
11498         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
11499         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
11500         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
11501         (aarch64_frecps<mode>): Likewise.
11502         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
11503         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
11504         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
11505         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
11506         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
11508 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
11510         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
11511         Allow HFmode constants if TARGET_FP_F16INST.
11513 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
11515         PR target/86014
11516         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
11517         No longer check last store for clobber of address register.
11519 2018-08-02  Martin Liska  <mliska@suse.cz>
11521         PR gcov-profile/86817
11522         * gcov.c (process_all_functions): New function.
11523         (main): Call it.
11524         (process_file): Move functions processing to
11525         process_all_functions.
11527 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
11529         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
11530         "const" to the "gimple *" and "rtx_insn *" parameters.
11531         * dumpfile.h (dump_user_location_t::dump_user_location_t):
11532         Likewise.
11533         (dump_location_t::dump_location_t): Likewise.
11535 2018-08-01  Martin Sebor  <msebor@redhat.com>
11537         PR tree-optimization/86650
11538         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
11539         rather than a "gcall *".  Directly pass the data of interest
11540         to percent_K_format, rather than building a temporary CALL_EXPR
11541         to hold it.
11542         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
11543         (gimple_fold_builtin_strncat): Adjust.
11544         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
11545         gcall* argument with gimple*.
11546         * gimple-ssa-warn-restrict.c (check_call): Same.
11547         (wrestrict_dom_walker::before_dom_children): Same.
11548         (builtin_access::builtin_access): Same.
11549         (check_bounds_or_overlap): Same
11550         (maybe_diag_overlap): Same.
11551         (maybe_diag_offset_bounds): Same.
11552         * tree-diagnostic.c (default_tree_printer): Move usage of
11553         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
11554         to this callsite.
11555         * tree-pretty-print.c (percent_K_format): Add argument.
11556         * tree-pretty-print.h: Add argument.
11557         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
11558         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
11559         (maybe_diag_stxncpy_trunc): Same.
11560         (handle_builtin_stxncpy): Same.
11561         (handle_builtin_strcat): Same.
11563 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11565         * match.pd: Optimise pointer range checks.
11567 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11569         PR tree-optimization/86758
11570         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
11571         to remove pattern statements.
11573 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11575         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
11576         result of dfs_enumerate_from when constructing stmt_vec_infos,
11577         instead of additionally calling get_loop_body.
11579 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11581         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
11582         parameter.
11583         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
11584         When creating an iv, assert that the step is not known to be zero.
11585         (vect_setup_realignment): Update call accordingly.
11586         * tree-vect-stmts.c (vectorizable_store): Likewise.
11587         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
11589 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11591         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
11592         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
11593         (vectorizable_reduction): Likewise.
11594         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11595         (vect_detect_hybrid_slp_stmts): Likewise.
11596         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11598 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
11600         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
11601         (wide_int_set_zero_nonzero_bits): ...this.
11602         (zero_nonzero_bits_from_vr): Rename to...
11603         (vrp_set_zero_nonzero_bits): ...this.
11604         (extract_range_from_multiplicative_op_1): Abstract wide int
11605         code...
11606         (wide_int_range_multiplicative_op): ...here.
11607         (extract_range_from_binary_expr_1): Extract wide int binary
11608         operations into their own functions.
11609         (wide_int_range_lshift): New.
11610         (wide_int_range_can_optimize_bit_op): New.
11611         (wide_int_range_shift_undefined_p): New.
11612         (wide_int_range_bit_xor): New.
11613         (wide_int_range_bit_ior): New.
11614         (wide_int_range_bit_and): New.
11615         (wide_int_range_trunc_mod): New.
11616         (extract_range_into_wide_ints): New.
11617         (vrp_shift_undefined_p): New.
11618         (extract_range_from_multiplicative_op): New.
11619         (vrp_can_optimize_bit_op): New.
11620         * tree-vrp.h (value_range::dump): New.
11621         (wide_int_range_multiplicative_op): New.
11622         (wide_int_range_lshift):New.
11623         (wide_int_range_shift_undefined_p): New.
11624         (wide_int_range_bit_xor): New.
11625         (wide_int_range_bit_ior): New.
11626         (wide_int_range_bit_and): New.
11627         (wide_int_range_trunc_mod): New.
11628         (zero_nonzero_bits_from_bounds): Rename to...
11629         (wide_int_set_zero_nonzero_bits): ...this.
11630         (zero_nonzero_bits_from_vr): Rename to...
11631         (vrp_set_zero_nonzero_bits): ...this.
11632         (range_easy_mask_min_max): Rename to...
11633         (wide_int_range_can_optimize_bit_op): this.
11634         * vr-values.c (simplify_bit_ops_using_ranges): Rename
11635         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
11637 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11639         * tree-vectorizer.h (vect_orig_stmt): New function.
11640         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
11641         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
11642         (vect_create_epilog_for_reduction): Likewise.
11643         (vectorizable_live_operation): Likewise.
11644         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
11645         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
11646         * tree-vect-stmts.c (vectorizable_call): Likewise.
11647         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
11649 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11651         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
11652         argument.
11653         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
11654         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
11655         (vect_transform_loop): Likewise.
11656         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
11658 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11660         * tree-vectorizer.h (vect_schedule_slp): Return void.
11661         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
11662         (vect_schedule_slp): Likewise.
11664 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11666         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
11667         argument.
11668         (vect_transform_loop): Update calls accordingly.  Schedule SLP
11669         instances before the main loop, if any exist.
11671 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
11673         PR tree-optimization/86749
11674         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
11675         If the lhs is used in a COND_EXPR, check that it is being used
11676         as the "then" or "else" value.
11678 2018-08-01  Tom de Vries  <tdevries@suse.de>
11680         PR target/86800
11681         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
11682         speculation_safe_value_not_needed.
11684 2018-08-01  Richard Biener  <rguenther@suse.de>
11686         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
11687         as base and offset.
11689 2018-08-01  Martin Liska  <mliska@suse.cz>
11691         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
11692         format how successful transformation is dumped.
11693         (gimple_mod_pow2_value_transform): Likewise.
11694         (gimple_mod_subtract_transform): Likewise.
11695         (gimple_stringops_transform): Likewise.
11697 2018-08-01  Martin Liska  <mliska@suse.cz>
11699         PR value-prof/35543
11700         * value-prof.c (interesting_stringop_to_profile_p):
11701         Simplify the code and add BUILT_IN_MEMMOVE.
11702         (gimple_stringops_transform): Likewise.
11704 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
11706         * config/aarch64/aarch64-simd.md
11707         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
11708         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
11709         use GPI iterator instead of SI mode.
11711 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
11713         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
11714         rs6000_speculation_barrier.
11715         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
11716         new barrier pattern name.
11718 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
11720         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
11721         (speculation_barrier): New insn.
11723 2018-08-01  Richard Biener  <rguenther@suse.de>
11725         PR bootstrap/86724
11726         * graphite.h: Include isl/id.h and isl/space.h to allow build
11727         with ISL 0.20.
11729 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
11731         PR target/86651
11732         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
11733         mode for COFF targets.
11734         * defaults.h (TARGET_COFF): Define.
11735         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
11736         TARGET_COFF): Define.
11737         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
11738         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
11739         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
11741 2018-07-31  Alexandre Oliva <oliva@adacore.com>
11743         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
11744         * gimple-streamer-out.c (output_bb): Save it.
11745         * lto-streamer-in.c (input_struct_function_base): Restore
11746         instance discriminator if available.  Create map on demand.
11747         * lto-streamer-out.c (output_struct_function_base): Save it if
11748         available.
11749         * final.c (decl_to_instance_map): Document LTO strategy.
11751 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
11752             Olivier Hainque  <hainque@adacore.com>
11754         * debug.h (decl_to_instance_map_t): New type.
11755         (decl_to_instance_map): Declare.
11756         (maybe_create_decl_to_instance_map): New inline function.
11757         * final.c (bb_discriminator, last_bb_discriminator): New statics,
11758         to track basic block discriminators.
11759         (final_start_function_1): Initialize them.
11760         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
11761         bb_discriminator.
11762         (decl_to_instance_map): New variable.
11763         (map_decl_to_instance, maybe_set_discriminator): New functions.
11764         (notice_source_line): Set discriminator.
11766 2018-07-31  Ian Lance Taylor  <iant@golang.org>
11768         * targhooks.c (default_have_speculation_safe_value): Add
11769         ATTRIBUTE_UNUSED.
11771 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
11773         * dump-context.h: Include "pretty-print.h".
11774         (dump_context::refresh_dumps_are_enabled): New decl.
11775         (dump_context::emit_item): New decl.
11776         (class dump_context): Add fields "m_test_pp" and
11777         "m_test_pp_flags".
11778         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
11779         (temp_dump_context::get_dumped_text): New decl.
11780         (class temp_dump_context): Add field "m_pp".
11781         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
11782         (dump_context::refresh_dumps_are_enabled): ...and add a test for
11783         m_test_pp.
11784         (set_dump_file): Update for above change.
11785         (set_alt_dump_file): Likewise.
11786         (dump_loc): New overload, taking a pretty_printer *.
11787         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
11788         to any test pretty-printer.
11789         (make_item_for_dump_gimple_stmt): New function, adapted from
11790         optinfo::add_gimple_stmt.
11791         (dump_context::dump_gimple_stmt): Call it, and use the result,
11792         eliminating the direct usage of dump_file and alt_dump_file in
11793         favor of indirectly using them via emit_item.
11794         (make_item_for_dump_gimple_expr): New function, adapted from
11795         optinfo::add_gimple_expr.
11796         (dump_context::dump_gimple_expr): Call it, and use the result,
11797         eliminating the direct usage of dump_file and alt_dump_file in
11798         favor of indirectly using them via emit_item.
11799         (make_item_for_dump_generic_expr): New function, adapted from
11800         optinfo::add_tree.
11801         (dump_context::dump_generic_expr): Call it, and use the result,
11802         eliminating the direct usage of dump_file and alt_dump_file in
11803         favor of indirectly using them via emit_item.
11804         (make_item_for_dump_printf_va): New function, adapted from
11805         optinfo::add_printf_va.
11806         (make_item_for_dump_printf): New function.
11807         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
11808         and use the result, eliminating the direct usage of dump_file and
11809         alt_dump_file in favor of indirectly using them via emit_item.
11810         (make_item_for_dump_dec): New function.
11811         (dump_context::dump_dec): Call it, and use the result,
11812         eliminating the direct usage of dump_file and alt_dump_file in
11813         favor of indirectly using them via emit_item.
11814         (make_item_for_dump_symtab_node): New function, adapted from
11815         optinfo::add_symtab_node.
11816         (dump_context::dump_symtab_node): Call it, and use the result,
11817         eliminating the direct usage of dump_file and alt_dump_file in
11818         favor of indirectly using them via emit_item.
11819         (dump_context::begin_scope): Reimplement, avoiding direct usage
11820         of dump_file and alt_dump_file in favor of indirectly using them
11821         via emit_item.
11822         (dump_context::emit_item): New member function.
11823         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
11824         Set up test pretty-printer on the underlying context.  Call
11825         refresh_dumps_are_enabled.
11826         (temp_dump_context::~temp_dump_context): Call
11827         refresh_dumps_are_enabled.
11828         (temp_dump_context::get_dumped_text): New member function.
11829         (selftest::verify_dumped_text): New function.
11830         (ASSERT_DUMPED_TEXT_EQ): New macro.
11831         (selftest::test_capture_of_dump_calls): Run all tests twice, with
11832         and then without optinfo enabled.  Add uses of
11833         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
11834         * dumpfile.h: Update comment for the dump_* API.
11835         * optinfo-emit-json.cc
11836         (selftest::test_building_json_from_dump_calls): Update for new
11837         param for temp_dump_context ctor.
11838         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
11839         and "m_owned" field.
11840         (optinfo_item::~optinfo_item): Likewise.
11841         (optinfo::add_item): New member function.
11842         (optinfo::emit): Update comment.
11843         (optinfo::add_string): Delete.
11844         (optinfo::add_printf): Delete.
11845         (optinfo::add_printf_va): Delete.
11846         (optinfo::add_gimple_stmt): Delete.
11847         (optinfo::add_gimple_expr): Delete.
11848         (optinfo::add_tree): Delete.
11849         (optinfo::add_symtab_node): Delete.
11850         (optinfo::add_dec): Delete.
11851         * optinfo.h (class dump_context): New forward decl.
11852         (optinfo::add_item): New decl.
11853         (optinfo::add_string): Delete.
11854         (optinfo::add_printf): Delete.
11855         (optinfo::add_printf_va): Delete.
11856         (optinfo::add_gimple_stmt): Delete.
11857         (optinfo::add_gimple_expr): Delete.
11858         (optinfo::add_tree): Delete.
11859         (optinfo::add_symtab_node): Delete.
11860         (optinfo::add_dec): Delete.
11861         (optinfo::add_poly_int): Delete.
11862         (optinfo_item::optinfo_item): Remove "owned" param.
11863         (class optinfo_item): Remove field "m_owned".
11865 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11867         PR middle-end/86705
11868         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
11869         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
11870         requested variable alignment.
11871         (expand_one_ssa_partition): Likewise.
11872         (expand_one_var): Likewise.
11874 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11876         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
11877         to speculation_safe_value_not_needed.
11879 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11881         * targhooks.h (speculation_safe_value_not_needed): New prototype.
11882         * targhooks.c (speculation_safe_value_not_needed): New function.
11883         * target.def (have_speculation_safe_value): Update documentation.
11884         * doc/tm.texi: Regenerated.
11886 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11888         * config/aarch64/iterators.md (ALLI_TI): New iterator.
11889         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
11890         expand.
11891         (despeculate_copy<ALLI:mode>_insn): New insn.
11892         (despeculate_copyti_insn): New insn.
11893         (despeculate_simple<ALLI:mode>): New insn
11894         (despeculate_simpleti): New insn.
11895         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
11896         function.
11897         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
11898         aarch64_speculation_safe_value.
11899         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
11901 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11903         * config/aarch64/aarch64-speculation.cc: New file.
11904         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
11905         before pass_reorder_blocks.
11906         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
11907         prototype.
11908         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
11909         X14 and X15 when tracking speculation.
11910         * config/aarch64/aarch64.md (register name constants): Add
11911         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
11912         (unspec): Add UNSPEC_SPECULATION_TRACKER.
11913         (speculation_barrier): New insn attribute.
11914         (cmp<mode>): Allow SP in comparisons.
11915         (speculation_tracker): New insn.
11916         (speculation_barrier): Add speculation_barrier attribute.
11917         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
11918         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
11919         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
11921 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11923         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
11924         aarch64_track_speculation is true.
11925         (tb<optab><mode>1): Likewise.
11926         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
11927         generate CB[N]Z when tracking speculation.
11928         (aarch64_split_compare_and_swap): Likewise.
11929         (aarch64_split_atomic_op): Likewise.
11931 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11933         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
11935 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11937         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
11938         (speculation_barrier): New insn.
11940 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11942         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
11943         * config/arm/arm.md (speculation_barrier): New expand.
11944         (speculation_barrier_insn): New pattern.
11946 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
11948         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
11949         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
11950         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
11951         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
11952         list.
11953         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
11954         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
11955         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
11956         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
11957         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
11958         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
11959         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
11960         * builtins.c (expand_speculation_safe_value): New function.
11961         (expand_builtin): Call it.
11962         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
11963         * doc/extend.texi: Document __builtin_speculation_safe_value.
11964         * doc/md.texi: Document "speculation_barrier" pattern.
11965         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
11966         TARGET_HAVE_SPECULATION_SAFE_VALUE.
11967         * doc/tm.texi: Regenerated.
11968         * target.def (have_speculation_safe_value, speculation_safe_value): New
11969         hooks.
11970         * targhooks.c (default_have_speculation_safe_value): New function.
11971         (default_speculation_safe_value): New function.
11972         * targhooks.h (default_have_speculation_safe_value): Add prototype.
11973         (default_speculation_safe_value): Add prototype.
11975 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
11977         * dump-context.h (dump_context::dump_loc): New decl.
11978         * dumpfile.c (dump_context::dump_loc): New member function.
11979         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
11980         and dump_gimple_stmt.
11981         (dump_context::dump_gimple_expr_loc): Likewise, using
11982         dump_gimple_expr.
11983         (dump_context::dump_generic_expr_loc): Likewise, using
11984         dump_generic_expr.
11985         (dump_context::dump_printf_loc_va): Likewise, using
11986         dump_printf_va.
11987         (dump_context::begin_scope): Explicitly using the global function
11988         "dump_loc", rather than the member function.
11990 2018-07-31  Martin Sebor  <msebor@redhat.com>
11992         PR tree-optimization/86741
11993         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
11995 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
11997         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
11998         depend on whether prefetch instructions will be emitted or not.
11999         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
12000         will be emitted or not.
12001         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
12002         (TARGET_SETMEM_PFD): New macros.
12004 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12006         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
12007         (NULL_STMT_VEC_INFO): Delete.
12008         (stmt_vec_info::operator*): Likewise.
12009         (stmt_vec_info::operator gimple *): Likewise.
12010         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
12011         of NULL_STMT_VEC_INFO.
12012         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
12013         (vect_reassociating_reduction_p): Likewise.
12014         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
12015         (vectorizable_store): Likewise.
12016         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
12017         (vec_info::free_stmt_vec_infos): Likewise.
12019 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12021         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
12022         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
12024 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12026         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
12027         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
12028         (vec_info::free_stmt_vec_info): New private member functions.
12029         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
12030         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
12031         * tree-parloops.c (gather_scalar_reductions): Remove calls to
12032         set_stmt_vec_info_vec and free_stmt_vec_infos.
12033         * tree-vect-loop.c (_loop_vec_info): Remove call to
12034         set_stmt_vec_info_vec.
12035         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
12036         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
12037         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
12038         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
12039         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
12040         assignments in {vec_info::,}new_stmt_vec_info that are redundant
12041         with the clearing in the xcalloc.
12042         (stmt_vec_info_vec): Delete.
12043         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
12044         (vectorize_loops): Likewise.
12045         (vec_info::~vec_info): Remove argument from call to
12046         free_stmt_vec_infos.
12047         (vec_info::add_stmt): Remove vinfo argument from call to
12048         new_stmt_vec_info.
12050 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12052         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
12053         rather than a gimple stmt.
12054         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
12055         information for pattern statements when passed the original
12056         statement; instead wait to be passed the pattern statement itself.
12057         Don't call set_vinfo_for_stmt here.
12058         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
12059         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
12060         stmt_vec_infos here.
12061         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
12062         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
12063         stmt_vec_infos entry.
12065 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12067         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
12068         * tree-vectorizer.c (vec_info::replace_stmt): New function.
12069         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
12070         * tree-vect-stmts.c (vectorizable_call): Likewise.
12071         (vectorizable_simd_clone_call): Likewise.
12073 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12075         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
12076         * tree-vectorizer.c (vec_info::remove_stmt): New function.
12077         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
12078         * tree-vect-loop.c (vect_transform_loop): Likewise.
12079         * tree-vect-slp.c (vect_schedule_slp): Likewise.
12080         * tree-vect-stmts.c (vect_remove_stores): Likewise.
12082 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12084         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
12085         (vect_dr_stmt): Delete.
12086         * tree-vectorizer.c (vec_info::lookup_dr): New function.
12087         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
12088         of DR_VECT_AUX.
12089         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12090         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
12091         (vect_verify_datarefs_alignment, vect_peeling_supportable)
12092         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
12093         (vect_analyze_data_refs): Likewise.
12094         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
12095         argument.
12096         (vect_find_same_alignment_drs): Likewise.
12097         (vect_slp_analyze_node_dependences): Update calls accordingly.
12098         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
12099         instead of DR_VECT_AUX.
12100         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
12101         of a vector data references.  Use vec_info::lookup_dr instead of
12102         DR_VECT_AUX.
12103         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
12104         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
12105         instead of DR_VECT_AUX.
12107 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12109         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
12110         dr_vec_info.
12111         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
12112         accordingly.
12113         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
12114         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
12115         (vect_gen_prolog_loop_niters): Likewise.
12117 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12119         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
12120         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
12121         (vect_known_alignment_in_bytes, vect_dr_behavior)
12122         (vect_get_scalar_dr_size): Take references as dr_vec_infos
12123         instead of data_references.  Update calls to other routines for
12124         which the same change has been made.
12125         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
12126         dr_vec_infos instead of stmt_vec_infos.
12127         (vect_analyze_data_ref_dependence): Update call accordingly.
12128         (vect_slp_analyze_data_ref_dependence)
12129         (vect_record_base_alignments): Use DR_VECT_AUX.
12130         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
12131         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
12132         (vector_alignment_reachable_p, vect_get_data_access_cost)
12133         (vect_peeling_supportable, vect_analyze_group_access_1)
12134         (vect_analyze_group_access, vect_analyze_data_ref_access)
12135         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
12136         (vect_compile_time_alias, vect_small_gap_p)
12137         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
12138         (vect_supportable_dr_alignment): Take references as dr_vec_infos
12139         instead of data_references.  Update calls to other routines for
12140         which the same change has been made.
12141         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
12142         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
12143         (vect_slp_analyze_and_verify_node_alignment)
12144         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
12145         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
12146         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
12147         above changes.
12148         (_vect_peel_info::dr): Replace with...
12149         (_vect_peel_info::dr_info): ...this new field.
12150         (vect_peeling_hash_get_most_frequent)
12151         (vect_peeling_hash_choose_best_peeling): Update accordingly.
12152         (vect_peeling_hash_get_lowest_cost):
12153         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
12154         routines for which the same change has been made.
12155         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
12156         data_reference.
12157         * tree-vect-loop-manip.c (get_misalign_in_elems)
12158         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
12159         above changes.
12160         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
12161         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
12162         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
12163         (get_group_load_store_type, get_negative_load_store_type)
12164         (vect_get_data_ptr_increment, vectorizable_store)
12165         (vectorizable_load): Likewise.
12166         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
12167         Update calls to other routines for which the same change has been made.
12169 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12171         * tree-vectorizer.h (vec_info::move_dr): New member function.
12172         (dataref_aux): Rename to...
12173         (dr_vec_info): ...this and add "dr" and "stmt" fields.
12174         (_stmt_vec_info::dr_aux): Update accordingly.
12175         (_stmt_vec_info::data_ref_info): Delete.
12176         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
12177         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
12178         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
12179         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
12180         of data_ref.
12181         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
12182         (STMT_VINFO_DR_INFO): New macro.
12183         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
12184         (set_dr_misalignment): Update after rename of dataref_aux.
12185         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
12186         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
12187         initialization of STMT_VINFO_DATA_REF.
12188         * tree-vectorizer.c (vec_info::move_dr): New function.
12189         * tree-vect-patterns.c (vect_recog_bool_pattern)
12190         (vect_recog_mask_conversion_pattern)
12191         (vect_recog_gather_scatter_pattern): Use it.
12192         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
12193         the "dr" and "stmt" fields of dr_vec_info instead of
12194         STMT_VINFO_DATA_REF.
12196 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12198         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
12199         (is_pattern_stmt_p): Use it.
12200         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
12201         on pattern statements.
12203 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12205         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
12206         original stmt as a stmt_vec_info rather than a gimple stmt.
12207         (vect_pattern_recog_1): Take the statement directly as a
12208         stmt_vec_info, rather than via a gimple_stmt_iterator.
12209         Update call to vect_mark_pattern_stmts.
12210         (vect_pattern_recog): Update calls accordingly.
12212 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12214         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
12215         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
12216         a vect_def_type for the first argument.
12217         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
12218         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
12219         operand if it isn't defined by a vectorized statement.
12220         (vect_build_gather_load_calls): Remove the mask_dt argument and
12221         update calls to vect_get_vec_def_for_stmt_copy.
12222         (vectorizable_bswap): Likewise the dt argument.
12223         (vectorizable_call): Update calls to vectorizable_bswap and
12224         vect_get_vec_def_for_stmt_copy.
12225         (vectorizable_simd_clone_call, vectorizable_assignment)
12226         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
12227         (vectorizable_comparison): Update calls to
12228         vect_get_vec_def_for_stmt_copy.
12229         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
12230         vect_is_simple_use.
12231         (vect_get_loop_based_defs): Remove dt argument and update call
12232         to vect_get_vec_def_for_stmt_copy.
12233         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
12234         and vect_get_vec_def_for_stmt_copy.
12235         (vectorizable_load): Update calls to vect_build_gather_load_calls
12236         and vect_get_vec_def_for_stmt_copy.
12237         * tree-vect-loop.c (vect_create_epilog_for_reduction)
12238         (vectorizable_reduction, vectorizable_live_operation): Update calls
12239         to vect_get_vec_def_for_stmt_copy.
12241 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12243         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
12244         and gimple stmt arguments with a stmt_vec_info.
12245         (vect_record_base_alignments): Update calls accordingly.
12246         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
12247         and gimple stmt arguments with a stmt_vec_info.
12248         (vect_build_slp_tree_1): Remove vinfo argument and update call
12249         to vect_record_max_nunits.
12250         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
12251         and vect_record_max_nunits.
12253 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12255         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
12256         file and take a stmt_vec_info instead of a gimple stmt.
12257         (supportable_widening_operation, vect_finish_replace_stmt)
12258         (vect_finish_stmt_generation, vect_get_store_rhs)
12259         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
12260         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
12261         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
12262         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
12263         (vect_create_data_ref_ptr, bump_vector_ptr)
12264         (vect_permute_store_chain, vect_setup_realignment)
12265         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
12266         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
12267         (vectorizable_reduction, vectorizable_induction)
12268         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
12269         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
12270         than gimple stmts as arguments.
12271         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
12272         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
12273         (can_group_stmts_p, vect_check_gather_scatter)
12274         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
12275         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
12276         (vect_permute_load_chain, vect_shift_permute_load_chain)
12277         (vect_transform_grouped_load)
12278         (vect_record_grouped_load_vectors): Likewise.
12279         * tree-vect-loop.c (vect_fixup_reduc_chain)
12280         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
12281         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
12282         (vectorizable_reduction, vectorizable_induction)
12283         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
12284         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
12285         (vect_get_load_store_mask): Likewise.
12286         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
12287         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
12288         * tree-vect-stmts.c (vect_mark_relevant)
12289         (is_simple_and_all_uses_invariant)
12290         (exist_non_indexing_operands_for_use_p, process_use)
12291         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
12292         (vect_get_vec_def_for_operand, vect_get_vec_defs)
12293         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
12294         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
12295         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
12296         (get_negative_load_store_type, get_load_store_type)
12297         (vect_check_load_store_mask, vect_check_store_rhs)
12298         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
12299         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
12300         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
12301         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
12302         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
12303         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
12304         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
12305         (supportable_widening_operation): Likewise.
12307 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12309         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
12310         a stmt_vec_info instead of a gcall.
12311         (vect_check_gather_scatter): Update call accordingly.
12312         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
12313         of a gphi.
12314         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
12315         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
12316         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
12317         instead of a gimple stmt.
12318         (vect_transform_loop): Update calls accordingly.
12319         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
12320         stmt_vec_infos instead of gimple stmts.
12321         (vect_analyze_slp_instance): Update use accordingly.
12322         * tree-vect-stmts.c (read_vector_array, write_vector_array)
12323         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
12324         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
12325         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
12326         (vect_gen_widened_results_half, vect_get_loop_based_defs)
12327         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
12328         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
12329         down to subroutines.
12331 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12333         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
12334         of the worklist from a vector of gimple stmts to a vector of
12335         stmt_vec_infos.
12336         * tree-vect-stmts.c (vect_mark_relevant, process_use)
12337         (vect_mark_stmts_to_be_vectorized): Likewise
12339 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12341         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
12342         statement before passing it to vect_analyze_stmt.
12343         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
12344         the chain of phi vector definitions.  Track the exit phi via its
12345         stmt_vec_info.
12346         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
12347         STMT_VINFO_REDUC_DEF.
12348         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
12349         stmt_vec_infos to handle the statement chains.
12350         (vect_get_slp_defs): Record the first statement in the node
12351         using a stmt_vec_info.
12352         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
12353         statements here and pass their stmt_vec_info down to subroutines.
12354         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
12355         down to vect_finish_stmt_generation.
12356         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
12357         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
12358         stmt_vec_infos to subroutines.
12359         (vect_remove_stores): Use stmt_vec_infos to handle the statement
12360         chains.
12362 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12364         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
12365         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
12366         (vect_permute_store_chain, vect_setup_realignment)
12367         (vect_permute_load_chain, vect_shift_permute_load_chain)
12368         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
12369         stmts internally, and when passing values to other vectorizer routines.
12370         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
12371         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
12372         (vect_analyze_loop_operations, get_initial_def_for_reduction)
12373         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
12374         (vectorizable_reduction, vectorizable_induction)
12375         (vectorizable_live_operation, vect_transform_loop_stmt)
12376         (vect_transform_loop): Likewise.
12377         * tree-vect-patterns.c (vect_reassociating_reduction_p)
12378         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
12379         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
12380         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12381         (vect_slp_analyze_node_operations_1): Likewise.
12382         * tree-vect-stmts.c (vect_mark_relevant, process_use)
12383         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
12384         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
12385         (vect_finish_stmt_generation_1, get_group_load_store_type)
12386         (get_load_store_type, vect_build_gather_load_calls)
12387         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
12388         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
12389         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
12390         (vectorizable_store, vectorizable_load, vectorizable_condition)
12391         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
12392         (supportable_widening_operation): Likewise.
12393         (vect_get_vector_types_for_stmt): Likewise.
12394         * tree-vectorizer.h (vect_dr_behavior): Likewise.
12396 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12398         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12399         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
12400         (vect_permute_store_chain, vect_permute_load_chain)
12401         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
12402         repeated stmt_vec_info lookups.
12403         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
12404         (vect_update_ivs_after_vectorizer): Likewise.
12405         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
12406         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
12407         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
12408         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12409         (vect_bb_slp_scalar_cost): Likewise.
12410         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
12412 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12414         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
12415         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
12416         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
12417         to get gassigns and gcalls, rather than operating on generc gimple
12418         stmts.
12419         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
12420         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
12421         (vectorizable_load, vect_analyze_stmt): Likewise.
12422         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
12424 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12426         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
12427         return stmt_vec_infos rather than gimple stmts.  Do not accept
12428         null arguments.
12429         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
12430         of a gimple stmt.
12431         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
12432         Update use of get_later_stmt.
12433         (vect_get_constant_vectors): Update call accordingly.
12434         (vect_schedule_slp_instance): Likewise
12435         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
12436         (vect_slp_analyze_instance_dependence): Likewise.
12437         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
12439 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12441         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
12442         (stmt_info_for_cost::stmt_info): ...this new field.
12443         (add_stmt_costs): Update accordingly.
12444         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
12445         (vect_get_known_peeling_cost): Likewise.
12446         (vect_estimate_min_profitable_iters): Likewise.
12447         * tree-vect-stmts.c (record_stmt_cost): Likewise.
12449 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12451         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
12452         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
12453         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
12454         accordingly.
12455         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
12457 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12459         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
12460         a gimple stmt to a stmt_vec_info.
12461         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
12463 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12465         * tree-vectorizer.h (vec_info::grouped_stores): Change from
12466         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
12467         (_loop_vec_info::reduction_chains): Likewise.
12468         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
12469         accordingly.
12470         * tree-vect-slp.c (vect_analyze_slp): Likewise.
12472 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12474         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
12475         a gimple stmt to a stmt_vec_info.
12476         (_stmt_vec_info::next_element): Likewise.
12477         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
12478         (vect_slp_analyze_and_verify_node_alignment)
12479         (vect_analyze_group_access_1, vect_analyze_group_access)
12480         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
12481         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
12482         (vect_supportable_dr_alignment): Update accordingly.
12483         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
12484         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
12485         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
12486         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
12487         * tree-vect-slp.c (vect_build_slp_tree_1)
12488         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
12489         (vect_split_slp_store_group, vect_analyze_slp_instance)
12490         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
12491         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
12492         (get_group_load_store_type, get_load_store_type)
12493         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
12494         (vect_transform_stmt, vect_remove_stores): Likewise.
12496 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12498         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
12499         than a gimple stmt.
12500         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12501         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
12502         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
12503         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
12504         (vector_alignment_reachable_p, vect_get_data_access_cost)
12505         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
12506         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
12507         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
12508         (vect_analyze_group_access_1, vect_analyze_group_access)
12509         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
12510         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
12511         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
12512         result of vect_dr_stmt and use the stmt_vec_info instead of
12513         the associated gimple stmt.
12514         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
12515         (vect_gen_prolog_loop_niters): Likewise.
12516         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
12518 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12520         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
12521         to a vec<stmt_vec_info>.
12522         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
12523         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
12524         vec<stmt_vec_info>.
12525         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
12526         to a vec<stmt_vec_info>.
12527         (bst_traits::value_type, bst_traits::value_type): Likewise.
12528         (bst_traits::hash): Update accordingly.
12529         (vect_get_and_check_slp_defs): Change the stmts parameter from
12530         a vec<gimple *> to a vec<stmt_vec_info>.
12531         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
12532         (vect_build_slp_tree): Likewise.
12533         (vect_build_slp_tree_2): Likewise.  Update uses of
12534         SLP_TREE_SCALAR_STMTS.
12535         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
12536         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
12537         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
12538         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
12539         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
12540         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
12541         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
12542         (vect_get_constant_vectors, vect_get_slp_defs)
12543         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
12544         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
12545         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
12546         instead of gimple stmts.
12547         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
12548         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
12549         (vect_slp_analyze_instance_dependence): Update uses of
12550         SLP_TREE_SCALAR_STMTS.
12551         (vect_slp_analyze_and_verify_node_alignment): Likewise.
12552         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
12553         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
12554         (get_initial_defs_for_reduction): Likewise.
12555         (vect_create_epilog_for_reduction): Likewise.
12556         (vectorize_fold_left_reduction): Likewise.
12557         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
12558         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
12559         (can_vectorize_live_stmts): Likewise.
12561 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12563         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
12564         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
12565         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
12566         than gimple stmts.
12567         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
12568         of a gimple stmt.
12569         (gather_scalar_reductions): Update after above interface changes.
12570         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
12571         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
12572         than gimple stmts.
12573         (vect_force_simple_reduction): Likewise.
12574         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
12575         LOOP_VINFO_REDUCTIONS.
12576         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12578 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12580         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
12581         a gimple stmt to a stmt_vec_info.
12582         * tree-vect-loop.c (vect_active_double_reduction_p)
12583         (vect_force_simple_reduction, vectorizable_reduction): Update
12584         accordingly.
12586 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12588         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
12589         vec<gimple *> to a vec<stmt_vec_info>.
12590         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
12591         the reduction_phis argument from a vec<gimple *> to a
12592         vec<stmt_vec_info>.
12593         (vectorizable_reduction): Likewise the phis local variable that
12594         is passed to vect_create_epilog_for_reduction.  Update for new type
12595         of SLP_TREE_VEC_STMTS.
12596         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
12597         (vectorizable_live_operation): Likewise.
12598         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
12599         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
12601 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12603         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
12604         a gimple stmt to a stmt_vec_info.
12605         (vectorizable_condition, vectorizable_live_operation)
12606         (vectorizable_reduction, vectorizable_induction): Pass back the
12607         vectorized statement as a stmt_vec_info.
12608         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
12609         use of STMT_VINFO_VEC_STMT.
12610         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
12611         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
12612         as stmt_vec_infos rather than gimple stmts.
12613         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
12614         to a stmt_vec_info.
12615         (vectorizable_live_operation): Likewise.
12616         (vectorizable_reduction, vectorizable_induction): Likewise,
12617         updating use of STMT_VINFO_VEC_STMT.
12618         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
12619         of STMT_VINFO_VEC_STMT.
12620         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
12621         (vectorizable_simd_clone_call, vectorizable_conversion)
12622         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
12623         (vectorizable_store, vectorizable_load, vectorizable_condition)
12624         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
12625         from a gimple stmt to a stmt_vec_info.
12626         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
12627         pointer to a stmt_vec_info to the vectorizable_* routines.
12629 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12631         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
12632         a gimple stmt to a stmt_vec_info.
12633         (is_pattern_stmt_p): Update accordingly.
12634         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
12635         (vect_record_grouped_load_vectors): Likewise.
12636         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
12637         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
12638         (vect_model_reduction_cost): Likewise.
12639         (vect_create_epilog_for_reduction): Likewise.
12640         (vectorizable_reduction, vectorizable_induction): Likewise.
12641         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
12642         Return the stmt_vec_info for the pattern statement.
12643         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
12644         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
12645         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
12646         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
12647         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
12648         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
12649         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
12650         (free_stmt_vec_info, vect_is_simple_use): Likewise.
12652 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12654         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
12655         (vect_finish_stmt_generation): Likewise.
12656         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
12657         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
12658         (vect_build_gather_load_calls): Use the return value of the above
12659         functions instead of a separate call to vinfo_for_stmt.  Use narrow
12660         scopes for the input gimple stmt and wider scopes for the associated
12661         stmt_vec_info.  Use vec_info::lookup_def when setting these
12662         stmt_vec_infos from an SSA_NAME definition.
12663         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
12664         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
12665         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
12666         (vectorizable_store, vectorizable_load, vectorizable_condition)
12667         (vectorizable_comparison): Likewise.
12668         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
12669         (vectorizable_reduction): Likewise.
12671 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12673         * tree-vectorizer.h (vect_is_simple_use): Add an optional
12674         stmt_vec_info * parameter before the optional gimple **.
12675         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
12676         (process_use, vect_get_vec_def_for_operand_1): Update callers.
12677         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
12678         * tree-vect-loop.c (vectorizable_reduction): Likewise.
12679         (vectorizable_live_operation): Likewise.
12680         * tree-vect-patterns.c (type_conversion_p): Likewise.
12681         (vect_look_through_possible_promotion): Likewise.
12682         (vect_recog_rotate_pattern): Likewise.
12683         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
12685 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12687         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
12688         a typedef to a wrapper class.
12689         (NULL_STMT_VEC_INFO): New macro.
12690         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
12691         (stmt_vec_info::operator*): New function.
12692         (stmt_vec_info::operator gimple *): Likewise.
12693         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
12694         (add_stmt_costs): Likewise.
12695         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
12696         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
12697         (vect_get_known_peeling_cost): Likewise.
12698         (vect_estimate_min_profitable_iters): Likewise.
12699         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
12700         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
12701         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
12702         (vectorizable_store, free_stmt_vec_infos): Likewise.
12703         (new_stmt_vec_info): Change return type of xcalloc to
12704         _stmt_vec_info *.
12706 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12708         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
12709         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
12710         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
12711         a single_imm_use-based sequence.
12712         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
12714 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12716         * tree-vectorizer.h (vec_info::lookup_def): Declare.
12717         * tree-vectorizer.c (vec_info::lookup_def): New function.
12718         * tree-vect-patterns.c (vect_get_internal_def): Use it.
12719         (vect_widened_op_tree): Likewise.
12720         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
12721         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
12722         (vectorizable_reduction): Likewise.
12723         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
12724         of a gimple *.
12725         (vect_is_slp_reduction): Update calls accordingly.  Use
12726         vec_info::lookup_def.
12727         (vect_is_simple_reduction): Likewise
12728         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
12730 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12732         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
12733         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
12734         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
12735         of vinfo_for_stmt.
12736         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
12737         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
12738         (vect_update_vf_for_slp, vect_analyze_loop_operations)
12739         (vect_is_slp_reduction, vectorizable_induction)
12740         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
12741         * tree-vect-patterns.c (vect_init_pattern_stmt):
12742         (vect_determine_min_output_precision_1, vect_determine_precisions)
12743         (vect_pattern_recog): Likewise.
12744         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
12745         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
12746         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
12747         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
12748         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
12749         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
12750         info field from a loop to a loop_vec_info.
12752 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12754         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
12755         (vec_info::add_stmt): Declare.
12756         * tree-vectorizer.c (vec_info::add_stmt): New function.
12757         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
12758         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
12759         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
12760         (vectorizable_induction): Likewise.
12761         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
12762         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
12763         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
12764         (vectorizable_load): Likewise.
12765         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
12766         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
12767         (vect_recog_gather_scatter_pattern): Likewise.
12768         (append_pattern_def_seq): Likewise.  Remove a check that is
12769         performed by add_stmt itself.
12771 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12773         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
12774         which make_ssa_name was called with new_stmt before new_stmt
12775         had been created.
12777 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12779         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
12780         split out from...
12781         (vect_is_slp_reduction): ...here...
12782         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
12783         that are already known to be false.
12785 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12787         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
12788         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
12789         STMT_VINFO_NUM_SLP_USES when it's true.
12790         (vect_free_slp_instance): Add a final_p parameter and pass it to
12791         vect_free_slp_tree.
12792         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
12793         (vect_analyze_slp_instance): Likewise.
12794         (vect_slp_analyze_operations): Likewise.
12795         (vect_slp_analyze_bb_1): Likewise.
12796         * tree-vectorizer.c (vec_info): Likewise.
12797         * tree-vect-loop.c (vect_transform_loop): Likewise.
12799 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12801         * tree-vect-loop.c (vectorizable_reduction): Assert that the
12802         function is not called for second and subsequent members of
12803         a reduction group.
12805 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
12807         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
12808         cases for nested loops from here to ...
12809         (vect_create_epilog_for_reduction): ...here.  Only call
12810         vect_is_simple_use for inner-loop reductions.
12812 2018-07-31  Martin Liska  <mliska@suse.cz>
12814         PR gcov-profile/85338
12815         PR gcov-profile/85350
12816         PR gcov-profile/85372
12817         * profile.c (struct location_triplet): New.
12818         (struct location_triplet_hash): Likewise.
12819         (output_location): Do not output a BB that
12820         is already recorded for a line.
12821         (branch_prob): Use streamed_locations.
12823 2018-07-31  Martin Liska  <mliska@suse.cz>
12825         PR gcov-profile/85370
12826         * coverage.c (coverage_begin_function): Do not mark target
12827         clones as artificial functions.
12829 2018-07-31  Martin Liska  <mliska@suse.cz>
12831         PR gcov-profile/83813
12832         PR gcov-profile/84758
12833         PR gcov-profile/85217
12834         PR gcov-profile/85332
12835         * profile.c (branch_prob): Do not record GOTO expressions
12836         for GIMPLE statements which locations are already streamed.
12838 2018-07-31  Olivier Hainque  <hainque@adacore.com>
12840         * gcc.c (handle_spec_function): Accept a soft_matched_part
12841         argument, as do_spec_1.  Pass it down to ...
12842         (eval_spec_function): Accept a soft_matched_part argument,
12843         and pass it down to ...
12844         (do_spec_2): Accept a soft_matched_part argument, and pass
12845         it down to do_spec_1.
12846         (do_spec_1): Pass soft_matched_part to handle_spec_function.
12847         (handle_braces): Update call to handle_spec_function.
12848         (driver::set_up_specs): Update calls to do_spec_2.
12849         (compare_debug_dump_opt_spec_function): Likewise.
12850         (compare_debug_self_opt_spec_function): Likewise.
12852 2018-07-31  Olivier Hainque  <hainque@adacore.com>
12854         * common.opt (nolibc): New option.
12855         * doc/invoke.texi (Link Options): Document it.
12856         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
12857         * config/alpha/linux.h: Likewise.
12858         * config/arc/elf.h: Likewise.
12859         * config/arm/uclinux-elf.h: Likewise.
12860         * config/arm/unknown-elf.h: Likewise.
12861         * config/avr/avrlibc.h: Likewise.
12862         * config/bfin/bfin.h: Likewise.
12863         * config/bfin/linux.h: Likewise.
12864         * config/bfin/uclinux.h: Likewise.
12865         * config/darwin.h: Likewise.
12866         * config/darwin10.h: Likewise.
12867         * config/darwin12.h: Likewise.
12868         * config/gnu-user.h: Likewise.
12869         * config/lm32/uclinux-elf.h: Likewise.
12870         * config/pa/pa-hpux11.h: Likewise.
12871         * config/pa/pa64-hpux.h: Likewise.
12872         * config/sparc/sparc.h: Likewise.
12874 2018-07-31  Olivier Hainque  <hainque@adacore.com>
12876         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
12877         undefined variables.
12879 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
12881         PR target/86640
12882         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
12883         instead of GEN_INT.
12885 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12887         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
12888         terminated string literal.
12890 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
12892         PR rtl-optimization/85160
12893         * combine.c (is_just_move): New function.
12894         (try_combine): Allow combining two instructions into two if neither of
12895         the original instructions was a move.
12897 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
12899         PR target/86673
12900         * doc/extend.texi (Global Register Variables): Discourage use of type
12901         qualifiers.
12902         (Local Register Variables): Likewise.
12904 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
12906         PR tree-optimization/86506
12907         * hwint.h (ceil_log2): Resync with hwint.c implementation.
12909 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
12911         PR target/86547
12912         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
12913         hard_regno, make sure no insn between `from` and `to` clobbers it.
12915 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
12916             Tom de Vries  <tdevries@suse.de>
12918         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
12919         (PTX_DEFAULT_RUNTIME_DIM): ... this.
12920         (nvptx_goacc_validate_dims): Set default worker and gang dims to
12921         PTX_DEFAULT_RUNTIME_DIM.
12922         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
12924 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
12926         * config/pa/pa.c (pa_output_addr_vec): Align address table.
12927         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
12928         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
12930 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
12932         * config/rs6000/constraints.md (wG constraint): Delete, no longer
12933         used.
12934         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
12935         predicate to reflect toc fusion has been deleted.
12936         (toc_fusion_mem_raw): Delete, no longer used.
12937         (toc_fusion_mem_wrapped): Likewise.
12938         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
12939         fusion mask bit.
12940         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
12941         Delete, no longer used.
12942         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
12943         meant to be used for toc fusion.
12944         (rs6000_debug_print_mode): Delete toc fusion debugging.
12945         (rs6000_debug_reg_global): Likewise.
12946         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
12947         fusion and secondary reload support that were never used.
12948         (rs6000_option_override_internal): Delete TOC fusion, that was only
12949         partially defined, and it did not work unless you also used the
12950         -mcmodel= switch.
12951         (rs6000_legitimate_address_p): Delete TOC fusion support.
12952         (rs6000_opt_masks): Likewise.
12953         (fusion_wrap_memory_address): Delete function, no longer used.
12954         (fusion_split_address); Delete TOC fusion support.
12955         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
12956         longer used with toc fusion being deleted.
12957         (TARGET_TOC_FUSION_FP): Likewise.
12958         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
12959         UNSPEC.
12960         (toc fusion spliter): Delete TOC fusion support.
12961         (toc_fusionload_<mode>): Likewise.
12962         (toc_fusionload_di): Likewise.
12963         (fusion_gpr_load_<mode>): Delete generator function, this insn no
12964         longer needs to be named.  Rename predicate to delete TOC fusion.
12965         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
12966         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
12967         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
12968         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
12969         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
12971 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12973         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
12974         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
12975         __int128 in built-in function prototypes.
12976         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
12977         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
12979 2018-07-27  Martin Sebor  <msebor@redhat.com>
12981         PR tree-optimization/86696
12982         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
12983         types, including enums.
12984         (handle_char_store): Be prepared for the above function to fail.
12986 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
12988         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
12989         when optimization level is lower than 2 or optimize for size.
12991 2018-07-26  Martin Sebor  <msebor@redhat.com>
12993         PR tree-optimization/86043
12994         PR tree-optimization/86042
12995         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
12996         (get_string_cst_length): Rename...
12997         (get_min_string_length): ...to this.  Add argument.
12998         (handle_char_store): Extend to handle multi-character stores by
12999         MEM_REF.
13000         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
13001         * tree.h (initializer_zerop): Add argument.
13003 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
13005         PR middle-end/86660
13006         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
13007         declare target to variables if they have always,{to,from,tofrom} map
13008         kinds.
13010 2018-07-26  Martin Liska  <mliska@suse.cz>
13012         PR lto/86548
13013         * lto-wrapper.c: Add linker_output as prefix
13014         for ltrans_output_file.
13016 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
13018         PR rtl-optimization/85805
13019         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
13020         value for hard registers if that was written in the same mode.
13022 2018-07-26  Martin Liska  <mliska@suse.cz>
13024         PR gcov-profile/86536
13025         * gcov.c (format_gcov): Use printf format %.*f directly
13026         and do not handle special values.
13028 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
13030         * common/config/arc/arc-common.c (arc_option_optimization_table):
13031         Update default optimizations for size.
13033 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
13035         * config/arc/arc.md (movsf_insn): Add short instruction selection.
13036         * config/arc/constraints.md (CfZ): New constraint.
13037         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
13038         (subsf3_fpu): Likewise.
13039         (cmpsf_fpu): Likewise.
13040         (cmpsf_fpu_uneq): Likewise.
13042 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
13044         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
13045         accesses as well.
13046         (arc_is_uncached_mem_p): uncached applies to both the variable and
13047         the pointer.
13049 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
13051         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
13052         register names.
13054 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
13056         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
13057         field "m_scopes" from vec to auto_vec.
13059 2018-07-25  Martin Liska  <mliska@suse.cz>
13061         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
13062         return type.
13064 2018-07-25  Richard Biener  <rguenther@suse.de>
13066         PR debug/86654
13067         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
13068         special wrt context_die late.
13069         (gen_subprogram_die): Re-use DIEs in local scope.
13071 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
13073         PR tree-optimization/86644
13074         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
13076 2018-07-25  Martin Liska  <mliska@suse.cz>
13078         PR middle-end/86645
13079         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
13080         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
13082 2018-07-25  Martin Liska  <mliska@suse.cz>
13084         PR sanitizer/79635
13085         * params.def: Explain ASan abbreviation and provide
13086         a documentation link.
13088 2018-07-24  Martin Sebor  <msebor@redhat.com>
13090         PR tree-optimization/86622
13091         PR tree-optimization/86532
13092         * builtins.h (string_length): Declare.
13093         * builtins.c (c_strlen): Correct handling of non-constant offsets.
13094         (check_access): Be prepared for non-constant length ranges.
13095         (string_length): Make extern.
13096         * expr.c (string_constant): Only handle the minor non-constant
13097         array index.  Use string_constant to compute the length of
13098         a generic string constant.
13100 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
13102         PR tree-optimization/86618
13103         * tree-vect-stmts.c (vectorizable_call): Don't take the address
13104         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
13106 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
13108         PR tree-optimization/86636
13109         * json.cc (json::object::set): Fix comment.  Add assertions.
13110         (json::array::append): Move here from json.h.  Add comment and an
13111         assertion.
13112         (json::string::string): Likewise.
13113         * json.h (json::array::append): Move to json.cc.
13114         (json::string::string): Likewise.
13115         * optinfo-emit-json.cc
13116         (optrecord_json_writer::impl_location_to_json): Assert that we
13117         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
13118         wrapper around it.  Expand the location once, rather than three
13119         times.
13120         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
13121         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
13122         wrappers.
13123         (optrecord_json_writer::optinfo_to_json): Likewise, in four
13124         places.  Fix some overlong lines.
13126 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
13128         * config/aarch64/aarch64-simd.md
13129         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
13130         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
13131         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
13132         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
13133         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
13134         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
13135         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
13136         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
13137         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
13139 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
13141         PR middle-end/86627
13142         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
13143         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
13144         and abs_d == d, do the power of two handling if profitable.
13146 2018-07-24  Richard Biener  <rguenther@suse.de>
13148         * match.pd: Add BIT_FIELD_REF canonicalizations.
13150 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13152         PR c/86617
13153         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
13155 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13157         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
13158         terminated STRING_CST object.
13160 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13162         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
13164 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
13166         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
13167         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
13168         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
13169         the elements into a register.
13170         (rs6000_split_v4si_init_di_reg): Delete.
13171         (rs6000_split_v4si_init): Delete.
13172         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
13173         (vsx_init_v4si): Rewrite as a define_expand.
13175 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
13177         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
13178         zero_extend argument from memory): New.
13180 2018-07-22  Martin Sebor  <msebor@redhat.com>
13182         PR bootstrap/86621
13183         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
13184         diagnosing calls with unknown arguments unless -Walloca-larger-than
13185         is restricted to less than PTRDIFF_MAX bytes.
13187 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
13189         * doc/gcov.texi (Invoking Gcov): Editorial changes.
13191 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
13193         * pretty-print.c (text_info::set_location): Remove redundant
13194         "line_table" parameter from call to rich_location::set_range.
13196 2018-07-20  Martin Sebor  <msebor@redhat.com>
13198         PR middle-end/82063
13199         * builtins.c (expand_builtin_alloca): Adjust.
13200         * calls.c (alloc_max_size): Simplify.
13201         * cgraphunit.c (cgraph_node::expand): Adjust.
13202         * common.opt (larger_than_size, warn_frame_larger_than): Remove
13203         variables.
13204         (frame_larger_than_size): Same.
13205         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
13206         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
13207         Initialize.
13208         * doc/invoke.texi (GCC Command Options): Document option arguments.
13209         Explain byte-size arguments and suffixes.
13210         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
13211         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
13212         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
13213         * doc/options.texi (UInteger): Expand.
13214         (Host_Wide_Int, ByteSize): Document new properties.
13215         * final.c (final_start_function_1): Include sizes in an error message.
13216         * function.c (frame_offset_overflow): Same.
13217         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
13218         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
13219         Diagnose unbounded alloca calls only for limits of less than
13220         PTRDIFF_MAX.
13221         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
13222         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
13223         for alloca(0).
13224         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
13225         only for limits of less than PTRDIFF_MAX.
13226         * langhooks-def.h (lhd_handle_option): Change function argument
13227         to HOST_WIDE_INT.
13228         * langhooks.c (lhd_handle_option): Same.
13229         * langhooks.h (handle_option): Same.
13230         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
13231         ByteSize flags.
13232         (var_type, var_type_struct): Same.
13233         (var_set): Handle ByteSize flag.
13234         * optc-gen.awk: Add comments to output to ease debugging.  Make
13235         use of HOST_WIDE_INT where appropriate.
13236         * opts-gen-save.awk:  Use %lx to format unsigned long.
13237         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
13238         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
13239         arguments.  Parse bytes-size suffixes.
13240         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
13241         (enum_value_to_arg): Same.
13242         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
13243         (handle_option): Adjust.
13244         (generate_option): Change function argument to HOST_WIDE_INT.
13245         (cmdline_handle_error): Adjust.
13246         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
13247         (set_option): Change function argument to HOST_WIDE_INT.
13248         (option_enabled): Handle cl_host_wide_int.
13249         (get_option_state): Handle CLVC_SIZE.
13250         (control_warning_option): Same.
13251         * opts.c (common_handle_option): Change function argument to
13252         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
13253         OPT_Wvla_larger_than_.
13254         * opts.h (enum cl_var_type): Add an enumerator.
13255         * stor-layout.c (layout_decl): Print a more meaningful warning.
13256         * toplev.c (output_stack_usage): Adjust.
13258 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
13260         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
13261         call to inline_expand_builtin_string_cmp.
13262         (expand_builtin_strcmp): Likewise.
13263         (expand_builtin_strncmp): Likewise.
13264         (inline_string_cmp): Delete the last parameter, change char_type_node
13265         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
13266         two operands.
13267         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
13268         the inlining expansion on target where the type of the call has same or
13269         narrower precision than unsigned char.
13271 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
13273         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
13274         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
13275         * common.opt (fsave-optimization-record): New option.
13276         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
13277         * doc/invoke.texi (-fsave-optimization-record): New option.
13278         * dumpfile.c: Include "optinfo-emit-json.h".
13279         (struct kv_pair): Move to coretypes.h.
13280         (optgroup_options): Make non-static.
13281         (dump_context::end_scope): Call
13282         optimization_records_maybe_pop_dump_scope.
13283         * dumpfile.h (optgroup_options): New decl.
13284         * json.cc: New file.
13285         * json.h: New file.
13286         * optinfo-emit-json.cc: New file.
13287         * optinfo-emit-json.h: New file.
13288         * optinfo.cc: Include "optinfo-emit-json.h".
13289         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
13290         (optinfo_enabled_p): Check optimization_records_enabled_p.
13291         (optinfo_wants_inlining_info_p): Likewise.
13292         * optinfo.h: Update comment.
13293         * profile-count.c (profile_quality_as_string): New function.
13294         * profile-count.h (profile_quality_as_string): New decl.
13295         (profile_count::quality): New accessor.
13296         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
13297         and optinfo_emit_json_cc_tests.
13298         * selftest.h (selftest::json_cc_tests): New decl.
13299         (selftest::optinfo_emit_json_cc_tests): New decl.
13300         * toplev.c: Include "optinfo-emit-json.h".
13301         (compile_file): Call optimization_records_finish.
13302         (do_compile): Call optimization_records_start.
13303         * tree-ssa-live.c: Include optinfo.h.
13304         (remove_unused_scope_block_p): Retain inlining information if
13305         optinfo_wants_inlining_info_p returns true.
13307 2018-07-20  Richard Biener  <rguenther@suse.de>
13309         PR debug/86585
13310         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
13311         to cover -flto-partition=none.
13313 2018-07-20  Martin Liska  <mliska@suse.cz>
13315         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
13316         (get_decl_source_range): Remove unused function.
13318 2018-07-20  Richard Biener  <rguenther@suse.de>
13320         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
13321         (struct vn_phi_s): Likewise.
13322         (struct vn_reference_s): Likewise.
13323         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
13324         for searching the slot of an entry known to be in the hash itself.
13325         (vn_phi_hasher::equal): Likewise.
13326         (vn_reference_hasher::equal): Likewise.
13327         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
13328         globals.
13329         (optimistic_info, current_info): Remove, keeping only valid_info.
13330         (vn_reference_lookup_1): Remove fallback lookup.
13331         (vn_reference_lookup_2): Likewise.
13332         (vn_nary_op_lookup_1): Likewise.
13333         (vn_phi_lookup): Likewise.
13334         (vn_nary_build_or_lookup_1): Make sure to not chain the built
13335         hash element.
13336         (vn_reference_insert): Adjust, chain the inserted hash element
13337         at last_inserted_ref.
13338         (vn_reference_insert_pieces): Likewise.
13339         (visit_reference_op_call): Likewise.
13340         (vn_nary_op_insert_into): Chain the inserted hash element at
13341         last_inserted_nary.
13342         (vn_nary_op_insert_pieces): Adjust.
13343         (vn_nary_op_insert): Likewise.
13344         (vn_nary_op_insert_stmt): Likewise.
13345         (vn_phi_insert): Adjust, chain the inserted hash element at
13346         last_inserted_phi.
13347         (process_scc): Remove clearing and copying the optimistic
13348         table.  Instead remove elements inserted during an optimistic
13349         iteration from the single table we maintain.
13350         (init_scc_vn): Adjust.
13351         (free_scc_vn): Likewise.
13352         (sccvn_dom_walker::record_cond): Likewise.
13353         (sccvn_dom_walker::after_dom_children): Likewise.
13355 2018-07-19  Martin Sebor  <msebor@redhat.com>
13357         PR tree-optimization/84047
13358         PR tree-optimization/83776
13359         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
13360         (check_array_bounds): Call it.
13362 2018-07-19  Martin Sebor  <msebor@redhat.com>
13364         * align.h (align_flags): Use member initialization.
13366 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
13368         * Makefile.in (OBJS): Add optinfo.o.
13369         * coretypes.h (class symtab_node): New forward decl.
13370         (struct cgraph_node): New forward decl.
13371         (class varpool_node): New forward decl.
13372         * dump-context.h: New file.
13373         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
13374         "tree-pass.h".
13375         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
13376         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
13377         (set_alt_dump_file): Likewise.
13378         (dump_context::~dump_context): New dtor.
13379         (dump_gimple_stmt): Move implementation to...
13380         (dump_context::dump_gimple_stmt): ...this new member function.
13381         Add the stmt to any pending optinfo, creating one if need be.
13382         (dump_gimple_stmt_loc): Move implementation to...
13383         (dump_context::dump_gimple_stmt_loc): ...this new member function.
13384         Start a new optinfo and add the stmt to it.
13385         (dump_gimple_expr): Move implementation to...
13386         (dump_context::dump_gimple_expr): ...this new member function.
13387         Add the stmt to any pending optinfo, creating one if need be.
13388         (dump_gimple_expr_loc): Move implementation to...
13389         (dump_context::dump_gimple_expr_loc): ...this new member function.
13390         Start a new optinfo and add the stmt to it.
13391         (dump_generic_expr): Move implementation to...
13392         (dump_context::dump_generic_expr): ...this new member function.
13393         Add the tree to any pending optinfo, creating one if need be.
13394         (dump_generic_expr_loc): Move implementation to...
13395         (dump_context::dump_generic_expr_loc): ...this new member
13396         function.  Add the tree to any pending optinfo, creating one if
13397         need be.
13398         (dump_printf): Move implementation to...
13399         (dump_context::dump_printf_va): ...this new member function.  Add
13400         the text to any pending optinfo, creating one if need be.
13401         (dump_printf_loc): Move implementation to...
13402         (dump_context::dump_printf_loc_va): ...this new member function.
13403         Start a new optinfo and add the stmt to it.
13404         (dump_dec): Move implementation to...
13405         (dump_context::dump_dec): ...this new member function.  Add the
13406         value to any pending optinfo, creating one if need be.
13407         (dump_context::dump_symtab_node): New member function.
13408         (dump_context::get_scope_depth): New member function.
13409         (dump_context::begin_scope): New member function.
13410         (dump_context::end_scope): New member function.
13411         (dump_context::ensure_pending_optinfo): New member function.
13412         (dump_context::begin_next_optinfo): New member function.
13413         (dump_context::end_any_optinfo): New member function.
13414         (dump_context::s_current): New global.
13415         (dump_context::s_default): New global.
13416         (dump_scope_depth): Delete global.
13417         (dumpfile_ensure_any_optinfo_are_flushed): New function.
13418         (dump_symtab_node): New function.
13419         (get_dump_scope_depth): Reimplement in terms of dump_context.
13420         (dump_begin_scope): Likewise.
13421         (dump_end_scope): Likewise.
13422         (selftest::temp_dump_context::temp_dump_context): New ctor.
13423         (selftest::temp_dump_context::~temp_dump_context): New dtor.
13424         (selftest::verify_item): New function.
13425         (ASSERT_IS_TEXT): New macro.
13426         (ASSERT_IS_TREE): New macro.
13427         (ASSERT_IS_GIMPLE): New macro.
13428         (selftest::test_capture_of_dump_calls): New test.
13429         (selftest::dumpfile_c_tests): Call it.
13430         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
13431         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
13432         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
13433         descriptive comment.
13434         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
13435         (dump_node, dump_bb): Move these unrelated decls.
13436         (class dump_manager): Add leading comment.
13437         * optinfo.cc: New file.
13438         * optinfo.h: New file.
13440 2018-07-19  Michael Collison  <michael.collison@arm.com>
13441             Richard Henderson <rth@redhat.com>
13443         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
13444         (subti): Handle op1 zero.
13445         (subvti4, usub4ti4): New.
13446         (*sub<GPI>3_compare1_imm): New.
13447         (sub<GPI>3_carryinCV): New.
13448         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
13449         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
13451 2018-07-19  Michael Collison  <michael.collison@arm.com>
13452             Richard Henderson <rth@redhat.com>
13454         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
13455         (addti3): Create simpler code if low part is already known to be 0.
13456         (addvti4, uaddvti4): New.
13457         (*add<GPI>3_compareC_cconly_imm): New.
13458         (*add<GPI>3_compareC_cconly): New.
13459         (*add<GPI>3_compareC_imm): New.
13460         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
13461         handle constants within this pattern..
13462         (*add<GPI>3_compareV_cconly_imm): New.
13463         (*add<GPI>3_compareV_cconly): New.
13464         (*add<GPI>3_compareV_imm): New.
13465         (add<GPI>3_compareV): New.
13466         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
13467         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
13468         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
13469         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
13470         with 'comparison' operator.
13471         (*add<GPI>3_compareV_cconly_imm): Ditto.
13472         (*add<GPI>3_compareV_cconly): Ditto.
13473         (*add<GPI>3_compareV_imm): Ditto.
13474         (add<GPI>3_compareV): Ditto.
13475         (add<mode>3_carryinC): Ditto.
13476         (*add<mode>3_carryinC_zero): Ditto.
13477         (*add<mode>3_carryinC): Ditto.
13478         (add<mode>3_carryinV): Ditto.
13479         (*add<mode>3_carryinV_zero): Ditto.
13480         (*add<mode>3_carryinV): Ditto.
13482 2018-07-19  Michael Collison  <michael.collison@arm.com>
13483             Richard Henderson <rth@redhat.com>
13485         * config/aarch64/aarch64-modes.def (CC_V): New.
13486         * config/aarch64/aarch64-protos.h
13487         (aarch64_addti_scratch_regs): Declare
13488         (aarch64_subvti_scratch_regs): Declare.
13489         (aarch64_expand_subvti): Declare.
13490         (aarch64_gen_unlikely_cbranch): Declare
13491         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
13492         for signed overflow using CC_Vmode.
13493         (aarch64_get_condition_code_1): Handle CC_Vmode.
13494         (aarch64_gen_unlikely_cbranch): New function.
13495         (aarch64_addti_scratch_regs): New function.
13496         (aarch64_subvti_scratch_regs): New function.
13497         (aarch64_expand_subvti): New function.
13499 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13501         * config/aarch64/aarch64-option-extensions.def: New entry for profile
13502         extension.
13503         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
13504         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
13505         extension.
13507 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13509         PR target/83009
13510         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13511         address check not strict.
13513 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13515         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
13516         Umq with Umn.
13517         (store_pair_lanes<mode>): Likewise.
13518         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
13519         enum value 'ADDR_QUERY_LDP_STP_N'.
13520         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
13521         (aarch64_print_address_internal): Add declaration.
13522         (aarch64_print_ldpstp_address): Remove.
13523         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
13524         (aarch64_print_operand): Change printing of 'y'.
13525         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
13526         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
13527         'true' rather than '1'.
13528         * config/aarch64/constraints.md (Uml): Likewise.
13529         (Uml): Rename to Umn.
13530         (Umq): Remove.
13532 2018-07-19  Richard Biener  <rguenther@suse.de>
13534         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
13535         a trailing array.
13536         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
13537         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
13538         (vn_reference_hasher): Likewise.
13539         (struct vn_tables_s): Remove obstack and alloc-pool members.
13540         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
13541         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
13542         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
13543         (vn_reference_insert_pieces): Likewise.
13544         (alloc_vn_nary_op_noinit): Adjust.
13545         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
13546         (vn_phi_eq): Adjust.
13547         (shared_lookup_phiargs): Remove.
13548         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
13549         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
13550         (visit_reference_op_call): Likewise.
13551         (copy_nary, copy_phi, copy_reference): Remove.
13552         (process_scc): Rewind the obstack when iterating.  Do not
13553         copy the elements to valid_info but just move them from one
13554         hashtable to the other.
13555         (allocate_vn_table): Adjust.
13556         (free_vn_table): Likewise.
13557         (init_scc_vn): Likewise.
13558         (free_scc_vn): Likewise.
13560 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
13562         PR target/86560
13563         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
13564         indirect_return as function type attribute.
13565         (ix86_attribute_table): Change indirect_return to function
13566         type attribute.
13567         * doc/extend.texi: Update indirect_return attribute.
13569 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
13571         * wide-int.h (widest2_int): New.
13572         * gimple-fold.c (arith_overflowed_p): Use it.
13573         * tree.h (widest2_int_cst): New.
13574         * tree-vrp.c (wide_int_binop_overflow): Rename from
13575         vrp_int_const_binop.
13576         Rewrite to work on trees.
13577         (extract_range_from_multiplicative_op_1): Abstract code to...
13578         (wide_int_range_min_max): ...here.
13579         (wide_int_range_cross_product): ...and here.
13580         (extract_range_from_binary_expr_1): Abstract overflow code to...
13581         (wide_int_range_mult_wrapping): ...here.
13582         * tree-vrp.h (wide_int_range_cross_product): New.
13583         (wide_int_range_mult_wrapping): New.
13585 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
13586             Julia Koval  <julia.koval@intel.com>
13588         * config/i386/x86-tune-costs.h (skylake_memcpy,
13589         skylake_memset): Replace rep_prefix with unrolling for size 512.
13591 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
13593         PR middle-end/86544
13594         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
13595         comparision with EQ_EXPR in last stmt.
13597 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13599         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
13600         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
13601         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
13602         previously known as "PowerPC AltiVec Built-in Functions".  Move
13603         some material to new subsubsections "PowerPC AltiVec Built-in
13604         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
13605         ISA 2.07".
13606         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
13607         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
13608         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
13609         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
13611 2018-07-18  Richard Biener  <rguenther@suse.de>
13613         PR tree-optimization/86557
13614         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
13615         EXACT_DIV_EXPR.
13617 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
13619         * config/s390/s390.c (s390_function_profiler): Generate CFI.
13621 2018-07-17  Jeff Law  <law@redhat.com>
13623         * config/arm/arm.c (get_label_padding): Update for recent
13624         changes to label_to_alignment.
13626         PR tree-optimization/86010
13627         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
13629         * config/mips/mips.c (vr4130_align_insns): Update for recent
13630         changes to label_to_alignment.
13632         * config/frv/frv.c (frv_label_align): Update for recent changes
13633         to label_to_alignment.
13635         * config/nios2/nios2.c (nios2_label_align): Update for recent
13636         changes which dropped ALIGN_LABELS_LOG.
13638 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
13640         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
13641         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
13643 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
13645         * config/arc/arc.c (arc_label_align): Use align_labels instead of
13646         deprecated align_labels_log.
13648 2018-07-17  Richard Biener  <rguenther@suse.de>
13650         PR lto/86456
13651         * dwarf2out.c (init_sections_and_labels): Always generate
13652         a debug_line_str_section for early LTO debug.
13653         (dwarf2out_finish): Reset debug_line_str_hash output early.
13654         Bump counter for extra dwarf5 .debug_loc labels to not conflict
13655         with fat LTO part.
13656         (dwarf2out_early_finish): Output debug_line_str.
13658 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
13660         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
13661         index register on z196 or later.
13663 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
13665         * config/s390/s390.c (s390_default_align): Set default function
13666         alignment to 16.
13667         (s390_override_options_after_change): Call s390_default align.
13668         (s390_option_override_internal): Call s390_default_align.
13669         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
13671 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
13673         PR middle-end/86542
13674         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
13675         to _looptemp_ clauses, other than the first two.
13677 2018-07-17  Martin Liska  <mliska@suse.cz>
13679         * opts.c: Do not enable OPT_falign_* for -Os.
13681 2018-07-17  Martin Liska  <mliska@suse.cz>
13683         * align.h (MAX_CODE_ALIGN): New.
13684         (MAX_CODE_ALIGN_VALUE): New.
13685         * common/config/i386/i386-common.c (ix86_handle_option):
13686         (MAX_CODE_ALIGN): Moved to align.h.
13687         * final.c (MAX_CODE_ALIGN): Likewise.
13688         * opts.c (parse_and_check_align_values):
13689         (MAX_CODE_ALIGN): Likewise.
13690         (MAX_CODE_ALIGN_VALUE): Likewise.
13692 2018-07-17  Martin Liska  <mliska@suse.cz>
13694         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
13695         in order to fulfil coding style.
13696         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
13697         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13698         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13699         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
13700         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
13701         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
13702         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
13703         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
13704         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13706 2018-07-17  Martin Liska  <mliska@suse.cz>
13708         * align.h: New file.
13709         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
13710         directly.
13711         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
13712         align_flags of label_to_alignment.
13713         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
13714         align_flags class.
13715         * config/m68k/m68k.c: Do not use removed align_labels_value and
13716         align_loops_value.
13717         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
13718         (LOOP_ALIGN): Likewise.
13719         (LABEL_ALIGN): Likewise.
13720         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
13721         Remove not used macro.
13722         (rs6000_loop_align): Change return type to align_flags.
13723         (rs6000_loop_align_max_skip): Remove.
13724         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
13725         Change return type to align_flags.
13726         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
13727         Remove not used macro.
13728         (rs6000_loop_align):  Change return type to align_flags.
13729         (rs6000_loop_align_max_skip): Remove.
13730         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
13731         * config/rx/rx-protos.h (rx_align_for_label): Make it
13732         static function.
13733         * config/rx/rx.c (rx_align_for_label): Change return type
13734         to align_flags.
13735         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
13736         macro definitions.
13737         into align_flags class.
13738         (LABEL_ALIGN): Likewise.
13739         (LOOP_ALIGN): Likewise.
13740         * config/s390/s390.c (s390_label_align): Use align_flags
13741         class member.
13742         (s390_asm_output_function_label): Likewise.
13743         * config/sh/sh.c (sh_override_options_after_change):
13744         Use align_flags class directly without macros.
13745         (find_barrier): Likewise.
13746         (barrier_align): Likewise.
13747         (sh_loop_align): Likewise.
13748         * config/spu/spu.c (spu_option_override):
13749         Use align_flags_tuple::get_value instead of removed macros.
13750         (spu_sched_init): Likewise.
13751         * config/spu/spu.h (GTY): Likewise.
13752         * config/visium/visium.c (visium_option_override):
13753         Set "8" as default secondary alignment.
13754         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
13755         in order to guarantee secondary alignment of 8.
13756         * coretypes.h: Include align.h header file.
13757         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
13758         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
13759         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
13760         * doc/tm.texi.in: Likewise.
13761         * final.c (struct label_alignment): Remove not used structure.
13762         (LABEL_ALIGN): Change type to align_flags.
13763         (LOOP_ALIGN): Likewise.
13764         (JUMP_ALIGN): Likewise.
13765         (default_loop_align_max_skip): Remove.
13766         (default_label_align_max_skip): Likewise.
13767         (default_jump_align_max_skip): Likewise.
13768         (default_label_align_after_barrier_max_skip):
13769         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
13770         (LABEL_TO_MAX_SKIP): Remove.
13771         (label_to_alignment): Return align_flags type instead of integer.
13772         (label_to_max_skip): Remove.
13773         (align_fuzz): Use align_flags type.
13774         (compute_alignments): Use align_flags type and use align_flags::max
13775         to combine multiple alignments.
13776         (grow_label_align): Grow vec instead of C array.
13777         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
13778         (shorten_branches):  Use align_flags type and use align_flags::max
13779         to combine multiple alignments.
13780         (final_scan_insn_1): Remove usage of secondary alignment that comes
13781         from label alignment, but instead use proper secondary alignment
13782         which is computed in grow_label_align.
13783         * flags.h (struct align_flags_tuple): Move to align.h.
13784         (struct align_flags): Likewise.
13785         (state_align_loops): Rename to align_loops.
13786         (state_align_jumps): Rename to align_jumps.
13787         (state_align_labels): Rename to align_labels.
13788         (state_align_functions): Rename to align_functions.
13789         (align_loops_log): Remove.
13790         (align_jumps_log): Remove.
13791         (align_labels_log): Remove.
13792         (align_functions_log): Remove.
13793         (align_loops_max_skip): Remove.
13794         (align_jumps_max_skip): Remove.
13795         (align_labels_max_skip): Remove.
13796         (align_functions_max_skip): Remove.
13797         (align_loops_value): Remove.
13798         (align_jumps_value): Remove.
13799         (align_labels_value): Remove.
13800         (align_functions_value): Remove.
13801         * output.h (label_to_alignment): Change return type to align_flags.
13802         (label_to_max_skip): Remove.
13803         * target.def: Remove loop_align_max_skip, label_align_max_skip,
13804         jump_align_max_skip macros.
13805         * targhooks.h (default_loop_align_max_skip): Remove.
13806         (default_label_align_max_skip): Likewise.
13807         (default_jump_align_max_skip): Likewise.
13808         (default_label_align_after_barrier_max_skip): Remove.
13809         * toplev.c (read_log_maxskip): Use ::normalize function.
13810         (parse_N_M): Remove not used argument and also call ::normalize.
13811         (parse_alignment_opts): Do not pass unused arguments.
13812         * varasm.c (assemble_start_function): Use directly align_functions
13813         instead of removed macros.
13814         * system.h: Do not poison removed macros.
13816 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
13818         PR middle-end/86539
13819         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
13820         and cond temporaries don't have reference type if iterator has
13821         pointer type.  For init use &for_pre_body instead of pre_p if
13822         for_pre_body is non-empty.
13824 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
13826         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
13827         double-double modes to SFmode directly directly.
13828         (trunc<mode>sf2_fprs): Delete.
13830 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
13832         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
13833         for conversions between IFmode and the decimal floating point modes.
13834         (init_float128_ieee): Use the correct names for conversions between
13835         KFmode and the decimal floating point modes.
13837 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
13839         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
13840         for the conversions between TDmode and IFmode.
13841         (init_float128_ieee): Use more correct names for the conversions
13842         between TDmode and KFmode.
13844 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
13846         PR tree-optimization/86526
13847         * builtins.c (expand_builtin_memcmp): Formatting fixes.
13848         (inline_expand_builtin_string_cmp): Likewise.
13849         (inline_string_cmp): Likewise.  Use c_readstr instead of
13850         builtin_memcpy_read_str.  Add unit_mode temporary.
13852 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
13854         PR middle-end/86528
13855         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
13856         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
13858 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13860         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
13861         Alphabetize prototypes of built-in functions, separating out
13862         built-in functions that are listed in this section but should be
13863         described elsewhere.
13865 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
13867         PR target/86511
13868         * expmed.c (emit_store_flag): Do not emit setcc followed by a
13869         conditional move when trapping comparison was split to a
13870         non-trapping one (and vice versa).
13872 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
13874         * config/s390/s390.c (s390_function_profiler): Generate nops
13875         instead of profiler call sequences.
13876         * config/s390/s390.opt: Add the new option.
13878 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
13880         * config/s390/s390.c (s390_function_profiler): Generate
13881         __mcount_loc section.
13882         * config/s390/s390.opt: Add the new option.
13884 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
13886         * common.opt: Add the new warning.
13887         * config/s390/s390.c (s390_function_profiler): Emit "brasl
13888         %r0,__fentry__" when -mfentry is specified.
13889         (s390_option_override_internal): Disallow -mfentry for 31-bit
13890         CPUs.
13891         * config/s390/s390.opt: Add the new option.
13893 2018-07-16  Richard Biener  <rguenther@suse.de>
13895         PR lto/86523
13896         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
13897         for function-local FUNCTION_DECL and RESULT_DECL immediately.
13899 2018-07-16  Martin Liska  <mliska@suse.cz>
13901         PR ipa/86529
13902         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
13903         to ::get_create.
13905 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
13907         * config/arc/arcHS.md: Update ARCHS scheduling rules.
13909 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
13911         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
13912         for ARCHS4x.
13913         * config/arc/arc-cpus.def (hs4x): New cpu.
13914         (hs4xd): Likewise.
13915         * config/arc/arc-tables.opt: Regenerate.
13916         * config/arc/arc.c (arc_sched_issue_rate): New function.
13917         (TARGET_SCHED_ISSUE_RATE): Define.
13918         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
13919         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
13920         fpu_cvt.
13921         (attr tune): Add ARCHS4x tune values.
13922         (attr tune_dspmpy): Define.
13923         (*tst): Correct instruction type.
13924         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
13925         * config/arc/arcHS4x.md: New file.
13926         * config/arc/fpu.md: Update instruction type attributes.
13927         * config/arc/t-multilib: Regenerate.
13929 2018-07-16  Tom de Vries  <tdevries@suse.de>
13931         PR debug/86455
13932         * var-tracking.c (vt_initialize): Fix pre_dec handling.
13934 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
13936         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
13937         early clobber.
13939 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
13941         PR tree-optimization/86514
13942         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
13943         conversion to a boolean type from a type with greater precision.
13945 2018-07-16  Tom de Vries  <tdevries@suse.de>
13947         * var-tracking.c (vt_initialize): Print adjusted insn slim if
13948         dump_flags request TDF_SLIM.
13950 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
13952         * fold-const.c (int_const_binop_1): Abstract...
13953         (wide_int_binop): ...wide int code here.
13954         (poly_int_binop): ...poly int code here.
13955         Abstract the rest of int_const_binop_1 into int_const_binop.
13956         * fold-const.h (wide_int_binop): New.
13957         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
13958         Remove useless PLUS/MINUS_EXPR case.
13959         (zero_nonzero_bits_from_vr): Move wide int code...
13960         (zero_nonzero_bits_from_bounds): ...here.
13961         (extract_range_from_binary_expr_1): Move mask optimization code...
13962         (range_easy_mask_min_max): ...here.
13963         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
13964         (range_easy_mask_min_max): New.
13966 2018-07-15  Jeff Law  <law@redhat.com>
13968         PR target/85993
13969         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
13970         block.
13972 2018-07-14  Jim Wilson  <jimw@sifive.com>
13974         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
13976 2018-07-14  Paul Koning  <ni1d@arrl.net>
13978         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
13980 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
13982         * lto-streamer-out.c (copy_function_or_variable): Dump info about
13983         copying section.
13985 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
13986             Steve Munroe  <munroesj52@gmail.com>
13988         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
13989         (_mm_andnot_si128): Likewise.
13990         (_mm_or_si128): Likewise.
13991         (_mm_xor_si128): Likewise.
13993 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
13995         PR middle-end/78809
13996         * builtins.c (expand_builtin_memcmp): Inline the calls first
13997         when result_eq is false.
13998         (expand_builtin_strcmp): Inline the calls first.
13999         (expand_builtin_strncmp): Likewise.
14000         (inline_string_cmp): New routine. Expand a string compare
14001         call by using a sequence of char comparison.
14002         (inline_expand_builtin_string_cmp): New routine. Inline expansion
14003         a call to str(n)cmp/memcmp.
14004         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
14005         New option.
14006         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
14008 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
14010         * config/arm/driver-arm.c: Include arm-native.h.
14011         (host_detect_local_cpu): Use auto-generated data tables.
14012         (vendors, arm_cpu_table): Delete.  Move part information to ...
14013         * config/arm/arm-cpus.in: ... here.
14014         * config/arm/parsecpu.awk (gen_native): New function.
14015         (vendor, part): New CPU fields.
14016         (END): Add support for building the native CPU detection tables.
14017         * config/arm/t-arm (arm-native.h): Add build rule.
14018         (driver-arm.o): Add dependency on arm-native.h.
14020 2018-07-13  Richard Biener  <rguenther@suse.de>
14022         PR middle-end/85974
14023         * match.pd (addr1 - addr2): Allow either of the operand to
14024         have a conversion.
14026 2018-07-13  Tom de Vries  <tdevries@suse.de>
14028         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
14029         in remap_ssa_name.
14031 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
14033         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
14034         arrays instead of numbered variables.
14036 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
14038         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
14039         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
14041 2018-07-13  Richard Biener  <rguenther@suse.de>
14043         PR debug/86452
14044         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
14045         instead of get_context_die.
14047 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
14048             Richard Biener  <rguenther@suse.de>
14050         PR middle-end/86489
14051         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
14052         that the loop latch destination where phi is defined.
14054 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
14056         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
14057         (riscv_expand_epilogue): Add assertion to check interrupt mode.
14058         (riscv_set_current_function): Extract getting interrupt type to new
14059         function.
14060         (riscv_get_interrupt_type): New function.
14061         (riscv_merge_decl_attributes): New function, checking interrupt type is
14062         same.
14063         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
14065 2018-07-12  Paul Koning  <ni1d@arrl.net>
14067         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
14068         directive.
14070 2018-07-12  Paul Koning  <ni1d@arrl.net>
14072         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
14073         zero reference, add doloop_end instead.
14074         * doc/md.texi (decrement_and_branch_until_zero): Remove.
14075         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
14076         detail for doloop_end.
14078 2018-07-12  Martin Sebor  <msebor@redhat.com>
14080         PR c/86453
14081         * attribs.c (decl_attributes): Reject conflicting attributes before
14082         calling attribute handlers.
14084 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
14086         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
14087          parameter.
14088         (gcc::dump_manager::get_dump_file_name): likewise.
14089         (dump_begin): Likewise.
14090         * dumpfile.h (dump_begin): Update prototype.
14091         (gcc::dump_manager::get_dump_file_name,
14092         gcc::dump_manager::get_dump_file_name): Update prototype.
14094 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
14096         * internal-fn.h (vectorizable_internal_fn_p): New function.
14097         * tree-vect-slp.c (compatible_calls_p): Likewise.
14098         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
14099         to internal functions.
14100         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
14102 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
14104         * fold-const.h (inverse_conditions_p): Declare.
14105         * fold-const.c (inverse_conditions_p): New function.
14106         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
14107         that test the inverse condition of a conditional internal function.
14108         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
14109         * internal-fn.c (internal_fn_mask_index): Handle conditional
14110         internal functions.
14111         (vectorized_internal_fn_supported_p): New function.
14112         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
14113         (any_pred_load_store): Replace with...
14114         (need_to_predicate): ...this new variable.
14115         (redundant_ssa_names): New variable.
14116         (ifcvt_can_use_mask_load_store): Move initial checks to...
14117         (ifcvt_can_predicate): ...this new function.  Handle tree codes
14118         for which a conditional internal function exists.
14119         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
14120         instead of ifcvt_can_use_mask_load_store.  Update after variable
14121         name change.
14122         (predicate_load_or_store): New function, split out from
14123         predicate_mem_writes.
14124         (check_redundant_cond_expr): New function.
14125         (value_available_p): Likewise.
14126         (predicate_rhs_code): Likewise.
14127         (predicate_mem_writes): Rename to...
14128         (predicate_statements): ...this.  Use predicate_load_or_store
14129         and predicate_rhs_code.
14130         (combine_blocks, tree_if_conversion): Update after above name changes.
14131         (ifcvt_local_dce): Handle redundant_ssa_names.
14132         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
14133         general conditional functions.
14134         * tree-vect-stmts.c (vectorizable_call): Likewise.
14136 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
14137             Alan Hayward  <alan.hayward@arm.com>
14138             David Sherwood  <david.sherwood@arm.com>
14140         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
14141         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
14142         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
14143         plus and minus and convert them into IFN_COND_FMA-based sequences.
14144         (convert_mult_to_fma): Handle conditional plus and minus.
14146 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
14148         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
14149         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
14150         (cond_fnms_optab): New optabs.
14151         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
14152         internal functions.
14153         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
14154         * internal-fn.h (get_conditional_internal_fn): Declare.
14155         (get_unconditional_internal_fn): Likewise.
14156         * internal-fn.c (cond_ternary_direct): New macro.
14157         (expand_cond_ternary_optab_fn): Likewise.
14158         (direct_cond_ternary_optab_supported_p): Likewise.
14159         (FOR_EACH_COND_FN_PAIR): Likewise.
14160         (get_conditional_internal_fn): New function.
14161         (get_unconditional_internal_fn): Likewise.
14162         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
14163         (gimple_match_op::gimple_match_op): Add a new overload for 5
14164         operands.
14165         (gimple_match_op::set_op): Likewise.
14166         (gimple_resimplify5): Declare.
14167         * genmatch.c (decision_tree::gen): Generate simplifications for
14168         5 operands.
14169         * gimple-match-head.c (gimple_simplify): Define an overload for
14170         5 operands.  Handle calls with 5 arguments in the top-level overload.
14171         (convert_conditional_op): Handle conversions from unconditional
14172         internal functions to conditional ones.
14173         (gimple_resimplify5): New function.
14174         (build_call_internal): Pass a fifth operand.
14175         (maybe_push_res_to_seq): Likewise.
14176         (try_conditional_simplification): Try converting conditional
14177         internal functions to unconditional internal functions.
14178         Handle 3-operand unconditional forms.
14179         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
14180         Define ternary equivalents of the current rules for binary conditional
14181         internal functions.
14182         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
14183         ternary operations.
14184         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
14185         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
14186         (optab): Handle them.
14187         (SVE_COND_FP_TERNARY): New int iterator.
14188         (sve_fmla_op, sve_fmad_op): New int attributes.
14189         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
14190         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
14191         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
14193 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
14195         * target.def (preferred_else_value): New target hook.
14196         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
14197         * doc/tm.texi: Regenerate.
14198         * targhooks.h (default_preferred_else_value): Declare.
14199         * targhooks.c (default_preferred_else_value): New function.
14200         * internal-fn.h (conditional_internal_fn_code): Declare.
14201         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
14202         (get_conditional_internal_fn): Use it.
14203         (conditional_internal_fn_code): New function.
14204         * gimple-match.h (gimple_match_cond): New struct.
14205         (gimple_match_op): Add a cond member function.
14206         (gimple_match_op::gimple_match_op): Update all forms to take a
14207         gimple_match_cond.
14208         * genmatch.c (expr::gen_transform): Use the same condition as res_op
14209         for the suboperation, but don't specify a particular else_value.
14210         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
14211         (visit_nary_op, visit_reference_op_load): Pass
14212         gimple_match_cond::UNCOND to the gimple_match_op constructor.
14213         * gimple-match-head.c: Include tree-eh.h
14214         (convert_conditional_op): New function.
14215         (maybe_resimplify_conditional_op): Likewise.
14216         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
14217         (gimple_resimplify2): Likewise.
14218         (gimple_resimplify3): Likewise.
14219         (gimple_resimplify4): Likewise.
14220         (maybe_push_res_to_seq): Return null for conditional operations.
14221         (try_conditional_simplification): New function.
14222         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
14223         constructor.
14224         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
14225         IFN_COND_* call.
14226         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
14227         function.
14228         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
14230 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
14232         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
14233         DECL_FCONTEXT
14234         (hash_tree): Do not hash DECL_FCONTEXT
14235         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
14236         Do not stream DECL_FCONTEXT.
14237         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
14238         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
14240 2018-07-12  Richard Biener  <rguenther@suse.de>
14242         PR debug/86462
14243         * dwarf2out.c (gen_block_die): Only output blocks when they have
14244         at least one !DECL_IGNORED_P variable.
14246 2018-07-12  Richard Biener  <rguenther@suse.de>
14248         PR target/84829
14249         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
14250         Remove -mieee-fp handling.
14252 2018-07-12  Richard Biener  <rguenther@suse.de>
14254         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
14255         left-over from last patch.
14257 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
14259         PR tree-optimization/86492
14260         * gimple-ssa-store-merging.c
14261         (imm_store_chain_info::coalesce_immediate_stores): Call
14262         check_no_overlap even for the merge_overlapping case.  Formatting fix.
14264 2018-07-12  Richard Biener  <rguenther@suse.de>
14266         PR middle-end/86479
14267         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
14268         move possibly trapping operations into the conditional.
14270 2018-07-12  Richard Biener  <rguenther@suse.de>
14272         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
14273         (vn_lookup_simplify_result): Remove recursion limit applied
14274         here.
14275         (vn_nary_build_or_lookup_1): Adjust.
14276         (try_to_simplify): Likewise.
14277         * gimple-match-head.c (gimple_resimplify1): Instead apply one
14278         here.
14279         (gimple_resimplify2): Likewise.
14280         (gimple_resimplify3): Likewise.
14281         (gimple_resimplify4): Likewise.
14283 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
14285         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
14286         Use __mmask64 type instead of __mmask8 for __M argument.
14287         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
14288         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
14289         __U argument.
14290         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
14291         __mmask16 for __M argument.
14292         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
14293         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
14294         to __mmask16 instead of __mmask8.
14295         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
14296         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
14297         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
14298         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
14299         instead of __mmask16 for __U argument.
14300         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
14301         __mmask16 instead of __mmask8 for __U argument.
14302         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
14303         __U argument.
14304         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
14305         __mmask16.
14306         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
14307         argument.
14308         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
14309         __U argument.
14310         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
14311         __mmask16.
14312         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
14313         of __mmask16.
14314         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
14315         __U argument.
14316         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
14317         __U argument.
14318         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
14319         __U argument.
14320         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
14321         __U argument.
14322         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
14323         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
14324         return type as well as __M argument type and all casts from __mmask8
14325         to __mmask32.
14326         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
14327         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
14328         return type as well as __M argument type and all casts from __mmask8
14329         to __mmask16.
14330         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
14331         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
14332         return type as well as __M argument type and all casts from __mmask8
14333         to __mmask32.
14334         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
14335         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
14336         return type as well as __M argument type and all casts from __mmask8
14337         to __mmask16.
14338         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
14339         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
14340         __mmask16.
14342 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
14344         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
14345         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
14346         for __U argument.
14348 2018-07-11  Paul Koning  <ni1d@arrl.net>
14350         * doc/md.texi (define_subst): Document how multiple occurrences of
14351         the same argument in the replacement pattern are handled.
14353 2018-07-11  Paul Koning  <ni1d@arrl.net>
14355         * doc/extend.texi (Common Variable Attributes): Move "mode" into
14356         alphabetical order.
14357         (Common Type Attributes): Add "mode" attribute.
14359 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
14361         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
14362         stream DECL_ORIGINAL_TYPE.
14363         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
14364         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
14365         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
14366         Do not walk original type.
14367         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
14368         external decls.
14369         (write_ts_decl_non_common_tree_pointers): Do not stream
14370         DECL_ORIGINAL_TYPE
14371         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
14372         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
14374 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
14376         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
14377         thread twice from the same starting edge.
14379 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
14381         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
14382         * gimple.c (gimple_call_nonnull_result_p): ...here...
14383         (gimple_call_nonnull_arg): ...and here.
14384         * gimple.h (gimple_call_nonnull_result_p): New.
14385         (gimple_call_nonnull_arg): New.
14387 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
14389         * config/arm/arm-cpus.in: Move information from fpu field of each
14390         cpu definition to the isa field.
14391         * config/arm/parsecpu.awk (fpu): Delete match rule.
14392         (gen_comm_data): Don't add bits from the CPU's FPU entry.
14394 2018-07-11  Richard Biener  <rguenther@suse.de>
14396         PR debug/86457
14397         * dwarf2out.c (init_sections_and_labels): Use
14398         output_asm_line_debug_info consistently.
14399         (dwarf2out_early_finish): Likewise.
14400         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
14401         type units.
14403 2018-07-11  Richard Biener  <rguenther@suse.de>
14405         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
14406         Rework father_bb setting in a way to avoid propagating constants
14407         multiple times on a loop body.
14409 2018-07-10  Mark Wielaard  <mark@klomp.org>
14411         PR debug/86459
14412         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
14414 2018-07-10  Richard Biener  <rguenther@suse.de>
14416         * hash-map.h (hash_map::iterator::operator*): Return
14417         references to key and value.
14419 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
14421         PR c++/86443
14422         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
14423         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
14424         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
14425         (gimplify_omp_for): For composite loops, move outer
14426         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
14427         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
14428         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
14429         TREE_LIST for both the original class iterator and the "last" helper
14430         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
14431         loop, remember has_decl_expr from outer composite loops for the
14432         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
14434 2018-07-09  Martin Sebor  <msebor@redhat.com>
14436         PR middle-end/77357
14437         PR middle-end/86428
14438         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
14439         accessing implicitly initialized array elements.
14440         * expr.c (string_constant): Handle string initializers of
14441         character arrays within aggregates.
14442         * gimple-fold.c (fold_array_ctor_reference): Add argument.
14443         Store element offset.  As a special case, handle zero size.
14444         (fold_nonarray_ctor_reference): Same.
14445         (fold_ctor_reference): Add argument.  Store subobject offset.
14446         * gimple-fold.h (fold_ctor_reference): Add argument.
14448 2018-07-09  Paul Koning  <ni1d@arrl.net>
14450         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
14451         (pdp11_insn_cost): New function.
14452         (pdp11_md_asm_adjust): New function.
14453         (TARGET_INVALID_WITHIN_DOLOOP): Define.
14454         (pdp11_rtx_costs): Update to match machine better.
14455         (output_addr_const_pdp11): Correct format mismatch warnings.
14456         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
14457         * config/pdp11/pdp11.md: General change to add base_cost and/or
14458         length attributes for use by new pdp11_insn_cost function.
14459         (MIN_BRANCH): Correct definition.
14460         (MIN_SOB): Ditto.
14461         (doloop_end): Use standard pattern name for looping pattern.
14462         (doloop_end_nocc): New.
14463         (movsf): Add another constraint alternative.
14464         (zero_extendqihi2): Add constraint alternatives for not in place
14465         extend.
14466         (zero_extendhisi2): Remove.
14467         (shift patterns): Add CC handling variants.
14468         (bswaphi2): New.
14469         (bswapsi2): New.
14470         (rothi3): New.
14471         (define_peephole2): New peephole to recognize mov that sets CC for
14472         subsequent test.
14474 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
14476         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
14477         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
14478         wi::add.
14480 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
14482         PR c/86420
14483         * real.c (real_nextafter): Return true if result is denormal.
14485 2018-07-09  Martin Liska  <mliska@suse.cz>
14487         * common.opt: Add back wrongly removed attribute.
14489 2018-07-09  Richard Biener  <rguenther@suse.de>
14491         PR debug/86413
14492         * dwarf2out.c (gen_block_die): For an early generated DIE
14493         always output high/low PC attributes.
14495 2018-07-09  Tom de Vries  <tdevries@suse.de>
14497         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
14498         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
14499         onto VAR_DECL with abstract origin.
14501 2018-07-07  Jim Wilson  <jimw@sifive.com>
14503         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
14505 2018-07-07  Tom de Vries  <tdevries@suse.de>
14507         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
14509 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
14511         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
14512         overflow_type.
14513         (combine_bound): Use wide-int overflow calculation instead of
14514         rolling our own.
14515         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
14516         overflow_type.
14517         * fold-const.c (int_const_binop_2): Same.
14518         (extract_muldiv_1): Same.
14519         (fold_div_compare): Same.
14520         (fold_abs_const): Same.
14521         * match.pd: Same.
14522         * poly-int.h (add): Same.
14523         (sub): Same.
14524         (neg): Same.
14525         (mul): Same.
14526         * predict.c (predict_iv_comparison): Same.
14527         * profile-count.c (slow_safe_scale_64bit): Same.
14528         * simplify-rtx.c (simplify_const_binary_operation): Same.
14529         * tree-chrec.c (tree_fold_binomial): Same.
14530         * tree-data-ref.c (split_constant_offset_1): Same.
14531         * tree-if-conv.c (idx_within_array_bound): Same.
14532         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
14533         * tree-ssa-phiopt.c (minmax_replacement): Same.
14534         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
14535         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
14536         * vr-values.c (vr_values::adjust_range_with_scev): Same.
14537         * wide-int.cc (wi::add_large): Same.
14538         (wi::mul_internal): Same.
14539         (wi::sub_large): Same.
14540         (wi::divmod_internal): Same.
14541         * wide-int.h: Change overflow type to overflow_type for neg, add,
14542         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
14543         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
14544         mul_internal, divmod_internal.
14545         (overflow_type): New enum.
14546         (accumulate_overflow): New.
14548 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
14550         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
14551         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
14553 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
14555         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
14556         argument is checked for zero before entering loop, avoid checking again.
14558 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
14560         * gimplify.h (generic_expr_could_trap_p): Set as global function.
14561         * gimplify.h (generic_expr_could_trap_p): Likwise.
14562         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
14564 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
14566         PR tree-optimization/86401
14567         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
14568         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
14569         (fold_bit_and_mask): ... here.  New helper function for match.pd.
14570         * fold-const.h (fold_bit_and_mask): Declare.
14571         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
14573 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
14575         PR target/86324
14576         * target.def (translate_mode_attribute): New hook.
14577         * targhooks.h (default_translate_mode_attribute): Declare.
14578         * targhooks.c (default_translate_mode_attribute): New function.
14579         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
14580         * doc/tm.texi: Regenerate.
14581         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
14582         (rs6000_translate_mode_attribute): New function.
14584 2018-07-06  Paul Koning  <ni1d@arrl.net>
14586         * doc/md.texi (define_split): Document DONE and FAIL.
14587         (define_peephole2): Ditto.
14589 2018-07-05  Jeff Law  <law@redhat.com>
14591         PR tree-optimization/86010
14592         * tree-ssa-dse.c (compute_trims): More aggressively trim at
14593         both the head and tail of mem* and str* calls.
14595 2018-07-05  Jim Wilson  <jimw@sifive.com>
14597         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
14599 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
14601         * config/aarch64/aarch64-simd.md: correct flags text for
14602         MIN_EXPR replacement.
14604 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
14606         * configure: Regenerated.
14608 2018-07-05  Carl Love  <cel@us.ibm.com>
14610         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
14611         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
14612         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
14613         VSX_BUILTIN_DOUBLEL_V4SF.
14615 2018-07-05  Martin Sebor  <msebor@redhat.com>
14617         PR c++/86400
14618         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
14619         than its domain to compute its the upper bound of a char array.
14621 2018-07-05  Nathan Sidwell  <nathan@acm.org>
14623         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
14624         * doc/cpp.texi: Update comment.
14625         * doc/tm.texi: Rebuilt.
14626         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
14627         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
14628         * doc/extend.texi (Backwards Compatibility): Clarify it is system
14629         headers affected by extern "C".
14630         * system.h: Poison NO_IMPLICIT_EXTERN_C.
14631         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
14632         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
14633         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
14634         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
14635         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
14636         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
14637         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
14638         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
14639         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
14640         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
14641         config/sparc/sp64-elf.h, config/spu/spu.h,
14642         config/stormy16/stormy16.h, config/v850/v850.h,
14643         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
14644         define NO_IMPLICIT_EXTERN_C.
14645         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
14647 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
14649         PR target/84711
14650         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14651         instead of GET_MODE_SIZE when comparing Units.
14653 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
14655         PR target/84711
14656         * rtlanal.c (set_noop_p): Constrain on mode change,
14657         include hard-reg-set.h
14659 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
14661         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
14663 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
14665         Revert
14666         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
14668         PR sanitizer/84250
14669         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
14670         libasan.
14671         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
14673 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
14675         PR sanitizer/84250
14676         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
14677         libasan.
14678         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
14680 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
14682         PR middle-end/86380
14683         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
14685 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
14687         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
14688         neg_*_op* variables.
14690 2018-07-04  Martin Liska  <mliska@suse.cz>
14692         * tree-switch-conversion.c: Define
14693         max_ratio_for_speed and max_ratio_for_size constants.
14695 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
14696             Martin Liska  <mliska@suse.cz>
14698         PR middle-end/66240
14699         PR target/45996
14700         PR c/84100
14701         * common.opt: Rename align options with 'str_' prefix.
14702         * common/config/i386/i386-common.c (set_malign_value): New
14703         function.
14704         (ix86_handle_option): Use it to set -falign-* options/
14705         * config/aarch64/aarch64-protos.h (struct tune_params): Change
14706         type from int to string.
14707         * config/aarch64/aarch64.c: Update default values from int
14708         to string.
14709         * config/alpha/alpha.c (alpha_override_options_after_change):
14710         Likewise.
14711         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
14712         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
14713         max skip conditionally.
14714         * config/i386/freebsd.h (SUBALIGN_LOG): New.
14715         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
14716         max skip conditionally.
14717         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
14718         max skip conditionally.
14719         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
14720         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
14721         max skip conditionally.
14722         * config/i386/i386.c (struct ptt): Change type from int to
14723         string.
14724         (ix86_default_align): Set default values.
14725         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
14726         max skip conditionally.
14727         * config/i386/iamcu.h (SUBALIGN_LOG): New.
14728         (ASM_OUTPUT_MAX_SKIP_ALIGN):
14729         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
14730         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
14731         max skip conditionally.
14732         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
14733         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
14734         * config/i386/x86-64.h (SUBALIGN_LOG): New.
14735         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
14736         max skip conditionally.
14737         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
14738         * config/ia64/ia64.c (ia64_option_override): Set default values
14739         for alignment options.
14740         * config/m68k/m68k.c: Handle new str_align_* options.
14741         * config/mips/mips.c (mips_set_compression_mode): Change
14742         type of constants.
14743         (mips_option_override): Set default values for options.
14744         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
14745         Likewise.
14746         * config/rs6000/rs6000.c (rs6000_option_override_internal):
14747         Likewise.
14748         * config/rx/rx.c (rx_option_override): Likewise.
14749         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
14750         (LABEL_ALIGN): Use align_labels_log.
14751         (LOOP_ALIGN): Use align_loops_align.
14752         * config/s390/s390.c (s390_asm_output_function_label): Use new
14753         macros.
14754         * config/sh/sh.c (sh_override_options_after_change):
14755         Change type of constants.
14756         * config/spu/spu.c (spu_sched_init): Likewise.
14757         * config/sparc/sparc.c (sparc_option_override): Set default
14758         values for options.
14759         * config/visium/visium.c (visium_option_override): Likewise.
14760         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
14761         emit p2align format with last argument if it's not needed.
14762         * doc/invoke.texi: Document extended format of -falign-*.
14763         * final.c: Use align_labels alignment.
14764         * flags.h (struct target_flag_state): Change type to use
14765         align_flags.
14766         (struct align_flags_tuple): New.
14767         (struct align_flags): Likewise.
14768         (align_loops_log): Redefine macro to use new types.
14769         (align_loops_max_skip): Redefine macro to use new types.
14770         (align_jumps_log): Redefine macro to use new types.
14771         (align_jumps_max_skip): Redefine macro to use new types.
14772         (align_labels_log): Redefine macro to use new types.
14773         (align_labels_max_skip): Redefine macro to use new types.
14774         (align_functions_log): Redefine macro to use new types.
14775         (align_loops): Redefine macro to use new types.
14776         (align_jumps): Redefine macro to use new types.
14777         (align_labels): Redefine macro to use new types.
14778         (align_functions): Redefine macro to use new types.
14779         (align_functions_max_skip): Redefine macro to use new types.
14780         (align_loops_value): New macro.
14781         (align_jumps_value): New macro.
14782         (align_labels_value): New macro.
14783         (align_functions_value): New macro.
14784         * function.c (invoke_set_current_function_hook): Propagate
14785         alignment values from flags to global variables default in
14786         topleev.h.
14787         * ipa-icf.c (sem_function::equals_wpa): Use
14788         cl_optimization_option_eq instead of memcmp.
14789         * lto-streamer.h (cl_optimization_stream_out): Support streaming
14790         of string types.
14791         (cl_optimization_stream_in): Likewise.
14792         * optc-save-gen.awk: Support strings in cl_optimization.
14793         * opth-gen.awk: Likewise.
14794         * opts.c (finish_options): Remove error checking of invalid
14795         value ranges.
14796         (MAX_CODE_ALIGN): Remove.
14797         (MAX_CODE_ALIGN_VALUE): Likewise.
14798         (parse_and_check_align_values): New function.
14799         (check_alignment_argument): Likewise.
14800         (common_handle_option): Use check_alignment_argument.
14801         * opts.h (parse_and_check_align_values): Declare.
14802         * toplev.c (init_alignments): Remove.
14803         (read_log_maxskip): New.
14804         (parse_N_M): Likewise.
14805         (parse_alignment_opts): Likewise.
14806         (backend_init_target): Remove usage of init_alignments.
14807         * toplev.h (parse_alignment_opts): Declare.
14808         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
14809         argument.
14810         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
14811         * tree.c (cl_option_hasher::equal): New.
14812         * varasm.c: Use new global macros.
14814 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
14816         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
14817         Use a simpler align directive also if MAXSKIP = ALIGN-1.
14818         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14819         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14820         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14821         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
14822         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
14823         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
14824         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
14825         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14826         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14827         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14828         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14830 2018-07-04  Martin Liska  <mliska@suse.cz>
14831             Jonathan Wakely  <jwakely@redhat.com>
14833         * coverage.c: Use correct type.
14834         * doc/invoke.texi: Language correction.
14836 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
14838         PR target/85620
14839         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
14840         ENDBRANCH for non-tail call which may return via indirect branch.
14841         * doc/extend.texi: Document indirect_return attribute.
14843 2018-07-03  Martin Sebor  <msebor@redhat.com>
14845         PR tree-optimization/86274
14846         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
14847         precondition.
14848         (format_floating): Correct handling of infinities and NaNs.
14850 2018-07-03  Martin Sebor  <msebor@redhat.com>
14852         * print-tree.c (print_real_cst): New function.
14853         (print_node_brief): Call it.
14854         (print_node): Ditto.
14856 2018-07-03  Jeff Law  <law@redhat.com>
14858         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
14859         into a single pattern.
14861         * config/h8300/h8300.md (ors code_iterator): New.
14862         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
14863         a single pattern and single splitter.
14864         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
14865         (iorqi3_1, xorqi3_1): Likewise.
14866         (iorqi3, xorqi3 expanders): Similarly.
14868         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
14869         (movmd_internal) into a single pattern using the P mode iterator.
14870         (movmd splitters): Similarly.
14871         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
14872         (movsd splitters): Similarly.
14874         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
14875         ADDB, ADDW and ADDL into a single ADD attribute which selects the
14876         right table based on the size of the operand.
14877         * config/h8300/h8300.md (length_table): Corresponding changes. All
14878         references to "addb", "addw" and "addl" changed to "add".
14879         (btst patterns): Merge two variants into a single pattern.
14880         (tstqi, tsthi): Likewise.
14881         (addhi3_incdec, addsi3_incdec): Likewise.
14882         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
14883         (mulhi3, mulsi3): Likewise.
14884         (udivhi3, udivsi3): Likewise.
14885         (divhi3, divsi3): Likewise.
14886         (andorqi3, andorhi3, andorsi3): Likewise.
14888 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
14890         PR target/85694
14891         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
14892         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
14894 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
14896         PR tree-optimization/85694
14897         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
14898         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
14899         UNSPEC_URHADD.
14900         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
14901         (<u>avg<mode>3_ceil): New patterns.
14903 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
14905         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
14906         scan-tree-dump directive.
14907         * gcc.dg/vect/slp-perm-2.c: Likewise.
14908         * gcc.dg/vect/slp-perm-3.c: Likewise.
14909         * gcc.dg/vect/slp-perm-5.c: Likewise.
14910         * gcc.dg/vect/slp-perm-6.c: Likewise.
14911         * gcc.dg/vect/slp-perm-7.c: Likewise.
14912         * gcc.dg/vect/slp-perm-8.c: Likewise.
14914 2018-07-03  Marek Polacek  <polacek@redhat.com>
14916         PR middle-end/86202
14917         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
14919 2018-07-03  Richard Biener  <rguenther@suse.de>
14921         PR ipa/86389
14922         * tree-ssa-structalias.c (find_func_clobbers): Properly
14923         handle indirect calls.
14925 2018-07-03  Jeff Law  <law@redhat.com>
14927         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
14928         (shifts): New code iterator.
14929         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
14930         expander.  Fix HImode handling on H8/SX.
14931         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
14932         (subqi3, subhi3, subsi3 expanders): Likewise.
14933         (andqi3, andhi3, andsi3 expanders): Likewise.
14934         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
14935         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
14936         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
14937         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
14938         (zero_extendqihi2, zero_extendqisi2): Likewise.
14939         (extendqihi2, extendqisi2): Likewise.
14940         (rotlqi3, rotlhi3, rotlsi3): Likewise.
14941         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
14942         (rotlqi3_1, rotlhi3_1): Likewise.
14943         (logicalhi3_sn, logicalsi3_sn): Likewise.
14944         (logicalhi3, logicalsi3): Likewise.
14946 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
14948         * tree-vect-patterns.c (vect_recog_rotate_pattern)
14949         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
14950         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
14951         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
14952         type to append_pattern_def_seq instead of creating a stmt_vec_info
14953         directly.
14954         (build_mask_conversion): Likewise.  Remove vinfo argument.
14955         (vect_add_conversion_to_patterm): Likewise, renaming to...
14956         (vect_add_conversion_to_pattern): ...this.
14957         (vect_recog_mask_conversion_pattern): Update call to
14958         build_mask_conversion.  Pass the vector type to
14959         append_pattern_def_seq here too.
14960         (vect_recog_gather_scatter_pattern): Update call to
14961         vect_add_conversion_to_pattern.
14963 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
14965         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
14966         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14967         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
14968         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
14969         STMT_VINFO_PATTERN_DEF_SEQ to null here.
14970         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
14971         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
14972         append_pattern_def_seq instead of new_pattern_def_seq.
14973         (vect_recog_divmod_pattern): Do both of the above.
14974         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
14975         is null.
14977 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
14979         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
14980         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
14981         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
14982         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
14983         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
14984         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
14985         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
14986         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
14987         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
14988         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
14989         parameter with a single stmt_vec_info.
14990         (vect_recog_func_ptr): Likewise.
14991         (vect_recog_gather_scatter_pattern): Likewise, folding in...
14992         (vect_try_gather_scatter_pattern): ...this.
14993         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
14994         the stmt_vec_info of the statement to be matched.  Don't clear
14995         STMT_VINFO_RELATED_STMT.
14996         (vect_pattern_recog): Update call accordingly.
14998 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
15000         PR tree-optimization/85694
15001         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
15002         (uavgM3_ceil): Document new optabs.
15003         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
15004         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
15005         functions.
15006         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
15007         (savg_ceil_optab): New optabs.
15008         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
15009         (vect_vect_recog_func_ptrs): Add it.
15010         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
15011         constant directly from the associated lhs.
15013 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
15015         * tree-vect-patterns.c (vect_split_statement): New function.
15016         (vect_convert_input): Use it to try to split an existing cast.
15018 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
15020         * poly-int.h (print_hex): New function.
15021         * dumpfile.h (dump_dec, dump_hex): Declare.
15022         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
15023         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
15024         min_input_precision, operation_precision and operation_sign.
15025         * tree-vect-patterns.c (vect_get_range_info): New function.
15026         (vect_same_loop_or_bb_p, vect_single_imm_use)
15027         (vect_operation_fits_smaller_type): Delete.
15028         (vect_look_through_possible_promotion): Add an optional
15029         single_use_p parameter.
15030         (vect_recog_over_widening_pattern): Rewrite to use new
15031         stmt_vec_info infomration.  Handle one operation at a time.
15032         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
15033         (vect_truncatable_operation_p, vect_set_operation_type)
15034         (vect_set_min_input_precision): New functions.
15035         (vect_determine_min_output_precision_1): Likewise.
15036         (vect_determine_min_output_precision): Likewise.
15037         (vect_determine_precisions_from_range): Likewise.
15038         (vect_determine_precisions_from_users): Likewise.
15039         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
15040         (vect_vect_recog_func_ptrs): Put over_widening first.
15041         Add cast_forwprop.
15042         (vect_pattern_recog): Call vect_determine_precisions.
15044 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
15046         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
15047         statements that have been replaced by further pattern statements.
15048         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
15050 2018-07-03  Richard Biener  <rguenther@suse.de>
15052         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
15053         always set *dt.  Dump vectype in vectype overload.
15054         * dumpfile.h (dump_gimple_expr): New function.
15055         (dump_gimple_expr_loc): Likewise.
15056         * dumpfile.c (dump_gimple_expr): New function.
15057         (dump_gimple_expr_loc): Likewise.
15059 2018-07-02  Jeff Law  <law@redhat.com>
15061         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
15062         the H8/300, H8/300H and H8/S variants into a single pattern.
15063         (movhi_h8300, movqi_h8300hs): Similarly.
15064         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
15065         (QHI mode iterator): New.
15067         * config/h8300/h8300.md: Remove trailing whitespace.
15069 2018-07-02  Jim Wilson  <jimw@sifive.com>
15071         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
15072         instead of emit_insn for interrupt returns.
15073         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
15074         (riscv_sret, riscv_uret): Likewise.
15076 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
15078         * pretty-print.c (selftest::test_pp_format): Move save and restore
15079         of quotes to class auto_fix_quotes, and add an instance.
15080         * selftest.c: Include "intl.h".
15081         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
15082         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
15083         * selftest.h (selftest::auto_fix_quotes): New class.
15085 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
15087         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
15088         (aarch64_sve_prepare_conditional_op): Remove.
15089         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
15090         Allow aarch64_simd_reg_or_zero as select operand; remove
15091         the aarch64_sve_prepare_conditional_op call.
15092         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
15093         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
15094         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
15095         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
15096         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
15097         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
15098         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
15099         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
15100         and a splitters to match all of the *_any patterns.
15101         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
15103         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
15104         (SVE_COND_FP_BINARY_REV): Remove.
15105         (sve_int_op_rev, sve_fp_op_rev): New.
15106         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
15107         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
15108         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
15109         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
15110         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
15111         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
15112         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
15113         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
15114         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
15116         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
15117         Remove match_dup 1 from the inner unspec.
15118         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
15120         * config/aarch64/aarch64.md (movprfx): New attr.
15121         (length): Default movprfx to 8.
15122         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
15123         (*madd<SVE_I>, *msub<SVE_I): Likewise.
15124         (*<su>mul<SVE_I>3_highpart): Likewise.
15125         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
15126         (*v<ASHIFT><SVE_I>3): Likewise.
15127         (*<su><MAXMIN><SVE_I>3): Likewise.
15128         (*<su><MAXMIN><SVE_F>3): Likewise.
15129         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
15130         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
15131         (*div<SVE_F>4): Likewise.
15133 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
15135         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
15136         in dump string.
15138 2018-07-02  Richard Biener  <rguenther@suse.de>
15140         PR tree-optimization/86363
15141         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
15142         memset argument refers to a non-variable address.
15144 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
15146         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
15147         {PLUS,MINUS}_EXPR code to...
15148         (adjust_symbolic_bound): ...here,
15149         (combine_bound): ...here,
15150         (set_value_range_with_overflow): ...and here.
15152 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
15154         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
15155         code...
15156         (extract_range_from_abs_expr): ...here.
15158 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
15160         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
15161         -fno-omit-frame-pointer when not optimizing.
15163 2018-07-02  Martin Liska  <mliska@suse.cz>
15165         PR ipa/86279
15166         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
15167         (propagate_nothrow): Likewise.
15169 2018-07-02  Martin Liska  <mliska@suse.cz>
15171         PR ipa/86323
15172         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
15174 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
15176         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
15177         function in r262149, changing "loc" param from source_location to
15178         const dump_location_t &.
15179         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
15180         declaration, as above.
15182 2018-07-01  Paul Koning  <ni1d@arrl.net>
15184         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
15185         -munit-asm, -mgnu-asm, -mdec-asm.
15186         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
15187         (pdp11_output_labelref): New.
15188         (pdp11_output_def): New.
15189         (pdp11_output_addr_vec_elt): New.
15190         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
15191         %# and %@ format codes.
15192         (pdp11_option_override): New.
15193         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
15194         (pdp11_output_ident): New.
15195         (pdp11_asm_named_section): New.
15196         (pdp11_asm_init_sections): New.
15197         (pdp11_file_start): New.
15198         (pdp11_file_end): New.
15199         (output_ascii): Use .ascii/.asciz for -mdec-asm.
15200         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
15201         %o, like %c but octal.
15202         (pdp11_option_override): New.
15203         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
15204         -mdec-asm.
15205         (DATA_SECTION_ASM_OP): Ditto.
15206         (READONLY_DATA_SECTION_ASM_OP): New.
15207         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
15208         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
15209         (ASM_OUTPUT_LABELREF): Ditto.
15210         (ASM_OUTPUT_DEF): Ditto.
15211         (ASM_OUTPUT_EXTERNAL): New.
15212         (ASM_OUTPUT_SOURCE_FILENAME): New.
15213         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
15214         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
15215         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
15216         %# and %@ format codes.
15217         * config/pdp11/pdp11.opt (mgnu-asm): New.
15218         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
15219         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
15220         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
15222 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
15224         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
15225         dereferencing path[] beyond its length.
15226         (debug_path): New.
15227         (debug_all_paths): New.
15228         (rewire_first_differing_edge): New.
15229         (adjust_paths_after_duplication): New.
15230         (duplicate_thread_path): Call adjust_paths_after_duplication.
15231         Add new argument.
15232         (thread_through_all_blocks): Add new argument to
15233         duplicate_thread_path.
15235 2018-06-30  Jim Wilson  <jimw@sifive.com>
15237         * config/riscv/predicates.md (p2m1_shift_operand): New.
15238         (high_mask_shift_operand): New.
15239         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
15240         pattern using p2m1_shift_operand.
15241         (lshsi3_zero_extend_3+2): New combiner pattern using
15242         high_mask_shift_operand.
15244 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
15246         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
15247         split out from...
15248         (vect_recog_rotate_pattern): ...here.
15249         (vect_convert_input): Try to insert casts of invariants in the
15250         preheader.
15251         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
15252         preheader to be empty.
15254 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
15256         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
15257         vector type.  If given, install it in the new statement's
15258         STMT_VINFO_VECTYPE.
15259         (vect_element_precision): New function.
15260         (vect_unpromoted_value): New struct.
15261         (vect_unpromoted_value::vect_unpromoted_value): New function.
15262         (vect_unpromoted_value::set_op): Likewise.
15263         (vect_look_through_possible_promotion): Likewise.
15264         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
15265         (vect_widened_op_tree, vect_convert_input): Likewise.
15266         (vect_convert_inputs, vect_convert_output): Likewise.
15267         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
15268         to handle the optional cast of the multiplication result and
15269         vect_widened_op_tree to detect the widened multiplication itself.
15270         Do not require the input and output of promotion casts to have
15271         the same sign, but base the signedness of the operation on the
15272         input rather than the result.  If the pattern includes two
15273         promotions, check that those promotions have the same sign.
15274         Do not restrict the MULT_EXPR handling to a double-width result;
15275         handle quadruple-width results and wider.  Use vect_convert_inputs
15276         to convert the inputs to the common type.
15277         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
15278         to handle the optional cast of the ABS result.  Also allow a sign
15279         change or a sign extension between the ABS and MINUS.
15280         Use vect_widened_op_tree to detect the widened subtraction and use
15281         vect_convert_inputs to convert the inputs to the common type.
15282         (vect_handle_widen_op_by_const): Delete.
15283         (vect_recog_widen_op_pattern): New function.
15284         (vect_recog_widen_mult_pattern): Use it.
15285         (vect_recog_widen_shift_pattern): Likewise.
15286         (vect_recog_widen_sum_pattern): Use
15287         vect_look_through_possible_promotion to handle the promoted
15288         PLUS_EXPR operand.
15290 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
15292         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
15293         the containing gimple_seq *.
15294         * gimple-iterator.h (gsi_for_stmt): Declare it.
15295         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
15296         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
15297         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
15298         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
15299         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
15300         checks.
15301         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
15302         split out from...
15303         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
15304         statement being replaced is part of an existing pattern
15305         definition sequence, inserting the new pattern statements before
15306         the original one.
15307         (vect_pattern_recog_1): Don't return a bool.  If the statement
15308         is already part of a pattern, instead apply pattern matching
15309         to the pattern definition statements.  Don't clear the
15310         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
15311         (vect_pattern_recog): Don't break after the first match;
15312         continue processing the pattern definition statements instead.
15313         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
15315 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
15317         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
15318         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
15319         (vect_recog_widen_sum_pattern): Use it.
15321 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
15323         * tree-vect-loop.c (vectorizable_reduction): Assert that the
15324         phi is not a pattern statement and has not been replaced by
15325         a pattern statement.
15326         * tree-vect-patterns.c (type_conversion_p): Don't check
15327         STMT_VINFO_IN_PATTERN_P.
15328         (vect_recog_vector_vector_shift_pattern): Likewise.
15329         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
15330         the pattern statement rather than the original statement; check
15331         directly for a WIDEN_MULT_EXPR here.
15332         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
15333         vect_is_simple_use to return the pattern statement rather
15334         than the original statement; use is_pattern_stmt_p to check
15335         for such a pattern statement.
15336         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
15337         to return the pattern statement rather than the original statement;
15338         don't do the same transformation here.
15339         (vect_is_simple_use): If the defining statement has been replaced
15340         by a pattern statement, return the pattern statement instead.
15341         Remove the corresponding (local) transformation from the vectype
15342         overload.
15344 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
15346         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
15347         end and default to null.
15348         * tree-vect-loop.c (vect_create_epilog_for_reduction)
15349         (vectorizable_reduction): Update calls accordingly, dropping the
15350         gimple ** argument if the passed-back statement isn't needed.
15351         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
15352         (vect_recog_rotate_pattern): Likewise.
15353         (vect_recog_mask_conversion_pattern): Likewise.
15354         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
15355         (vect_mask_constant_operand_p): Likewise.
15356         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
15357         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
15358         (get_group_load_store_type, get_load_store_type): Likewise.
15359         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
15360         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
15361         (vectorizable_conversion, vectorizable_assignment): Likewise.
15362         (vectorizable_shift, vectorizable_operation): Likewise.
15363         (vectorizable_store, vect_is_simple_cond): Likewise.
15364         (vectorizable_condition, vectorizable_comparison): Likewise.
15365         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
15366         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
15367         and move it to the end.  Cope with null def_stmt_outs.
15369 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15371         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
15373 2018-06-29  Jeff Law  <law@redhat.com>
15375         * config/v850/v850.c (v850_legitimate_address_p): Handle large
15376         displacements for TARGET_V850E2V3 and newer.
15377         (TARGET_LRA_P): Remove.  Defaults to LRA now.
15378         * config/v850/v850.md (sign23byte_load): Remove.
15379         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
15380         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
15382 2018-06-29  Martin Liska  <mliska@suse.cz>
15384         PR lto/85759
15385         * coverage.c (coverage_init): Mangle full path name.
15386         * doc/invoke.texi: Document the change.
15387         * gcov-io.c (mangle_path): New.
15388         * gcov-io.h (mangle_path): Likewise.
15389         * gcov.c (mangle_name): Use mangle_path for path mangling.
15391 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15393         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
15394         if starting source register is not even.
15396 2018-06-29  Martin Liska  <mliska@suse.cz>
15398         PR tree-optimization/86263
15399         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
15400         Make edge redirection.
15402 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
15404         * dumpfile.c (dump_loc): Add indentation based on scope depth.
15405         (dump_scope_depth): New variable.
15406         (get_dump_scope_depth): New function.
15407         (dump_begin_scope): New function.
15408         (dump_end_scope): New function.
15409         * dumpfile.h (get_dump_scope_depth): New declaration.
15410         (dump_begin_scope): New declaration.
15411         (dump_end_scope): New declaration.
15412         (class auto_dump_scope): New class.
15413         (AUTO_DUMP_SCOPE): New macro.
15414         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
15415         AUTO_DUMP_SCOPE.
15417 2018-06-29  Richard Biener  <rguenther@suse.de>
15419         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
15420         compute_all_dependences succeeds.
15421         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
15422         exceed --param loop-max-datarefs-for-datadeps.
15424 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
15426         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
15428 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
15430         PR target/86348
15431         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
15432         alternative 0 in preferred_for_speed attribute.
15434 2018-06-28  Paul Koning  <ni1d@arrl.net>
15436         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
15437         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
15438         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
15439         * config/pdp11/pdp11.md: Correct "length" attribute calculation
15440         for shift insn patterns.
15442 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
15444         * cgraph.c (cgraph_node::get_body): Replace assignments to
15445         "dump_file" with calls to set_dump_file.
15446         * dumpfile.c (alt_dump_file): Make static, and group with...
15447         (alt_flags): ...this definition.
15448         (dumps_are_enabled): New variable.
15449         (refresh_dumps_are_enabled): New function.
15450         (set_dump_file): New function.
15451         (set_alt_dump_file): New function.
15452         (gcc::dump_manager::dump_start): Replace assignments to
15453         "dump_file" and "alt_dump_file" with calls to set_dump_file and
15454         set_alt_dump_file.
15455         (gcc::dump_manager::dump_finish): Likewise.
15456         * dumpfile.h (alt_dump_file): Delete decl.
15457         (dumps_are_enabled): New variable decl.
15458         (set_dump_file): New function decl.
15459         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
15460         global.
15461         * tree-nested.c (lower_nested_functions): Replace assignments to
15462         "dump_file" with calls to set_dump_file.
15464 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
15466         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
15467         goto_locus of each outgoing edge of each basic block.
15469 2018-06-28  Richard Biener  <rguenther@suse.de>
15471         * dwarf2out.c (decl_scope_table): Remove.
15472         (push_decl_scope): Likewise.
15473         (pop_decl_scope): Likewise.
15474         (gen_type_die_for_member): Do not call push/pop_decl_scope.
15475         (gen_struct_or_union_type_die): Likewise.
15476         (gen_tagged_type_die): Likewise.
15477         (dwarf2out_init): Do not initialize decl_scope_table.
15478         (dwarf2out_c_finalize): Do not free it.
15480 2018-06-28  Richard Biener  <rguenther@suse.de>
15482         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
15483         deciding whether to not re-use a DIE.
15485 2018-06-28  Richard Biener  <rguenther@suse.de>
15487         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
15488         DW_AT_abstract_origin attribute.
15490 2018-06-28  Martin Liska  <mliska@suse.cz>
15492         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
15493         Use newly introduced constants.
15494         * tree-switch-conversion.h (struct jump_table_cluster):
15495         Define max_ratio_for_size and max_ratio_for_speed.
15497 2018-06-28  Martin Liska  <mliska@suse.cz>
15499         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15500         Add new checking assert to catch invalid state.
15501         (jump_table_cluster::can_be_handled): Handle single case
15502         clusters.
15503         (jump_table_cluster::is_beneficial): Bail out for such case.
15504         (bit_test_cluster::find_bit_tests):
15505         Add new checking assert to catch invalid state.
15506         (bit_test_cluster::can_be_handled): Handle single case
15507         clusters.
15508         (bit_test_cluster::is_beneficial): Bail out for such case.
15509         (switch_decision_tree::analyze_switch_statement):
15510         Fix comment.
15512 2018-06-28  Martin Liska  <mliska@suse.cz>
15514         * common.opt: Introduce -completion option.
15515         * gcc.c (driver_handle_option): Handle it.
15516         (driver::main): Print completions if completion
15517         is set.
15518         * opt-suggestions.c (option_proposer::get_completions):
15519         New function.
15520         (option_proposer::suggest_completion): Likewise.
15521         (option_proposer::find_param_completions): Likewise.
15522         (verify_autocompletions): Likewise.
15523         (test_completion_valid_options): Likewise.
15524         (test_completion_valid_params): Likewise.
15525         (in_completion_p): Likewise.
15526         (empty_completion_p): Likewise.
15527         (test_completion_partial_match): Likewise.
15528         (test_completion_garbage): Likewise.
15529         (opt_proposer_c_tests): Likewise.
15530         * opt-suggestions.h: Declare new functions.
15531         * opts.c (common_handle_option): Handle OPT__completion_.
15532         * selftest-run-tests.c (selftest::run_tests): Add
15533         opt_proposer_c_tests.
15534         * selftest.c (assert_str_startswith): New.
15535         * selftest.h (assert_str_startswith): Likewise.
15536         (opt_proposer_c_tests): New.
15537         (ASSERT_STR_STARTSWITH): Likewise.
15539 2018-06-28  Martin Liska  <mliska@suse.cz>
15541         * Makefile.in: Add opt-suggestions.o.
15542         * gcc-main.c: Include opt-suggestions.h.
15543         * gcc.c (driver::driver): Likewise.
15544         (driver::~driver): Remove m_option_suggestions.
15545         (driver::build_option_suggestions): Moved to option_proposer.
15546         (driver::suggest_option): Likewise.
15547         (driver::handle_unrecognized_options): Use option_proposer.
15548         * gcc.h (class driver): Add new memver m_option_proposer.
15549         * opt-suggestions.c: New file.
15550         * opt-suggestions.h: New file.
15552 2018-06-28  Martin Liska  <mliska@suse.cz>
15554         * vec.h (class auto_string_vec): New (moved from auto_argvec).
15555         (auto_string_vec::~auto_string_vec): Likewise.
15557 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
15559         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
15560         prevent_decl_creation_for_types fields up and add reset_location field.
15561         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
15562         statement if id->reset_location is true.
15563         (copy_edges_for_bb): Do not set goto_locus on the new edges if
15564         id->reset_location is true.
15565         (copy_phis_for_bb): Force input_location on the arguments if
15566         id->reset_location is true.
15567         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
15568         is set on the function to be inlined.
15570 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
15572         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
15574 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
15576         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
15577         registers for Pmode.
15578         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
15579         hard registers for the clobbered pseudo.
15581 2018-06-27  Paul Koning  <ni1d@arrl.net>
15583         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
15584         mutually exclusive options.
15585         * config/pdp11/constraints.md (h): New constraint.
15586         (O): Update definition to match shift code generation.
15587         (D): New constraint.
15588         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
15589         (CCFP): Remove.
15590         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
15591         function.
15592         (output_jump): Change arguments.
15593         (pdp11_fixed_cc_regs): New function.
15594         (pdp11_cc_mode): Ditto.
15595         (pdp11_expand_shift): Ditto.
15596         (pdp11_assemble_shift): Ditto.
15597         (pdp11_small_shift): Ditto.
15598         (pdp11_branch_cost): Remove.
15599         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
15600         from output.
15601         (pdp11_register_move_cost): Update for CC registers.
15602         (pdp11_rtx_costs): Add case for LSHIFTRT.
15603         (pdp11_output_jump): Add CCNZ mode conditional branches.
15604         (notice_update_cc_on_set): Remove.
15605         (pdp11_cc_mode): New function.
15606         (simple_memory_operand): Correct pre/post decrement case.
15607         (no_side_effect_operand): New function.
15608         (pdp11_regno_reg_class): Add CC_REGS class.
15609         (pdp11_fixed_cc_regs): New function.
15610         (pdp11_small_shift): New function.
15611         (pdp11_expand_shift): New function to expand shift insns.
15612         (pdp11_assemble_shift): New function to output shifts.
15613         (pdp11_branch_cost): Remove.
15614         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
15615         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
15616         (WCHAR_TYPE): Ditto.
15617         (PTRDIFF_TYPE): Ditto.
15618         (ADJUST_INSN_LENGTH): New macro.
15619         (FIXED_REGISTERS): Add CC registers.
15620         (CALL_USED_REGISTERS): Ditto.
15621         (reg_class): Ditto.
15622         (REG_CLASS_NAMES): Ditto.
15623         (REG_CLASS_CONTENTS): Ditto.
15624         (SELECT_CC_MODE): Use new function.
15625         (TARGET_FLAGS_REGNUM): New macro.
15626         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
15627         (cc0_reg_rtx): Remove.
15628         (CC_STATUS_MDEP): Remove.
15629         (CC_STATUS_MDEFP_INIT): Remove.
15630         (CC_IN_FPU): Remove.
15631         (NOTICE_UPDATE_CC): Remove.
15632         (REGISTER_NAMES): Add CC registers.
15633         (BRANCH_COST): Change to constant 1.
15634         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
15635         handling.
15636         * config/pdp11/pdp11.opt (mbcopy): Remove.
15637         (mbcopy-builtin): Remove.
15638         (mbranch-cheap): Remove.
15639         (mbranch-expensive): Remove.
15640         * config/pdp11/predicates.md (expand_shift_operand): Update to
15641         match shift code generation.
15642         (ccnz_operator): New predicate.
15643         * doc/invoke.texi (PDP-11 Options): Remove deleted options
15644         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
15645         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
15646         exclusive options.
15647         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
15648         description of O constraint.
15650 2018-06-27  Jeff Law  <law@redhat.com>
15651             Austin Law  <austinklaw@gmail.com>
15653         * config/v850/v850.md (addsi3_set_flags): New pattern.
15654         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
15655         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
15656         (zero_extendhisi2_v850_set_flags): Likewise.
15657         (zero_extendqisi2_v850_set_flags): Likewise.
15658         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
15659         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
15660         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
15662         * config/v850/v850-protos.h (notice_update_cc): Remove.
15663         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
15664         (v850_print_operand): Handle 'D' and "d".
15665         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
15666         Add handling of arithmetic/logical operations compared against zero.
15667         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
15668         Do not look at v850_compare_op, instead get mode from last argument.
15669         (v850_gen_compare): Remove
15670         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
15671         after reload for prologue insns.
15672         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
15673         patterns.
15674         (construct_save_jarl): Likewise.
15675         (TARGET_FLAGS_REGNUM): Define.
15676         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
15677         (NOTICE_UPDATE_CC): Remove.
15678         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
15679         than cc0.  Conditionalize on reload_completed.
15680         (cmpsi_insn, setfcc_insn): Likewise.
15681         (tst1 splitter): Turn into define_and_split which sets the flags
15682         after reload.
15683         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
15684         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
15685         (cstoresf4, cstoredf4): Clobber the flags.
15686         (cmpsi, cmpsf, cmpdf): Remove expanders.
15687         (setf_insn): Remove pattern.
15688         (addsi3): Turn into define_and_split which clobbers the flags after
15689         reload and a suitable pattern (addsi3_clobber_flags) for use after
15690         reload.
15691         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
15692         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
15693         (ashrsi3, ashrsi3_v850e2): Likewise.
15694         (bins): Clobber the flags.
15695         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
15696         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
15697         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
15698         (call_value_internal_short, call_value_internal_long): Likewise.
15699         (callt_save_interrupt, callt_return_interrupt): Likewise.
15700         (save_interrupt, return_interrupt): Likewise.
15701         (callt_save_all_interrupt, save_all_interrupt): Likewise.
15702         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
15703         (restore_all_interrupt, _restore_all_interrupt): Likewise.
15704         (All FP comparisons): Only allow after reload has completed.
15705         (trfsr): Likewise.
15706         (divh, divhu): Tweak output template.
15707         (branch_z_normal, branch_z_invert): Remove
15708         (branch_nz_normal, branch_nz_invert): Likewise.
15709         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
15711         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
15712         * config/v850/v850.c (notice_update_cc): Remove.
15713         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
15714         (CC_NO_CARRY): Likewise.
15715         (NOTICE_UPDATE_CC): Define to nothing.
15716         * config/v850/v850.md: Remove block comment on cc0 handling
15717         Remove "cc" attribute from all patterns.  Remove cc_status handling
15718         from all patterns.  Minor formatting fixes.
15720 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15722         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
15723         (cortex-a76.cortex-a55): Likewise.
15724         * config/aarch64/aarch64-tune.md: Regenerate.
15725         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
15726         cortex-a76.cortex-a55.
15728 2018-06-27  Jeff Law  <law@redhat.com>
15730         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
15731         (MULTILIB_DIRNAMES): Similarly.
15733 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
15735         * gimple.h (gimple_return_retbnd): Delete.
15736         (gimple_return_set_retbnd): Likewise.
15737         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
15738         gimple_return_set_retbnd.
15739         * gimple-pretty-print.c (dump_gimple_return): Remove call to
15740         gimple_return_retbnd and adjust.
15741         * tree-inline.h (struct copy_body_data): Remove retbnd field.
15742         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
15743         Explicitly return NULL in a couple more cases.  Move assertion
15744         on debug statements and remove unreachable code.
15745         (reset_debug_binding): Do not test id->retbnd.
15746         (expand_call_inline): Do not set it.
15748 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
15750         * configure.ac: Add --disable-gcov option.
15751         * configure: Regenerate.
15752         * Makefile.in: Honour @enable_gcov@.
15753         * doc/install.texi: Document --disable-gcov.
15755 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15757         * config/arm/arm-cpus.in (cortex-a76): New entry.
15758         (cortex-a76.cortex-a55): Likewise.
15759         * config/arm/arm-tables.opt: Regenerate.
15760         * config/arm/arm-tune.md: Likewise.
15761         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
15762         * doc/invoke.texi (ARM Options): Document cortex-a76 and
15763         cortex-a76.cortex-a55.
15765 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
15767         PR target/85769
15768         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
15770 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
15772         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
15773         comment.
15774         (EPILOGUE_USES): Likewise.
15776 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
15778         * tree-inline.c (remap_location): New function extracted from...
15779         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
15780         (copy_phis_for_bb): ...here.  Call remap_location.
15781         (copy_cfg_body): Adjust call to copy_edges_for_bb.
15783 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
15785         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
15786         unaligned vsx for 16B memset.
15788 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
15790         PR target/86285
15791         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
15792         ieee128_float_type_node to long_double_type_node unless
15793         TARGET_LONG_DOUBLE_128 is set.
15795 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
15797         * cfgloop.c (get_loop_location): Convert return type from
15798         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
15799         by implicit construction from rtx_insn *, and using
15800         dump_user_location_t::from_function_decl for the fallback case.
15801         * cfgloop.h (get_loop_location): Convert return type from
15802         location_t to dump_user_location_t.
15803         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
15804         dump_printf_loc to pass in a dump_location_t rather than a
15805         location_t, via the gimple stmt.
15806         * coverage.c (get_coverage_counts): Update calls to
15807         dump_printf_loc to pass in dump_location_t rather than a
15808         location_t.
15809         * doc/optinfo.texi (Dump types): Convert example of
15810         dump_printf_loc from taking "locus" to taking "insn".  Update
15811         description of the "_loc" calls to cover dump_location_t.
15812         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
15813         "selftest.h".
15814         (dump_user_location_t::dump_user_location_t): New constructors,
15815         from gimple *stmt and rtx_insn *.
15816         (dump_user_location_t::from_function_decl): New function.
15817         (dump_loc): Make static.
15818         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
15819         const dump_location_t &.
15820         (dump_generic_expr_loc): Delete.
15821         (dump_printf_loc): Convert param "loc" from location_t to
15822         const dump_location_t &.
15823         (selftest::test_impl_location): New function.
15824         (selftest::dumpfile_c_tests): New function.
15825         * dumpfile.h: Include "profile-count.h".
15826         (class dump_user_location_t): New class.
15827         (struct dump_impl_location_t): New struct.
15828         (class dump_location_t): New class.
15829         (dump_printf_loc): Convert 2nd param from source_location to
15830         const dump_location_t &.
15831         (dump_generic_expr_loc): Delete.
15832         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
15833         const dump_location_t &.
15834         * gimple-fold.c (fold_gimple_assign): Update call to
15835         dump_printf_loc to pass in a dump_location_t rather than a
15836         location_t, via the gimple stmt.
15837         (gimple_fold_call): Likewise.
15838         * gimple-loop-interchange.cc
15839         (loop_cand::analyze_iloop_reduction_var): Update for change to
15840         check_reduction_path.
15841         (tree_loop_interchange::interchange): Update for change to
15842         find_loop_location.
15843         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
15844         change in return-type of find_loop_location.
15845         (graphite_regenerate_ast_isl): Likewise.
15846         * graphite-optimize-isl.c (optimize_isl): Likewise.
15847         * graphite.c (graphite_transform_loops): Likewise.
15848         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
15849         pass in a dump_location_t rather than a location_t, via the
15850         gimple stmt.
15851         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15852         * ipa.c (walk_polymorphic_call_targets): Likewise.
15853         * loop-unroll.c (report_unroll): Convert "locus" param from
15854         location_t to dump_location_t.
15855         (decide_unrolling): Update for change to get_loop_location's
15856         return type.
15857         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
15858         location_t to dump_user_location_t.
15859         (grid_find_single_omp_among_assignments_1): Updates calls to
15860         dump_printf_loc to pass in a dump_location_t rather than a
15861         location_t, via the gimple stmt.
15862         (grid_parallel_clauses_gridifiable): Convert "tloc" from
15863         location_t to dump_location_t.  Updates calls to dump_printf_loc
15864         to pass in a dump_location_t rather than a location_t, via the
15865         gimple stmt.
15866         (grid_inner_loop_gridifiable_p): Likewise.
15867         (grid_dist_follows_simple_pattern): Likewise.
15868         (grid_gfor_follows_tiling_pattern): Likewise.
15869         (grid_target_follows_gridifiable_pattern): Likewise.
15870         (grid_attempt_target_gridification): Convert initialization
15871         of local "grid" from memset to zero-initialization; FIXME: does
15872         this require C++11?  Update call to dump_printf_loc to pass in a
15873         optinfo_location rather than a location_t, via the gimple stmt.
15874         * profile.c (read_profile_edge_counts): Updates call to
15875         dump_printf_loc to pass in a dump_location_t rather than a
15876         location_t
15877         (compute_branch_probabilities): Likewise.
15878         * selftest-run-tests.c (selftest::run_tests): Call
15879         dumpfile_c_tests.
15880         * selftest.h (dumpfile_c_tests): New decl.
15881         * tree-loop-distribution.c (pass_loop_distribution::execute):
15882         Update for change in return type of find_loop_location.
15883         * tree-parloops.c (parallelize_loops): Likewise.
15884         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
15885         "locus" from location_t to dump_user_location_t.
15886         (canonicalize_loop_induction_variables): Likewise.
15887         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
15888         for change in return type of find_loop_location.
15889         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
15890         to dump_printf_loc to pass in a dump_location_t rather than a
15891         location_t, via the stmt.
15892         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
15893         Likewise.
15894         * tree-vect-loop-manip.c (find_loop_location): Convert return
15895         type from source_location to dump_user_location_t.
15896         (vect_do_peeling): Update for above change.
15897         (vect_loop_versioning): Update for change in type of
15898         vect_location.
15899         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
15900         from location_t to dump_user_location_t.
15901         (vect_estimate_min_profitable_iters): Update for change in type
15902         of vect_location.
15903         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
15904         location_t to dump_location_t.
15905         (vect_slp_bb): Update for change in type of vect_location.
15906         * tree-vectorizer.c (vect_location): Convert from source_location
15907         to dump_user_location_t.
15908         (try_vectorize_loop_1): Update for change in vect_location's type.
15909         (vectorize_loops): Likewise.
15910         (increase_alignment): Likewise.
15911         * tree-vectorizer.h (vect_location): Convert from source_location
15912         to dump_user_location_t.
15913         (find_loop_location): Convert return type from source_location to
15914         dump_user_location_t.
15915         (check_reduction_path): Convert 1st param from location_t to
15916         dump_user_location_t.
15917         * value-prof.c (check_counter): Update call to dump_printf_loc to
15918         pass in a dump_user_location_t rather than a location_t; update
15919         call to error_at for change in type of "locus".
15920         (check_ic_target): Update call to dump_printf_loc to
15921         pass in a dump_user_location_t rather than a location_t, via the
15922         call_stmt.
15924 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
15926         * config/s390/s390.h (enum processor_flags): Do not use
15927         default tune parameter when -march was specified.
15929 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
15931         PR target/86314
15932         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
15933         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
15934         operands.
15936 2018-06-26  Richard Biener  <rguenther@suse.de>
15938         PR tree-optimization/86287
15939         PR bootstrap/86316
15940         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
15941         (vect_analyze_loop): Initialize n_stmts.
15943 2018-06-26  Richard Biener  <rguenther@suse.de>
15945         PR middle-end/86271
15946         * fold-const.c (fold_convertible_p): Pointer extension
15947         isn't valid.
15949 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
15951         PR debug/86064
15952         * dwarf2out.c (loc_list_has_views): Adjust comments.
15953         (dw_loc_list): Split single cross-partition range with
15954         nonzero locview.
15956 2018-06-25  Jeff Law  <law@redhat.com>
15958         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
15959         on -mbig-switch by default.
15961         * config/v850/predicates.md (const_float_1_operand): Fix match_code
15962         test.
15963         (const_float_0_operand): Remove unused predicate.
15964         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
15965         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
15966         (recipsf2): New expander.  Original pattern now called
15967         (recipsf2_insn).
15968         (recipdf2, recipdf2_insn): Similarly.
15969         (rsqrtsf2, rsqrtsf2_insn): Similarly
15970         (rsqrtdf2, rsqrtdf2_insn): Similarly
15972 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
15974         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
15975         Simplify logic for FreeBSD (twice).
15977 2018-06-25  Martin Sebor  <msebor@redhat.com>
15979         PR tree-optimization/86204
15980         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
15981         a strnlen result if it's less than the length of the string.
15983 2018-06-25  Martin Sebor  <msebor@redhat.com>
15985         PR tree-optimization/85700
15986         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
15987         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
15988         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
15990 2018-06-25  Martin Sebor  <msebor@redhat.com>
15992         * doc/extend.texi (Zero-length arrays): Update and clarify.
15994 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
15996         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
15997         added IEEE/IBM long double multilib support on PowerPC little
15998         endian Linux systems.
15999         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
16000         (MULTILIB_DEFAULTS): Likewise.
16001         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16002         Likewise.
16003         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
16004         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
16005         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
16007 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
16009         PR middle-end/86311
16010         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
16011         (REORDER_45): Likewise.
16013 2018-06-25  Jeff Law  <law@redhat.com>
16015         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
16016         dividend to 32 bits.  Adjust length.
16017         (udivmodhi4): Cleanup output template.  Fix length.
16019 2018-06-25  Carl Love  <cel@us.ibm.com>
16021         * config/rs6000/vsx.md: Change word selector to prefered location.
16023 2018-06-25  Richard Biener  <rguenther@suse.de>
16025         PR tree-optimization/86304
16026         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
16027         epilogue-if-converted loops as well.
16029 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
16031         * lto-section-out.c (lto_begin_section): Do not print section
16032         name for noaddr and unnumbered dumps.
16034 2018-06-25  Richard Biener  <rguenther@suse.de>
16036         * tree-vectorizer.h (struct vec_info_shared): New structure
16037         with parts split out from struct vec_info and loop_nest from
16038         struct _loop_vec_info.
16039         (struct vec_info): Adjust accordingly.
16040         (struct _loop_vec_info): Likewise.
16041         (LOOP_VINFO_LOOP_NEST): Adjust.
16042         (LOOP_VINFO_DATAREFS): Likewise.
16043         (LOOP_VINFO_DDRS): Likewise.
16044         (struct _bb_vec_info): Likewise.
16045         (BB_VINFO_DATAREFS): Likewise.
16046         (BB_VINFO_DDRS): Likewise.
16047         (struct _stmt_vec_info): Add dr_aux member.
16048         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
16049         (DR_MISALIGNMENT_UNINITIALIZED): New.
16050         (set_dr_misalignment): Adjust.
16051         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
16052         (vect_analyze_loop): Adjust prototype.
16053         (vect_analyze_loop_form): Likewise.
16054         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
16055         Compute dependences lazily.
16056         (vect_record_base_alignments): Use shared datarefs/ddrs.
16057         (vect_verify_datarefs_alignment): Likewise.
16058         (vect_analyze_data_refs_alignment): Likewise.
16059         (vect_analyze_data_ref_accesses): Likewise.
16060         (vect_analyze_data_refs): Likewise.
16061         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
16062         constructor parameter for shared part.
16063         (vect_analyze_loop_form): Pass in shared part and adjust.
16064         (vect_analyze_loop_2): Pass in storage for the number of
16065         stmts.  Move loop nest finding to the caller.  Compute
16066         datarefs lazily.
16067         (vect_analyze_loop): Pass in shared part.
16068         (vect_transform_loop): Verify shared datarefs are unchanged.
16069         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
16070         constructor parameter for shared part.
16071         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
16072         (vect_slp_bb): Verify shared datarefs are unchanged before
16073         transform.
16074         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
16075         change.
16076         (new_stmt_vec_info): Initialize DR_AUX misalignment to
16077         DR_MISALIGNMENT_UNINITIALIZED.
16078         * tree-vectorizer.c (vec_info::vec_info): Add constructor
16079         parameter for shared part.
16080         (vec_info::~vec_info): Adjust.
16081         (vec_info_shared::vec_info_shared): New.
16082         (vec_info_shared::~vec_info_shared): Likewise.
16083         (vec_info_shared::save_datarefs): Likewise.
16084         (vec_info_shared::check_datarefs): Likewise.
16085         (try_vectorize_loop_1): Construct shared part live for analyses
16086         of a single loop for multiple vector sizes.
16087         * tree-parloops.c (gather_scalar_reductions): Adjust.
16089 2018-06-25  Richard Biener  <rguenther@suse.de>
16091         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
16092         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
16093         (vect_analyze_data_refs): Remove similar code from here and
16094         simplify accordingly.
16096 2018-06-25  Richard Biener  <rguenther@suse.de>
16098         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
16099         for reverse storage order accesses rather than asserting
16100         they cannot happen here.
16102 2018-06-25  Tom de Vries  <tdevries@suse.de>
16104         PR debug/86257
16105         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
16106         Use data16 instead of .byte for insn prefix.
16108 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
16110         PR C++/86082
16111         * parser.c (make_char_string_pack): Pass this literal chars
16112         through cpp_interpret_string.
16113         (cp_parser_userdef_numeric_literal): Check the result of
16114         make_char_string_pack.
16116 2018-06-24  Maya Rashish  <coypu@sdf.org>
16118         * ginclude/stddef.h: Simplify conditions around avoiding
16119         re-definition of __size_t.
16121 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
16123         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
16124         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
16126 2018-06-22  Maya Rashish  <coypu@sdf.org>
16128         * doc/invoke.texi (mno-fancy-math-387): Update for changes
16129         made to OpenBSD and NetBSD through the years.
16131 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16133         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
16134         behavior of vec_pack (vector double, vector double) to match
16135         behavior of vec_float2 (vector double, vector double).
16137 2018-06-22  Olivier Hainque  <hainque@adacore.com>
16139         * gimplify.c (gimplify_function_tree): Prevent creation
16140         of a trampoline for the address of the current function
16141         passed to entry/exit instrumentation hooks.
16143 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
16145         PR target/86222
16146         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
16147         correctly.
16149 2018-06-22  Martin Liska  <mliska@suse.cz>
16151         PR tree-optimization/86263
16152         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
16153         Bail out if is_enabled is false.
16154         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
16155         New declaration.
16156         (jump_table_cluster::is_enabled): New function.
16158 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
16160         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
16161         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
16162         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
16163         (lto_input_ts_binfo_tree_pointers): Likewise.
16164         * tree-streamer-out.c (streamer_write_tree_bitfields,
16165         write_ts_binfo_tree_pointers): Likewise.
16166         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
16168 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
16170         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
16172 2018-06-22  Martin Liska  <mliska@suse.cz>
16174         * symbol-summary.h (get): Make it pure and inline move
16175         functionality from ::get function.
16176         (get): Remove and inline into ::get and ::get_create.
16177         (get_create): Move code from ::get function.
16179 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16181         PR target/85994
16182         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
16183         -x assembler-with-cpp.
16185 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16187         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
16188         _FILE_OFFSET_BITS=64 for C++.
16190 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
16192         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
16193         conversion insn that shows up when pr85657-3.c is compiled using
16194         IEEE 128-bit long double.
16195         (neg<mode>2_internal): Use the correct mode to check whether the
16196         mode is IBM extended.
16197         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
16198         multiply and divide external functions from being created more
16199         than once.
16201 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
16203         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
16204         functions.
16205         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
16206         the edge can be forwarded.
16207         (cfg_layout_merge_blocks): Likewise.
16209 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
16211         * except.c (finish_eh_generation): Commit edge insertions only after
16212         the EH edges have been redirected from post-landing to landing pads.
16214 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
16216         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
16217         create_tmp_var_for to create the FRAME decl.
16218         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
16220 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
16222         * tree-inline.c (copy_edges_for_bb): Minor tweak.
16223         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
16224         debug statement when resetting its value.
16225         (expand_call_inline): Copy the locus of the call onto the assignment
16226         of the return value, if any.  Use local variable in more cases.
16228 2018-06-21  Martin Liska  <mliska@suse.cz>
16230         * ipa-pure-const.c (propagate_nothrow): Use
16231         funct_state_summaries->get.
16232         (dump_malloc_lattice): Likewise.
16233         (propagate_malloc): Likewise.
16235 2018-06-21  Richard Biener  <rguenther@suse.de>
16237         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
16238         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
16239         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
16240         comment.
16241         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
16242         BLOCK_ABSTRACT_ORIGIN unconditionally.
16244 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
16246         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
16247         deleting it.
16248         * ipa-reference.c (ipa_reference_c_finalize): Delete
16249         ipa_ref_opt_sum_summaries and set it to NULL.
16251 2018-06-21  Tom de Vries  <tdevries@suse.de>
16253         PR tree-optimization/85859
16254         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
16255         test with comment from bb_no_side_effects_p.
16257 2018-06-21  Richard Biener  <rguenther@suse.de>
16259         PR tree-optimization/86232
16260         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
16261         max for constant niter.
16263 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16265         * config/aarch64/aarch64-simd.md
16266         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
16268 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
16270         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
16271         Make opernads of the unspec commutative.
16273 2018-06-21  Richard Biener  <rguenther@suse.de>
16275         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
16276         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
16277         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
16278         (vect_analyze_data_ref_dependence): Re-order checks to deal with
16279         NULL DR_STEP.
16280         (vect_record_base_alignments): Do not record base alignment
16281         for gathers or scatters.
16282         (vect_compute_data_ref_alignment): Drop return value that is always
16283         true.  Bail out early for gathers or scatters.
16284         (vect_enhance_data_refs_alignment): Bail out early for gathers
16285         or scatters.
16286         (vect_find_same_alignment_drs): Likewise.
16287         (vect_analyze_data_refs_alignment): Remove dead code.
16288         (vect_slp_analyze_and_verify_node_alignment): Likewise.
16289         (vect_analyze_data_refs): For possible gathers or scatters do
16290         not create an alternate DR, just check their possible validity
16291         and mark them.  Adjust DECL_NONALIASED handling to not rely
16292         on DR_BASE_ADDRESS.
16293         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
16294         update inits of gathers or scatters.
16295         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
16296         Also copy gather/scatter flag to pattern vinfo.
16298 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16300         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
16301         behavior of vec_packsu (vector unsigned long long, vector unsigned
16302         long long) to match behavior of vec_packs with same signature.
16304 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
16305             Thomas Schwinge <thomas@codesourcery.com>
16306             Cesar Philippidis  <cesar@codesourcery.com>
16308         * gimplify.c (gimplify_scan_omp_clauses): Add support for
16309         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
16310         (gimplify_adjust_omp_clauses): Likewise.
16311         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
16312         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
16313         (gimplify_omp_target_update): Update handling of acc update and
16314         enter/exit data.
16315         * omp-low.c (install_var_field): Remove unused parameter
16316         base_pointers_restrict.
16317         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
16318         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
16319         FINALIZE}
16320         (omp_target_base_pointers_restrict_p): Delete.
16321         (scan_omp_target): Update call to scan_sharing_clauses.
16322         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
16323         FINALIZE}.
16324         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
16325         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
16326         (convert_local_omp_clauses): Likewise.
16327         * tree-pretty-print.c (dump_omp_clause): Likewise.
16328         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
16329         FINALIZE}.
16330         (omp_clause_code_name): Likewise.
16332 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
16334         PR debug/86194
16335         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
16336         be narrowed.
16338         PR tree-optimization/86231
16339         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
16340         anti-range don't overwrite *vr0min before using it to compute *vr0max.
16342 2018-06-20  Tom de Vries  <tdevries@suse.de>
16344         PR tree-optimization/86097
16345         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
16346         iv type if signedness of iv type is not the same as that of *nit.
16348 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
16350         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
16351         EDGE_EH edges, verify they are all EDGE_EH.
16353 2018-06-20  Maya Rashish  <coypu@sdf.org>
16355         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
16357 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16359         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
16360         * config/aarch64/aarch64.c (xgene1_tunings): Add
16361         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
16362         (aarch64_mode_valid_for_sched_fusion_p):
16363         Allow 16-byte modes.
16364         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
16365         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
16366         128-bit modes.
16367         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
16368         New pattern.
16369         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
16370         * config/aarch64/iterators.md (VQ2): New mode iterator.
16372 2018-06-20  Martin Liska  <mliska@suse.cz>
16374         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
16375         Change default ratio from 10 to 8.
16377 2018-06-20  Martin Liska  <mliska@suse.cz>
16379         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
16380         New.
16381         (bit_test_cluster::find_bit_tests): Likewise.
16382         (switch_decision_tree::analyze_switch_statement): Find clusters.
16383         * tree-switch-conversion.h (struct jump_table_cluster): Document
16384         hierarchy.
16386 2018-06-20  Martin Liska  <mliska@suse.cz>
16388         * tree-switch-conversion.c (switch_conversion::collect):
16389         Record m_uniq property.
16390         (switch_conversion::expand): Bail out for special conditions.
16391         (group_cluster::~group_cluster): New.
16392         (group_cluster::group_cluster): Likewise.
16393         (group_cluster::dump): Likewise.
16394         (jump_table_cluster::emit): New.
16395         (switch_decision_tree::fix_phi_operands_for_edges): New.
16396         (struct case_node): Remove struct.
16397         (jump_table_cluster::can_be_handled): New.
16398         (case_values_threshold): Moved to header.
16399         (reset_out_edges_aux): Likewise.
16400         (jump_table_cluster::is_beneficial): New.
16401         (bit_test_cluster::can_be_handled): Likewise.
16402         (add_case_node): Remove.
16403         (bit_test_cluster::is_beneficial): New.
16404         (case_bit_test::cmp): New.
16405         (bit_test_cluster::emit): New.
16406         (expand_switch_as_decision_tree_p): Remove.
16407         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
16408         (fix_phi_operands_for_edge): Likewise.
16409         (switch_decision_tree::analyze_switch_statement): New.
16410         (compute_cases_per_edge): Move ...
16411         (switch_decision_tree::compute_cases_per_edge): ... here.
16412         (try_switch_expansion): Likewise.
16413         (switch_decision_tree::try_switch_expansion): Likewise.
16414         (record_phi_operand_mapping): Likewise.
16415         (switch_decision_tree::record_phi_operand_mapping): Likewise.
16416         (emit_case_decision_tree): Likewise.
16417         (switch_decision_tree::emit): Likewise.
16418         (balance_case_nodes): Likewise.
16419         (switch_decision_tree::balance_case_nodes): Likewise.
16420         (dump_case_nodes): Likewise.
16421         (switch_decision_tree::dump_case_nodes): Likewise.
16422         (emit_jump): Likewise.
16423         (switch_decision_tree::emit_jump): Likewise.
16424         (emit_cmp_and_jump_insns): Likewise.
16425         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
16426         (emit_case_nodes): Likewise.
16427         (switch_decision_tree::emit_case_nodes): Likewise.
16428         (conditional_probability): Remove.
16429         * tree-switch-conversion.h (enum cluster_type): New.
16430         (PRINT_CASE): New.
16431         (struct cluster): Likewise.
16432         (cluster::cluster): Likewise.
16433         (struct simple_cluster): Likewise.
16434         (simple_cluster::simple_cluster): Likewise.
16435         (struct group_cluster): Likewise.
16436         (struct jump_table_cluster): Likewise.
16437         (struct bit_test_cluster): Likewise.
16438         (struct min_cluster_item): Likewise.
16439         (struct case_tree_node): Likewise.
16440         (case_tree_node::case_tree_node): Likewise.
16441         (jump_table_cluster::case_values_threshold): Likewise.
16442         (struct case_bit_test): Likewise.
16443         (struct switch_decision_tree): Likewise.
16444         (struct switch_conversion): Likewise.
16445         (switch_decision_tree::reset_out_edges_aux): Likewise.
16447 2018-06-20  Martin Liska  <mliska@suse.cz>
16449         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
16450         (hoist_edge_and_branch_if_true): Likewise.
16451         (expand_switch_using_bit_tests_p): Likewise.
16452         (struct case_bit_test): Likewise.
16453         (case_bit_test_cmp): Likewise.
16454         (emit_case_bit_tests): Likewise.
16455         (switch_conversion::switch_conversion): New class.
16456         (struct switch_conv_info): Remove old struct.
16457         (collect_switch_conv_info): More to ...
16458         (switch_conversion::collect): ... this.
16459         (check_range): Likewise.
16460         (switch_conversion::check_range): Likewise.
16461         (check_all_empty_except_final): Likewise.
16462         (switch_conversion::check_all_empty_except_final): Likewise.
16463         (check_final_bb): Likewise.
16464         (switch_conversion::check_final_bb): Likewise.
16465         (create_temp_arrays): Likewise.
16466         (switch_conversion::create_temp_arrays): Likewise.
16467         (free_temp_arrays): Likewise.
16468         (gather_default_values): Likewise.
16469         (switch_conversion::gather_default_values): Likewise.
16470         (build_constructors): Likewise.
16471         (switch_conversion::build_constructors): Likewise.
16472         (constructor_contains_same_values_p): Likewise.
16473         (switch_conversion::contains_same_values_p): Likewise.
16474         (array_value_type): Likewise.
16475         (switch_conversion::array_value_type): Likewise.
16476         (build_one_array): Likewise.
16477         (switch_conversion::build_one_array): Likewise.
16478         (build_arrays): Likewise.
16479         (switch_conversion::build_arrays): Likewise.
16480         (gen_def_assigns): Likewise.
16481         (switch_conversion::gen_def_assigns): Likewise.
16482         (prune_bbs): Likewise.
16483         (switch_conversion::prune_bbs): Likewise.
16484         (fix_phi_nodes): Likewise.
16485         (switch_conversion::fix_phi_nodes): Likewise.
16486         (gen_inbound_check): Likewise.
16487         (switch_conversion::gen_inbound_check): Likewise.
16488         (process_switch): Use the newly created class.
16489         (switch_conversion::expand): New.
16490         (switch_conversion::~switch_conversion): New.
16491         * tree-switch-conversion.h: New file.
16493 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16495         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
16496         tree-vect-patterns.c.
16497         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
16498         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
16499         (vect_recog_sad_pattern): Likewise.
16500         (vect_recog_widen_sum_pattern): Likewise.
16501         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
16502         (vect_recog_widen_shift_pattern): Remove the type_in argument.
16503         (vect_recog_rotate_pattern): Likewise.
16504         (vect_recog_mult_pattern): Likewise.
16505         (vect_recog_vector_vector_shift_pattern): Likewise.
16506         (vect_recog_divmod_pattern): Likewise.
16507         (vect_recog_mixed_size_cond_pattern): Likewise.
16508         (vect_recog_bool_pattern): Likewise.
16509         (vect_recog_mask_conversion_pattern): Likewise.
16510         (vect_try_gather_scatter_pattern): Likewise.
16511         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
16512         (vect_recog_over_widening_pattern): Likewise.
16513         (vect_recog_gather_scatter_pattern): Likewise.
16514         (vect_recog_func_ptr): Move from tree-vectorizer.h
16515         (vect_vect_recog_func_ptrs): Move further down the file.
16516         (vect_recog_func): Likewise.  Remove the third argument.
16517         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
16518         (vect_pattern_recog_1): Expect the pattern function to do any
16519         necessary target tests.  Also expect it to provide a vector type.
16520         Remove the type_in handling.
16522 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16524         * tree-vect-patterns.c (vect_pattern_detected): New function.
16525         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
16526         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
16527         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
16528         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
16529         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
16530         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
16531         (vect_recog_mask_conversion_pattern)
16532         (vect_try_gather_scatter_pattern): Likewise.
16534 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16536         * tree-vect-patterns.c (vect_get_internal_def): New function.
16537         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
16538         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
16539         (search_type_for_mask_1): Use it.
16541 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16543         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
16544         redundant WIDEN_SUM_EXPR handling.
16545         (vect_recog_sad_pattern): Likewise.
16547 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16549         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
16550         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
16551         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
16552         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
16553         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
16555 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16557         * tree-vect-stmts.c (vectorizable_call): Make sure that we
16558         use the stmt_vec_info of the original bb statement for the
16559         new zero assignment, even if the call is part of a pattern.
16561 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16563         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
16564         that the sequence is attached to the original statement rather
16565         than the pattern statement.
16566         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
16567         PATTERN_DEF_SEQ from the original statement rather than
16568         the main pattern statement.
16569         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
16570         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
16571         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
16573 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
16575         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
16576         definition statements before the early exit for statements that aren't
16577         live or relevant.
16578         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
16579         split out from...
16580         (vect_transform_loop): ...here.  Process pattern definition
16581         statements without first checking whether the main pattern
16582         statement is live or relevant.
16584 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
16586         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
16587         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
16589 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
16591         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
16592         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
16593         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
16594         (expand_block_compare): Change select_block_compare_mode call.
16595         (expand_strncmp_align_check): Use new functions, fix comment.
16596         (emit_final_str_compare_gpr): New function.
16597         (expand_strn_compare): Refactor and clean up code.
16598         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
16600 2018-06-19  Tony Reix  <tony.reix@atos.com>
16601             Damien Bergamini  <damien.bergamini@atos.com>
16602             David Edelsohn  <dje.gcc@gmail.com>
16604         * collect2.c (static_obj): New variable.
16605         (static_libs): New variable.
16606         (is_in_list): Uncomment declaration.
16607         (main): Track AIX libraries linked statically.
16608         (is_in_list): Uncomment definition.
16609         (scan_prog_file): Don't add AIX shared libraries initializer
16610         to constructor list if linking statically.
16612 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
16614         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
16615         constant.
16616         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
16618 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
16620         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
16621         blocks.
16623 2018-06-19  Martin Liska  <mliska@suse.cz>
16625         * config/i386/i386.c (ix86_can_inline_p): Do not use
16626         ipa_fn_summaries::get_create.
16627         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
16628         get.
16629         (devirtualization_time_bonus): Likewise.
16630         (ipcp_propagate_stage): Likewise.
16631         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16632         (edge_set_predicate): Likewise.
16633         (evaluate_conditions_for_known_args): Likewise.
16634         (evaluate_properties_for_edge): Likewise.
16635         (ipa_call_summary::reset): Tranform to ...
16636         (ipa_call_summary::~ipa_call_summary): ... this.
16637         (ipa_fn_summary::reset): Transform to ...
16638         (ipa_fn_summary::~ipa_fn_summary): ... this.
16639         (ipa_fn_summary_t::remove): Rename to ...
16640         (ipa_fn_summary_t::remove_callees): ... this.
16641         (ipa_fn_summary_t::duplicate): Use placement new
16642         instead of memory copy.
16643         (ipa_call_summary_t::duplicate): Likewise.
16644         (ipa_call_summary_t::remove): Remove.
16645         (dump_ipa_call_summary): Change get_create to get.
16646         (ipa_dump_fn_summary): Dump only when summary exists.
16647         (analyze_function_body): Use symbol_summary::get instead
16648         of get_create.
16649         (compute_fn_summary): Likewise.
16650         (estimate_edge_devirt_benefit): Likewise.
16651         (estimate_edge_size_and_time): Likewise.
16652         (inline_update_callee_summaries): Likewise.
16653         (remap_edge_change_prob): Likewise.
16654         (remap_edge_summaries): Likewise.
16655         (ipa_merge_fn_summary_after_inlining): Likewise.
16656         (write_ipa_call_summary): Likewise.
16657         (ipa_fn_summary_write): Likewise.
16658         (ipa_free_fn_summary): Likewise.
16659         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
16660         (struct ipa_call_summary): Likewise.
16661         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
16662         of get_create.
16663         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
16664         (estimate_size_after_inlining): Likewise.
16665         (estimate_growth): Likewise.
16666         (growth_likely_positive): Likewise.
16667         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16668         (inline_call): Likewise.
16669         * ipa-inline.c (caller_growth_limits): Likewise.
16670         (can_inline_edge_p): Likewise.
16671         (can_inline_edge_by_limits_p): Likewise.
16672         (compute_uninlined_call_time): Likewise.
16673         (compute_inlined_call_time): Likewise.
16674         (want_inline_small_function_p): Likewise.
16675         (edge_badness): Likewise.
16676         (update_caller_keys): Likewise.
16677         (update_callee_keys): Likewise.
16678         (inline_small_functions): Likewise.
16679         (inline_to_all_callers_1): Likewise.
16680         (dump_overall_stats): Likewise.
16681         (early_inline_small_functions): Likewise.
16682         (early_inliner): Likewise.
16683         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16684         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16685         * ipa-pure-const.c (malloc_candidate_p): Likewise.
16686         * ipa-split.c (execute_split_functions): Likewise.
16687         * symbol-summary.h: Likewise.
16688         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16690 2018-06-19  Richard Biener  <rguenther@suse.de>
16692         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
16693         (vectorize_loops): ... here.  Fix dbgcnt handling.
16694         (try_vectorize_loop): Wrap try_vectorize_loop_1.
16696 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
16698         PR target/86197
16699         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
16700         ieee128 argument takes up only one (vector) register, not two (floating
16701         point) registers.
16703 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
16705         * gimplify.c (gimplify_init_constructor): Really never clear for an
16706         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
16708 2018-06-19  Richard Biener  <rguenther@suse.de>
16710         PR tree-optimization/86179
16711         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
16712         after failed recognition.
16714 2018-06-18  Martin Sebor  <msebor@redhat.com>
16716         PR middle-end/85602
16717         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
16718         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
16719         Handle integer subtraction.
16720         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
16721         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
16723 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
16725         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
16726         param from rtx to rtx_insn *.
16727         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
16728         param.
16729         (frv_ifcvt_modify_insn): Likwise.
16730         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
16731         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
16732         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
16733         as_a <rtx_insn *> cast to local "unprotected_region" once
16734         it's been established that it's not NULL or pc_rtx.
16735         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
16736         param "sethi" from rtx to rtx_insn *.
16737         (nds32_group_float_insns): Likewise for param "insn".
16738         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
16739         param.
16740         (vax_output_int_subtract): Likewise.
16741         * config/vax/vax.c (vax_output_int_add): Likewise for param
16742         "insn".
16743         (vax_output_int_subtract): Likewise.
16744         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
16745         (emit_pattern_after): Likewise for param "after".
16746         (emit_insn_after): Likewise.
16747         (emit_jump_insn_after): Likewise.
16748         (emit_call_insn_after): Likewise.
16749         (emit_debug_insn_after): Likewise.
16750         (emit_pattern_before): Likewise for param "before".
16751         (emit_insn_before): Likewise.
16752         (emit_jump_insn_before): Likewise.
16753         * final.c (get_insn_template): Likewise for param "insn", removing
16754         a cast.
16755         * output.h (get_insn_template): Likewise for 2nd param.
16756         * rtl.h (emit_insn_before): Likewise.
16757         (emit_jump_insn_before): Likewise.
16758         (emit_debug_insn_before_noloc): Likewise.
16759         (emit_insn_after): Likewise.
16760         (emit_jump_insn_after): Likewise.
16761         (emit_call_insn_after): Likewise.
16762         (emit_debug_insn_after): Likewise.
16763         (set_insn_deleted): Likewise for param.
16765 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
16767         PR target/85358
16768         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
16769         floating point modes, so that IFmode is numerically greater than
16770         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
16771         to declare the ordering.  This prevents IFmode from being
16772         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
16773         machine.  Include rs6000-modes.h to share the fractional values
16774         between genmodes* and the rest of the compiler.
16775         (IFmode): Likewise.
16776         (KFmode): Likewise.
16777         (TFmode): Likewise.
16778         * config/rs6000/rs6000-modes.h: New file.
16779         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
16780         meaning of rs6000_long_double_size so that 126..128 selects an
16781         appropriate 128-bit floating point type.
16782         (rs6000_option_override_internal): Likewise.
16783         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
16784         (TARGET_LONG_DOUBLE_128): Change the meaning of
16785         rs6000_long_double_size so that 126..128 selects an appropriate
16786         128-bit floating point type.
16787         (LONG_DOUBLE_TYPE_SIZE): Update comment.
16788         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
16789         source and destination to match the standard usage.
16790         (truncifkf2): Likewise.
16791         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
16792         ISA 2.07 to use an explicit clobber, instead of passing in a
16793         temporary.
16794         (copysign<mode>3_soft): Likewise.
16796 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
16798         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
16799         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
16800         (vect_slp_analyze_instance_dependence): Likewise.
16801         (vect_enhance_data_refs_alignment): Likewise.
16802         (vect_analyze_data_refs_alignment): Likewise.
16803         (vect_slp_analyze_and_verify_instance_alignment
16804         (vect_analyze_data_ref_accesses): Likewise.
16805         (vect_prune_runtime_alias_test_list): Likewise.
16806         (vect_analyze_data_refs): Likewise.
16807         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
16808         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
16809         (vect_analyze_scalar_cycles_1): Likewise.
16810         (vect_get_loop_niters): Likewise.
16811         (vect_analyze_loop_form_1): Likewise.
16812         (vect_update_vf_for_slp): Likewise.
16813         (vect_analyze_loop_operations): Likewise.
16814         (vect_analyze_loop): Likewise.
16815         (vectorizable_induction): Likewise.
16816         (vect_transform_loop): Likewise.
16817         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
16818         * tree-vect-slp.c (vect_analyze_slp): Likewise.
16819         (vect_make_slp_decision): Likewise.
16820         (vect_detect_hybrid_slp): Likewise.
16821         (vect_slp_analyze_operations): Likewise.
16822         (vect_slp_bb): Likewise.
16823         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
16824         (vectorizable_bswap): Likewise.
16825         (vectorizable_call): Likewise.
16826         (vectorizable_simd_clone_call): Likewise.
16827         (vectorizable_conversion): Likewise.
16828         (vectorizable_assignment): Likewise.
16829         (vectorizable_shift): Likewise.
16830         (vectorizable_operation): Likewise.
16831         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
16833 2018-06-18  Martin Sebor  <msebor@redhat.com>
16835         PR tree-optimization/81384
16836         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
16837         * builtins.c (expand_builtin_strnlen): New function.
16838         (expand_builtin): Call it.
16839         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
16840         * builtins.def (BUILT_IN_STRNLEN): New.
16841         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
16842         Warn for bounds in excess of maximum object size.
16843         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
16844         single-value ranges.  Handle strnlen.
16845         (handle_builtin_strlen): Handle strnlen.
16846         (strlen_check_and_optimize_stmt): Same.
16847         * doc/extend.texi (Other Builtins): Document strnlen.
16849 2018-06-18  Maya Rashish  <coypu@sdf.org>
16851         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
16852         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
16853         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
16855         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
16856         here to ...
16857         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
16859 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
16861         * tree.c (escaped_string::escape): Replace cast to char * by
16862         const_cast<char *> (unescaped).
16864 2018-06-18  Nick Clifton  <nickc@redhat.com>
16866         PR 84195
16867         * tree.c (escaped_string): New class.  Converts an unescaped
16868         string into its escaped equivalent.
16869         (warn_deprecated_use): Use the new class to convert the
16870         deprecation message, if present.
16871         (test_escaped_strings): New self test.
16872         (test_c_tests): Add test_escaped_strings.
16873         * doc/extend.texi (deprecated): Add a note that the
16874         deprecation message is affected by the -fmessage-length
16875         option, and that control characters will be escaped.
16876         (#pragma GCC error): Document this pragma.
16877         (#pragma GCC warning): Likewise.
16878         * doc/invoke.texi (-fmessage-length): Document this option's
16879         effect on the #warning and #error preprocessor directives and
16880         the deprecated attribute.
16882 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
16884         * tree.c (decl_value_expr_lookup): Revert latest change.
16885         (decl_value_expr_insert): Likewise.
16887 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
16889         * gimplify.c (nonlocal_vlas): Delete.
16890         (nonlocal_vla_vars): Likewise.
16891         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
16892         referenced VLAs.
16893         (gimplify_body): Do not create and destroy nonlocal_vlas.
16894         * tree-nested.c: Include diagnostic.h.
16895         (use_pointer_in_frame): Tweak.
16896         (lookup_field_for_decl): Add assertion and declare the transformation.
16897         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
16898         internal error when the reference is in a wrong context.  Do not
16899         create a debug decl by default.
16900         (note_nonlocal_block_vlas): Delete.
16901         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
16902         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
16903         create a debug decl by default.
16904         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
16905         call is in a wrong context.
16906         (fixup_vla_decls): New function.
16907         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
16908         debug variables were created.
16909         * tree.c (decl_value_expr_lookup): Add checking assertion.
16910         (decl_value_expr_insert): Likewise.
16912 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
16914         PR middle-end/82479
16915         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
16916         * tree-scalar-evolution.c (interpret_expr): Likewise.
16917         (expression_expensive_p): Likewise.
16918         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
16919         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
16920         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
16921         (ssa_defined_by_minus_one_stmt_p): New.
16923 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
16925         PR middle-end/64946
16926         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
16927         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
16928         * dojump.c (do_jump): Likewise.
16929         * expr.c (expand_expr_real_2): Check operand type's sign.
16930         * fold-const.c (const_unop): Handle ABSU_EXPR.
16931         (fold_abs_const): Likewise.
16932         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
16933         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
16934         (strip_sign_op_1): Likesise.
16935         * match.pd: Add new pattern to generate ABSU_EXPR.
16936         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
16937         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
16938         * tree-eh.c (operation_could_trap_helper_p): Likewise.
16939         * tree-inline.c (estimate_operator_cost): Likewise.
16940         * tree-pretty-print.c (dump_generic_node): Likewise.
16941         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
16942         * tree.def (ABSU_EXPR): New.
16944 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
16946         PR middle-end/86095
16947         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
16948         documented as preserved for backward compatibility only.
16949         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
16951         PR rtl-optimization/86108
16952         * bb-reorder.c (create_forwarder_block): Renamed to ...
16953         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
16954         jump from new landing pad to the second part.
16955         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
16956         Adjust callers.
16958 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
16960         PR middle-end/85878
16961         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
16962         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
16963         Only call store_expr for halves if the mode is the same.
16965         PR middle-end/86123
16966         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
16967         Fix up comment formatting.
16969 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
16971         * typed-splay-tree.h (typed_splay_tree::remove): New function.
16972         (typed_splay_tree::closure,
16973         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
16974         (typed_splay_tree::typed_splay_tree,
16975         typed_splay_tree::operator =): Declared private.
16976         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
16977         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
16978         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
16979         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
16980         typed_splay_tree::splay_tree_splay,
16981         typed_splay_tree::splay_tree_foreach_helper,
16982         typed_splay_tree::splay_tree_insert,
16983         typed_splay_tree::splay_tree_remove,
16984         typed_splay_tree::splay_tree_lookup,
16985         typed_splay_tree::splay_tree_predecessor,
16986         typed_splay_tree::splay_tree_successor,
16987         typed_splay_tree::splay_tree_min,
16988         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
16989         (typed_splay_tree::root, typed_splay_tree::comp,
16990         typed_splay_tree::delete_key,
16991         typed_splay_tree::delete_value): New data members.
16992         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
16993         typed_splay_tree::remove.
16995 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
16997         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
16998         -mginv and -mno-ginv to the assembler.
16999         * config/mips/mips.opt (-mcrc): New option.
17000         (-mginv): Likewise.
17001         * doc/invoke.text (-mcrc): Document.
17002         (-mginv): Likewise.
17004 2018-06-15  Nick Clifton  <nickc@redhat.com>
17006         PR 84195
17007         * tree.c (escaped_string): New class.  Converts an unescaped
17008         string into its escaped equivalent.
17009         (warn_deprecated_use): Use the new class to convert the
17010         deprecation message, if present.
17011         (test_escaped_strings): New self test.
17012         (test_c_tests): Add test_escaped_strings.
17013         * doc/extend.texi (deprecated): Add a note that the
17014         deprecation message is affected by the -fmessage-length
17015         option, and that control characters will be escaped.
17016         (#pragma GCC error): Document this pragma.
17017         (#pragma GCC warning): Likewise.
17018         * doc/invoke.texi (-fmessage-length): Document this option's
17019         effect on the #warning and #error preprocessor directives and
17020         the deprecated attribute.
17022 2018-06-15  Richard Biener  <rguenther@suse.de>
17024         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
17025         here, also noting vector size used.
17026         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
17027         size used in MSG_OPTIMIZED_LOCATIONS dump.
17028         (pass_slp_vectorize::execute): Adjust.
17030 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
17032         PR target/85968
17033         * config/arc/arc.c (arc_return_address_register): Fix
17034         if-condition.
17036 2018-06-15  Richard Biener  <rguenther@suse.de>
17038         PR middle-end/86159
17039         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
17040         leave useless conversion stripping to force_gimple_operand_gsi.
17041         (gimplify_build2): Likewise.
17042         (gimplify_build1): Likewise.
17044 2018-06-15  Richard Biener  <rguenther@suse.de>
17046         PR middle-end/86076
17047         * tree-cfg.c (move_stmt_op): unshare invariant addresses
17048         before adjusting their block.
17050 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
17052         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
17053         multilibs for *-*-rtems*.
17054         * config/riscv/t-rtems: New file.
17056 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
17058         PR middle-end/86122
17059         * match.pd ((A +- CST1) +- CST2): Punt if last resort
17060         unsigned_type_for returns NULL.
17062         PR target/85945
17063         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
17064         subregs of multi-word pseudos unless the float mode has word size.
17066 2018-06-14  Richard Biener  <rguenther@suse.de>
17068         PR middle-end/86139
17069         * tree-vect-generic.c (build_word_mode_vector_type): Remove
17070         duplicate and harmful type_hash_canon.
17071         * tree.c (type_hash_canon): Assert we didn't find ourselves.
17073 2018-06-14  Richard Biener  <rguenther@suse.de>
17075         PR ipa/86124
17076         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
17077         NULL cgraph_node.
17079 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
17081         * config/rtems.h (STDINT_LONG32): Define.
17083 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
17084             Prachi Godbole  <prachi.godbole@imgtec.com>
17086         * config/mips/mips-cpus.def: Define P6600.
17087         * config/mips/mips-tables.opt: Regenerate.
17088         * config/mips/mips.c (mips_ucbranch_type): New enum.
17089         (mips_rtx_cost_data): Add support for P6600.
17090         (mips_issue_rate): Likewise.
17091         (mips_multipass_dfa_lookahead): Likewise.
17092         (mips_avoid_hazard): Likewise.
17093         (mips_reorg_process_insns): Likewise.
17094         (mips_classify_branch_p6600): New function.
17095         * config/mips/mips.h (TUNE_P6600): New define.
17096         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
17097         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
17098         * config/mips/mips.md: Include p6600.md.
17099         (processor): Add p6600.
17100         * config/mips/p6600.md: New file.
17101         * doc/invoke.texi: Add p6600 to supported architectures.
17103 2018-06-13  Martin Sebor  <msebor@redhat.com>
17105         PR tree-optimization/86114
17106         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
17107         of integer types.
17108         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
17110 2018-06-13  Richard Biener  <rguenther@suse.de>
17112         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
17113         Properly set vector type of the intermediate stmt.
17114         * tree-vect-stmts.c (vectorizable_operation): The destination
17115         var always has vectype_out type.
17117 2018-06-13  Jeff Law  <law@redhat.com>
17119         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
17120         integer 0 for argument to print_rtl_with_bb.
17121         (rl78_reorg): Likewise.
17123 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
17125         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
17126         from rtx to rtx_insn *.
17127         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
17128         "label".
17129         (add_sched_insns_for_speculation): Likewise for local "target",
17130         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
17131         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
17132         from rtx_insn ** to rtx_code_label **.
17133         (reorg_emit_nops): Likewise.
17134         (c6x_reorg): Likewise for local "call_labels".
17135         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
17136         rtx to rtx_insn *.
17137         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
17138         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
17139         the loops over LABEL_REFS.
17140         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
17141         braf_label.
17142         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
17143         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
17144         (split_branches): Strengthen local "olabel" from rtx to
17145         rtx_insn *, adding a safe_as_a cast.
17146         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
17147         to "rtx_insn *".
17148         (add_insn_after): Likewise for first two params.
17149         (add_insn_before): Likewise.
17150         (remove_insn): Likewise for param.
17151         (emit_pattern_before_noloc): Likewise for second and third params.
17152         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
17153         (emit_call_insn_before_noloc): Likewise.
17154         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
17155         to "rtx_insn *".
17156         (emit_barrier_before): Likewise.
17157         (emit_label_before): Strengthen "label" param from "rtx" to
17158         "rtx_code_label *".  Strengthen "before" param from "rtx" to
17159         "rtx_insn *".
17160         (emit_insn_after_1): Strengthen "after" param from "rtx" to
17161         "rtx_insn *".
17162         (emit_pattern_after_noloc): Likewise.
17163         (emit_insn_after_noloc): Likewise.
17164         (emit_jump_insn_after_noloc): Likewise.
17165         (emit_call_insn_after_noloc): Likewise.
17166         (emit_debug_insn_after_noloc): Likewise.
17167         (emit_barrier_after): Likewise.
17168         (emit_label_after): Likewise for both params.
17169         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
17170         "loc" param from "int" to "location_t".
17171         (emit_insn_after_setloc): Likewise.
17172         (emit_jump_insn_after_setloc): Likewise.
17173         (emit_call_insn_after_setloc): Likewise.
17174         (emit_debug_insn_after_setloc): Likewise.
17175         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
17176         "loc" param from "int" to "location_t".
17177         (emit_pattern_before): Convert NULL_RTX to NULL.
17178         (emit_insn_before_setloc): Convert "loc" param from "int" to
17179         "location_t".
17180         (emit_jump_insn_before_setloc): Likewise.
17181         (emit_call_insn_before_setloc): Likewise.
17182         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
17183         rtx_insn *.  Convert "loc" param from "int" to "location_t".
17184         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
17185         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
17186         Convert 3rd param from "int" to "location_t".
17187         (emit_barrier_before, emit_barrier_after, next_real_insn):
17188         Strengthen param from rtx to rtx_insn *.
17189         (emit_label_before): Strengthen 1st param from "rtx" to
17190         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
17191         "rtx_insn *".
17192         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
17193         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
17194         Strengthen 2nd param from "rtx" to "rtx_insn *".
17195         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
17196         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
17197         Likewise. Convert 3rd param from "int" to "location_t".
17198         (emit_label_after): Strengthen 1st param from "rtx" to
17199         "rtx_code_label *".
17200         (next_real_insn, remove_insn): Strengthen param from "rtx" to
17201         "rtx_insn *".
17202         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
17203         from "rtx" to "rtx_insn *".
17205 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
17207         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
17208         bodies streamed in with -Q.
17209         * dumpfile.c (dump_files): Add lto-stream-out dump file.
17210         * dumpfile.h (tree_dump_index): Add lto_stream_out.
17211         * gimple-streamer-out.c: Include gimple-pretty-print.h
17212         (output_bb): Dump stmts streamed.
17213         * lto-section-out.c: Include print-tree.h
17214         (lto_begin_section): Dump sections created.
17215         (lto_output_decl_index): Dump decl encoded.
17216         * lto-streamer-out.c: Include print-tree.h
17217         (create_output_block): Dump output block created.
17218         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
17219         (output_function): Dump function output.
17220         (output_constructor): Dump constructor streamed.
17221         (write_global_stream): Output indexes encoded.
17222         (produce_asm_for_decls): Dump streams encoded.
17223         * lto-streamer.c (streamer_dump_file): New global var.
17224         * lto-streamer.h (streamer_dump_file): Declare.
17225         * passes.c (ipa_write_summaries): Initialize streamer dump.
17226         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
17227         in.
17229 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
17231         PR target/86048
17232         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
17233         offsets for register save directives.  Emit a second batch of save
17234         directives, if need be, when the function accesses prior frames.
17236 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
17238         * config/arc/fpu.md (fmasf4): Force operand to register.
17239         (fnmasf4): Likewise.
17241 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
17243         * config/arc/arc-protos.h (arc_pad_return): Remove.
17244         * config/arc/arc.c (machine_function): Remove force_short_suffix
17245         and size_reason.
17246         (arc_print_operand): Adjust printing of '&'.
17247         (arc_verify_short): Remove conditional printing of short suffix.
17248         (arc_final_prescan_insn): Remove reference to size_reason.
17249         (pad_return): New function.
17250         (arc_reorg): Call pad_return.
17251         (arc_pad_return): Remove.
17252         (arc_init_machine_status): Remove reference to force_short_suffix.
17253         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
17254         (attr length): When attribute iscompact is true force to 2
17255         regardless; in the case of maybe check if we want to force the
17256         instruction to have 4 bytes length.
17257         (nopv): Change it to generate 4 byte long nop as well.
17258         (blockage): New pattern.
17259         (simple_return): Remove call to arc_pad_return.
17260         (p_return_i): Likewise.
17262 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
17264         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
17266 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
17268         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
17269         ARC cores.
17271 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
17273         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
17274         for ARC700 and ARCv2.
17276 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
17278         PR target/86076
17279         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
17280         operands[2] instead of operands[1].
17283 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
17285         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
17286         case, check whether the outer register overlaps an unallocatable
17287         register, not just whether it fits the required class.
17289 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
17291         * poly-int.h (can_div_trunc_p): Add new overload in which all values
17292         are poly_ints.
17293         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
17294         (memrefs_conflict_p): Likewise.
17295         (init_alias_analysis): Likewise.
17296         * cfgexpand.c (expand_debug_expr): Likewise.
17297         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
17298         * cse.c (fold_rtx): Likewise.
17299         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
17300         * expr.c (emit_block_move_hints): Likewise.
17301         (clear_storage_hints, push_block, emit_push_insn): Likewise.
17302         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
17303         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
17304         (emit_group_store): Likewise.
17305         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
17306         to read the PRE/POST_MODIFY increment.
17307         * calls.c (store_one_arg): Use strip_offset.
17308         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
17309         poly_int_rtx_p.
17310         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
17311         by a VEC_SELECT.
17312         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
17313         (simplify_binary_operation_1): Extend CONST_INT handling to
17314         poly_int_rtx_p.
17315         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
17316         than a HOST_WIDE_INT.
17317         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
17318         poly_int64.
17319         (adjust_mems, add_stores): Update accodingly.
17320         (vt_canonicalize_addr): Track polynomial offsets.
17321         (emit_note_insn_var_location): Likewise.
17322         (vt_add_function_parameter): Likewise.
17323         (vt_initialize): Likewise.
17325 2018-06-12  Jeff Law  <law@redhat.com>
17327         * config.gcc (alpha*-*-freebsd*): Remove.
17328         * config/alpha/freebsd.h: Remove.
17330 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
17332         PR other/69968
17333         * spellcheck-tree.c (levenshtein_distance): Rename to...
17334         (get_edit_distance): ...this, and update for underlying renaming.
17335         * spellcheck-tree.h (levenshtein_distance): Rename to...
17336         (get_edit_distance): ...this.
17337         * spellcheck.c (levenshtein_distance): Rename to...
17338         (get_edit_distance): ...this.  Convert from Levenshtein distance
17339         to Damerau-Levenshtein distance by supporting transpositions of
17340         adjacent characters.  Rename "v1" to "v_next" and "v0" to
17341         "v_one_ago".
17342         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
17343         (selftest::test_edit_distance_unit_test_oneway): ...this, and
17344         update for underlying renaming.
17345         (selftest::levenshtein_distance_unit_test): Rename to...
17346         (selftest::test_get_edit_distance_unit): ...this, and update for
17347         underlying renaming.
17348         (selftest::test_find_closest_string): Add example from PR 69968
17349         where transposition helps
17350         (selftest::test_metric_conditions): Update for renaming.
17351         (selftest::test_metric_conditions): Likewise.
17352         (selftest::spellcheck_c_tests): Likewise.
17353         * spellcheck.h (levenshtein_distance): Rename both overloads to...
17354         (get_edit_distance): ...this.
17355         (best_match::consider): Update for renaming.
17357 2018-06-12  Martin Sebor  <msebor@redhat.com>
17359         PR tree-optimization/85259
17360         * builtins.c (compute_objsize): Handle constant offsets.
17361         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
17362         true iff a warning has been issued.
17363         * gimple.h (gimple_nonartificial_location): New function.
17364         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
17365         gimple_nonartificial_location and handle -Wno-system-headers.
17366         (handle_builtin_stxncpy): Same.
17368 2018-06-12  Martin Sebor  <msebor@redhat.com>
17370         PR c/85931
17371         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
17373 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
17375         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
17376         BUILTIN_VEC_XST entries for pointer to double and long long.
17378 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
17380         PR target/85990
17381         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
17382         Update comments.
17383         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
17384         Likewise.
17386 2018-06-12  Martin Liska  <mliska@suse.cz>
17388         * doc/options.texi: Document IntegerRange.
17390 2018-06-12  Martin Liska  <mliska@suse.cz>
17392         * config/i386/i386.opt: Make MPX-related options as Deprecated.
17393         * opt-functions.awk: Handle Deprecated flag.
17394         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
17395         and report error.
17396         (read_cmdline_option): Report warning for a deprecated option.
17397         * opts.h (struct cl_option): Add new field cl_deprecated.
17398         (CL_ERR_DEPRECATED): New.
17400 2018-06-12  Martin Liska  <mliska@suse.cz>
17402         * doc/options.texi: Document Deprecated option flag.
17404 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
17406         * config/arc/arc-arch.h (arc_extras): New enum.
17407         (arc_cpu_t):Add field extra.
17408         (arc_cpu_types): Consider the extras.
17409         * config/arc/arc-cpus.def: Add extras info.
17410         * config/arc/arc-opts.h (processor_type): Consider extra field.
17411         * config/arc/arc.c (arc_override_options): Handle extra field.
17413 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
17415         * config/arc/arc-arch.h: Update ARC_OPTX macro.
17416         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
17417         field.
17418         * config/arc/arc.c (arc_init): Update pic warning.
17419         (irq_range): Update irq range parsing warnings.
17420         (arc_override_options): Update various warning messages.
17421         (arc_handle_aux_attribute): Likewise.
17423 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
17425         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
17427 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17429         * doc/sourcebuild.texi: Document usage of line number 0 in verify
17430         compiler messages directives.
17432 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
17434         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
17435         * config/mips/mips-tables.opt: Regenerate.
17436         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
17437         mips64r6.
17438         * doc/invoke.texi: Document -march=i6500.
17440 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
17442         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
17443         (i6400_gpmul): Add cpu_unit.
17444         (i6400_gpdiv): Likewise.
17445         (i6400_msa_add_d): Update reservations.
17446         (i6400_msa_int_add) Likewise.
17447         (i6400_msa_short_logic3) Likewise.
17448         (i6400_msa_short_logic2) Likewise.
17449         (i6400_msa_short_logic) Likewise.
17450         (i6400_msa_move) Likewise.
17451         (i6400_msa_cmp) Likewise.
17452         (i6400_msa_short_float2) Likewise.
17453         (i6400_msa_div_d) Likewise.
17454         (i6400_msa_long_logic1) Likewise.
17455         (i6400_msa_long_logic2) Likewise.
17456         (i6400_msa_mult) Likewise.
17457         (i6400_msa_long_float2) Likewise.
17458         (i6400_msa_long_float4) Likewise.
17459         (i6400_msa_long_float5) Likewise.
17460         (i6400_msa_long_float8) Likewise.
17461         (i6400_fpu_fadd): Include frint type.
17462         (i6400_fpu_store): New define_insn_reservation.
17463         (i6400_fpu_load): Likewise.
17464         (i6400_fpu_move): Likewise.
17465         (i6400_fpu_fcmp): Likewise.
17466         (i6400_fpu_fmadd): Likewise.
17467         (i6400_int_mult): Include imul3nc type and update reservation.
17468         (i6400_int_div): Include idiv3 type and update reservation.
17469         (i6400_int_load): Update to check type not move_type.
17470         (i6400_int_store): Likewise.
17471         (i6400_int_prefetch): Set zero latency.
17473 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
17475         * gcc.c: Document new %@{...} sequence.
17476         (LINK_COMMAND_SPEC): Use it for the -L switches.
17477         (cpp_unique_options): Use it for the -I switches.
17478         (at_file_argbuf): New global variable.
17479         (in_at_file): Likewise.
17480         (alloc_args): Create at_file_argbuf.
17481         (clear_args): Truncate at_file_argbuf.
17482         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
17483         (open_at_file): New function.
17484         (close_at_file): Likewise.
17485         (create_at_file): Delete.
17486         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
17487         <'o'>: Likewise.
17488         <'@'>: New case.
17489         (validate_switches_from_spec): Deal with %@{...} sequence.
17490         (validate_switches): Likewise.
17491         (driver::finalize): Call clear_args.
17493 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
17495         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
17497 2018-06-11  Martin Sebor  <msebor@redhat.com>
17499         * doc/invoke.texi (-Wall): List -Wc++17-compat.
17500         (Wno-class-memaccess): Add @opindex.
17501         (Wno-templates, Wno-multiple-inheritance): Same.
17502         (Wno-virtual-inheritance, Wno-namespaces): Same.
17503         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
17504         (Wno-format-overflow, Wno-format-truncation): Same.
17505         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
17506         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
17507         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
17508         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
17509         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
17510         (Wno-misspelled-isr): Same.
17512 2018-06-11  Martin Sebor  <msebor@redhat.com>
17514         * PR tree-optimization/86083
17515         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
17517 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
17519         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
17521 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
17523         PR target/85755
17524         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
17525         on the correct operand.
17526         (*movdi_internal64): Ditto.
17528 2018-06-11  Martin Liska  <mliska@suse.cz>
17530         PR tree-optimization/86089
17531         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
17533 2018-06-11  Julia Koval  <julia.koval@intel.com>
17535         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
17536         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
17537         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
17539 2018-06-11  Olivier Hainque  <hainque@adacore.com>
17541         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
17542         for Ada with strict dwarf2.
17544 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
17546         PR target/85755
17547         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
17548         addresses.
17550 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
17552         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
17554 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
17556         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
17557         TARGET_ELF.
17559 2018-06-08  Martin Liska  <mliska@suse.cz>
17561         * tree-cfg.h (debug_function): Fix argument type to match
17562         implementation.
17564 2018-06-08  Martin Liska  <mliska@suse.cz>
17566         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
17567         Remove usage of MPX-related (and removed) fields.
17568         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
17570 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
17572         * cfg.c (debug): Use TDF_NONE rather than 0.
17573         * cfghooks.c (debug): Likewise.
17574         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
17575         (struct dump_option_value_info): Convert to...
17576         (struct kv_pair): ...this template type.
17577         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
17578         rather than 0.
17579         (optinfo_verbosity_options): Likewise.
17580         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
17581         OPTGROUP_NONE.
17582         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
17583         than int for "optgroup_flags" param.
17584         (dump_generic_expr_loc): Use dump_flags_t rather than int for
17585         "dump_kind" param.
17586         (dump_dec): Likewise.
17587         (dump_finish): Use TDF_NONE rather than 0.
17588         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
17589         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
17590         than 0.  Update for change to option_ptr.
17591         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
17592         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
17593         0.  Update for changes to optinfo_verbosity_options and
17594         optgroup_options.
17595         (opt_info_switch_p): Convert optgroup_flags from int to
17596         optgroup_flags_t.
17597         (dump_basic_block): Use dump_flags_t rather than int
17598         for "dump_kind" param.
17599         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
17600         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
17601         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
17602         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
17603         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
17604         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
17605         TDF_NONE): Convert from macros to...
17606         (enum dump_flag): ...this new enum.
17607         (dump_flags_t): Update to use enum.
17608         (operator|, operator&, operator~, operator|=, operator&=):
17609         Implement for dump_flags_t.
17610         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
17611         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
17612         Convert from macros to...
17613         (enum optgroup_flag): ...this new enum.
17614         (optgroup_flags_t): New typedef.
17615         (operator|, operator|=): Implement for optgroup_flags_t.
17616         (struct dump_file_info): Convert field "alt_flags" to
17617         dump_flags_t.  Convert field "optgroup_flags" to
17618         optgroup_flags_t.
17619         (dump_basic_block): Use dump_flags_t rather than int for param.
17620         (dump_generic_expr_loc): Likewise.
17621         (dump_dec): Likewise.
17622         (dump_register): Convert param "optgroup_flags" to
17623         optgroup_flags_t.
17624         (opt_info_enable_passes): Likewise.
17625         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
17626         than 0.
17627         * gimple-pretty-print.c (debug): Likewise.
17628         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
17629         (merged_store_group::apply_stores): Likewise.
17630         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
17631         * gimple.c (verify_gimple_pp): Likewise.
17632         * graphite-poly.c (print_pbb_body): Likewise.
17633         * passes.c (pass_manager::register_one_dump_file): Convert
17634         local "optgroup_flags" to optgroup_flags_t.
17635         * print-tree.c (print_node): Use TDF_NONE rather than 0.
17636         (debug): Likewise.
17637         (debug_body): Likewise.
17638         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
17639         to optgroup_flags_t.
17640         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
17641         than 0.
17642         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
17643         (convert_mult_to_fma): Likewise.
17644         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
17645         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
17646         * tree-vect-data-refs.c (dump_lower_bound): Convert param
17647         "dump_kind" to dump_flags_t.
17649 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
17651         * config/rs6000/rs6000.c (min, max): Delete.
17653 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
17655         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
17656         -mabi=spe and -mabi=no-spe.
17658 2018-06-08  Martin Liska  <mliska@suse.cz>
17660         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
17661         where we expect an existing summary.
17663 2018-06-08  Martin Liska  <mliska@suse.cz>
17665         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
17666         * ipa-inline.h (estimate_edge_growth): Likewise.
17668 2018-06-08  Martin Liska  <mliska@suse.cz>
17670         * cgraph.c (function_version_hasher::hash): Use
17671         cgraph_node::get_uid ().
17672         (function_version_hasher::equal):
17673         * cgraph.h (cgraph_node::get_uid): New method.
17674         * ipa-inline.c (update_caller_keys): Use
17675         cgraph_node::get_uid ().
17676         (update_callee_keys): Likewise.
17677         * ipa-utils.c (searchc): Likewise.
17678         (ipa_reduced_postorder): Likewise.
17679         * lto-cgraph.c (input_node): Likewise.
17680         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
17681         * symbol-summary.h (symtab_insertion): Likewise.
17682         (symtab_removal): Likewise.
17683         (symtab_duplication): Likewise.
17684         * tree-pretty-print.c (dump_function_header): Likewise.
17685         * tree-sra.c (convert_callers_for_node): Likewise.
17687 2018-06-08  Martin Liska  <mliska@suse.cz>
17689         * cgraph.c (symbol_table::create_edge): Always assign a new
17690         unique number.
17691         (symbol_table::free_edge): Do not recycle numbers.
17692         * cgraph.h (cgraph_edge::get): New method.
17693         * symbol-summary.h (symtab_removal): Use it.
17694         (symtab_duplication): Likewise.
17695         (call_summary::hashable_uid): Remove.
17697 2018-06-08  Martin Liska  <mliska@suse.cz>
17699         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
17700         (initialize_growth_caches): Remove.
17701         (free_growth_caches): Likewise.
17702         (do_estimate_edge_time): Use edge_growth_cache.
17703         (do_estimate_edge_size): Likewise.
17704         (do_estimate_edge_hints): Likewise.
17705         * ipa-inline.c (reset_edge_caches): Likewise.
17706         (recursive_inlining): Likewise.
17707         (inline_small_functions): Likewise.
17708         * ipa-inline.h (initialize_growth_caches): Remove.
17709         (estimate_edge_size): Likewise.
17710         (estimate_edge_time): Likewise.
17711         (estimate_edge_hints): Likewise.
17712         (reset_edge_growth_cache): Likewise.
17713         * symbol-summary.h (call_summary::remove): New method.
17715 2018-06-08  Martin Liska  <mliska@suse.cz>
17717         * ipa-cp.c (class edge_clone_summary): New summary.
17718         (grow_edge_clone_vectors): Remove.
17719         (ipcp_edge_duplication_hook): Remove.
17720         (class edge_clone_summary_t): New call_summary class.
17721         (ipcp_edge_removal_hook): Remove.
17722         (edge_clone_summary_t::duplicate): New function.
17723         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
17724         (create_specialized_node): Likewise.
17725         (ipcp_driver): Initialize edge_clone_summaries and do not
17726         register hooks.
17728 2018-06-08  Martin Liska  <mliska@suse.cz>
17730         * symbol-summary.h (get): New function.
17731         (call_summary::m_initialize_when_cloning): New class member.
17733 2018-06-08  Martin Liska  <mliska@suse.cz>
17735         * cgraph.c (cgraph_node::remove): Do not recycle uid.
17736         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
17737         (symbol_table::allocate_cgraph_symbol): Do not set uid.
17738         * passes.c (uid_hash_t): Record removed_nodes by their uids.
17739         (remove_cgraph_node_from_order): Use the removed_nodes set.
17740         (do_per_function_toporder): Likwise.
17741         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
17742         instead of summary_uid.
17743         (symtab_removal): Likewise.
17744         (symtab_duplication): Likewise.
17746 2018-06-08  Martin Liska  <mliska@suse.cz>
17748         * ipa-cp.c (ipcp_store_bits_results): Use
17749         ipcp_transformation_sum.
17750         (ipcp_store_vr_results): Likewise.
17751         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
17752         to ...
17753         (ipcp_transformation_initialize): ... this.
17754         (ipa_set_node_agg_value_chain):
17755         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
17756         (write_ipcp_transformation_info): Likewise.
17757         (read_ipcp_transformation_info): Likewise.
17758         (ipcp_update_bits): Likewise.
17759         (ipcp_update_vr): Likewise.
17760         (ipcp_transform_function): Likewise.
17761         * ipa-prop.h: Rename ipcp_transformation_summary to
17762         ipcp_transformation.
17763         (class ipcp_transformation_t): New function summary.
17764         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
17765         (ipa_get_agg_replacements_for_node): Likewise.
17767 2018-06-08  Martin Liska  <mliska@suse.cz>
17769         * ipa-pure-const.c (struct funct_state_d): Do it class instead
17770         of struct.
17771         (class funct_state_summary_t): New function_summary class.
17772         (has_function_state): Remove.
17773         (get_function_state): Likewise.
17774         (set_function_state): Likewise.
17775         (add_new_function): Likewise.
17776         (funct_state_summary_t::insert): New function.
17777         (duplicate_node_data): Remove.
17778         (remove_node_data): Remove.
17779         (funct_state_summary_t::duplicate): New function.
17780         (register_hooks): Create new funct_state_summaries.
17781         (pure_const_generate_summary): Use it.
17782         (pure_const_write_summary): Likewise.
17783         (pure_const_read_summary): Likewise.
17784         (propagate_pure_const): Likewise.
17785         (propagate_nothrow): Likewise.
17786         (dump_malloc_lattice): Likewise.
17787         (propagate_malloc): Likewise.
17788         (execute): Do not register hooks, just remove summary
17789         instead.
17790         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
17791         constructor.
17793 2018-06-08  Martin Liska  <mliska@suse.cz>
17795         * ipa-reference.c (remove_node_data): Remove.
17796         (duplicate_node_data): Likewise.
17797         (class ipa_ref_var_info_summary_t): New class.
17798         (class ipa_ref_opt_summary_t): Likewise.
17799         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
17800         (get_reference_optimization_summary): Use
17801         ipa_ref_opt_sum_summaries.
17802         (set_reference_vars_info): Remove.
17803         (set_reference_optimization_summary): Likewise.
17804         (ipa_init): Create summaries.
17805         (init_function_info): Use function summary.
17806         (ipa_ref_opt_summary_t::duplicate): New function.
17807         (ipa_ref_opt_summary_t::remove): New function.
17808         (get_read_write_all_from_node): Fix GNU coding style.
17809         (propagate): Use function summary.
17810         (write_node_summary_p): Fix GNU coding style.
17811         (stream_out_bitmap): Likewise.
17812         (ipa_reference_read_optimization_summary): Use function summary.
17813         (ipa_reference_c_finalize): Do not release hooks.
17815 2018-06-08  Martin Liska  <mliska@suse.cz>
17817         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
17818         (analyze_function_body): Extract multiple calls of get_create.
17819         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
17820         * ipa-inline.c (recursive_inlining): Use ::get method.
17821         * ipa-inline.h (estimate_edge_growth): Likewise.
17823 2018-06-08  Martin Liska  <mliska@suse.cz>
17825         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
17826         HSA_INVALID.
17827         (hsa_function_summary::hsa_function_summary): Use the new enum
17828         value.
17829         (hsa_gpu_implementation_p): Use hsa_summaries::get.
17830         * hsa-gen.c (hsa_get_host_function): Likewise.
17831         (get_brig_function_name): Likewise.
17832         * ipa-hsa.c (process_hsa_functions): Likewise.
17833         (ipa_hsa_write_summary): Likewise.
17834         * symbol-summary.h (symtab_duplication): Use ::get function/
17835         (get): New function.
17837 2018-06-08  Martin Liska  <mliska@suse.cz>
17839         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
17840         of get.
17841         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
17842         (hsa_register_kernel): Likewise.
17843         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
17844         * hsa-gen.c (hsa_get_host_function): Likewise.
17845         (get_brig_function_name): Likewise.
17846         (generate_hsa): Likewise.
17847         (pass_gen_hsail::execute): Likewise.
17848         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
17849         (devirtualization_time_bonus): Likewise.
17850         (ipcp_propagate_stage): Likewise.
17851         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
17852         (edge_set_predicate): Likewise.
17853         (evaluate_conditions_for_known_args): Likewise.
17854         (evaluate_properties_for_edge): Likewise.
17855         (ipa_fn_summary::reset): Likewise.
17856         (ipa_fn_summary_t::duplicate): Likewise.
17857         (dump_ipa_call_summary): Likewise.
17858         (ipa_dump_fn_summary): Likewise.
17859         (analyze_function_body): Likewise.
17860         (compute_fn_summary): Likewise.
17861         (estimate_edge_devirt_benefit): Likewise.
17862         (estimate_edge_size_and_time): Likewise.
17863         (estimate_calls_size_and_time): Likewise.
17864         (estimate_node_size_and_time): Likewise.
17865         (inline_update_callee_summaries): Likewise.
17866         (remap_edge_change_prob): Likewise.
17867         (remap_edge_summaries): Likewise.
17868         (ipa_merge_fn_summary_after_inlining): Likewise.
17869         (ipa_update_overall_fn_summary): Likewise.
17870         (read_ipa_call_summary): Likewise.
17871         (inline_read_section): Likewise.
17872         (write_ipa_call_summary): Likewise.
17873         (ipa_fn_summary_write): Likewise.
17874         (ipa_free_fn_summary): Likewise.
17875         * ipa-hsa.c (process_hsa_functions): Likewise.
17876         (ipa_hsa_write_summary): Likewise.
17877         (ipa_hsa_read_section): Likewise.
17878         * ipa-icf.c (sem_function::merge): Likewise.
17879         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
17880         (do_estimate_edge_time): Likewise.
17881         (estimate_size_after_inlining): Likewise.
17882         (estimate_growth): Likewise.
17883         (growth_likely_positive): Likewise.
17884         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
17885         (inline_call): Likewise.
17886         * ipa-inline.c (caller_growth_limits): Likewise.
17887         (can_inline_edge_p): Likewise.
17888         (can_inline_edge_by_limits_p): Likewise.
17889         (compute_uninlined_call_time): Likewise.
17890         (compute_inlined_call_time): Likewise.
17891         (want_inline_small_function_p): Likewise.
17892         (edge_badness): Likewise.
17893         (update_caller_keys): Likewise.
17894         (update_callee_keys): Likewise.
17895         (recursive_inlining): Likewise.
17896         (inline_small_functions): Likewise.
17897         (inline_to_all_callers_1): Likewise.
17898         (dump_overall_stats): Likewise.
17899         (early_inline_small_functions): Likewise.
17900         (early_inliner): Likewise.
17901         * ipa-inline.h (estimate_edge_growth): Likewise.
17902         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
17903         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
17904         * ipa-prop.h (IPA_NODE_REF): Likewise.
17905         (IPA_EDGE_REF): Likewise.
17906         * ipa-pure-const.c (malloc_candidate_p): Likewise.
17907         (propagate_malloc): Likewise.
17908         * ipa-split.c (execute_split_functions): Likewise.
17909         * symbol-summary.h: Rename get to get_create.
17910         (get): Likewise.
17911         (get_create): Likewise.
17912         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
17914 2018-06-08  Martin Liska  <mliska@suse.cz>
17916         * symbol-summary.h (release): Move definition out of class
17917         declaration.
17918         (symtab_removal): Likewise.
17919         (symtab_duplication): Likewise.
17921 2018-06-08  Martin Liska  <mliska@suse.cz>
17923         * symbol-summary.h (function_summary): Move constructor
17924         implementation out of class declaration.
17925         (release): Likewise.
17926         (symtab_insertion): Likewise.
17927         (symtab_removal): Likewise.
17928         (symtab_duplication): Likewise.
17929         (get): Likewise.
17931 2018-06-08  Martin Liska  <mliska@suse.cz>
17933         * Makefile.in: Remove support for MPX (macros, related functions,
17934         fields in cgraph_node, ...).
17935         * builtin-types.def (BT_BND): Likewise.
17936         (BT_FN_BND_CONST_PTR): Likewise.
17937         (BT_FN_CONST_PTR_BND): Likewise.
17938         (BT_FN_VOID_PTR_BND): Likewise.
17939         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
17940         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
17941         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
17942         (expand_builtin_mempcpy_with_bounds): Likewise.
17943         (expand_builtin_memset_with_bounds): Likewise.
17944         (expand_builtin_memset_args): Likewise.
17945         (std_expand_builtin_va_start): Likewise.
17946         (expand_builtin): Likewise.
17947         (expand_builtin_with_bounds): Likewise.
17948         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
17949         (DEF_LIB_BUILTIN_CHKP): Likewise.
17950         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
17951         (DEF_CHKP_BUILTIN): Likewise.
17952         (BUILT_IN_MEMCPY): Likewise.
17953         (BUILT_IN_MEMMOVE): Likewise.
17954         (BUILT_IN_MEMPCPY): Likewise.
17955         (BUILT_IN_MEMSET): Likewise.
17956         (BUILT_IN_STPCPY): Likewise.
17957         (BUILT_IN_STRCAT): Likewise.
17958         (BUILT_IN_STRCHR): Likewise.
17959         (BUILT_IN_STRCPY): Likewise.
17960         (BUILT_IN_STRLEN): Likewise.
17961         (BUILT_IN_MEMCPY_CHK): Likewise.
17962         (BUILT_IN_MEMMOVE_CHK): Likewise.
17963         (BUILT_IN_MEMPCPY_CHK): Likewise.
17964         (BUILT_IN_MEMSET_CHK): Likewise.
17965         (BUILT_IN_STPCPY_CHK): Likewise.
17966         (BUILT_IN_STRCAT_CHK): Likewise.
17967         (BUILT_IN_STRCPY_CHK): Likewise.
17968         * calls.c (store_bounds): Likewise.
17969         (emit_call_1): Likewise.
17970         (special_function_p): Likewise.
17971         (maybe_warn_nonstring_arg): Likewise.
17972         (initialize_argument_information): Likewise.
17973         (finalize_must_preallocate): Likewise.
17974         (compute_argument_addresses): Likewise.
17975         (expand_call): Likewise.
17976         * cfgexpand.c (expand_call_stmt): Likewise.
17977         (expand_return): Likewise.
17978         (expand_gimple_stmt_1): Likewise.
17979         (pass_expand::execute): Likewise.
17980         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
17981         (cgraph_node::remove): Likewise.
17982         (cgraph_node::dump): Likewise.
17983         (cgraph_node::verify_node): Likewise.
17984         * cgraph.h (chkp_function_instrumented_p): Likewise.
17985         (symtab_node::get_alias_target): Likewise.
17986         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
17987         (cgraph_local_p): Likewise.
17988         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
17989         (cgraph_edge::rebuild_references): Likewise.
17990         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
17991         (walk_polymorphic_call_targets): Likewise.
17992         (cgraph_node::expand_thunk): Likewise.
17993         (symbol_table::output_weakrefs): Likewise.
17994         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
17995         (ix86_handle_option): Likewise.
17996         * config/i386/constraints.md: Likewise.
17997         * config/i386/i386-builtin-types.def (BND): Likewise.
17998         (VOID): Likewise.
17999         (PVOID): Likewise.
18000         (ULONG): Likewise.
18001         * config/i386/i386-builtin.def (BDESC_END): Likewise.
18002         (BDESC_FIRST): Likewise.
18003         (BDESC): Likewise.
18004         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
18005         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
18006         * config/i386/i386.c (enum reg_class): Likewise.
18007         (ix86_target_string): Likewise.
18008         (ix86_option_override_internal): Likewise.
18009         (ix86_conditional_register_usage): Likewise.
18010         (ix86_valid_target_attribute_inner_p): Likewise.
18011         (ix86_set_indirect_branch_type): Likewise.
18012         (ix86_set_current_function): Likewise.
18013         (ix86_function_arg_regno_p): Likewise.
18014         (init_cumulative_args): Likewise.
18015         (ix86_function_arg_advance): Likewise.
18016         (ix86_function_arg): Likewise.
18017         (ix86_pass_by_reference): Likewise.
18018         (ix86_function_value_regno_p): Likewise.
18019         (ix86_function_value_1): Likewise.
18020         (ix86_function_value_bounds): Likewise.
18021         (ix86_return_in_memory): Likewise.
18022         (ix86_setup_incoming_vararg_bounds): Likewise.
18023         (ix86_va_start): Likewise.
18024         (indirect_thunk_need_prefix): Likewise.
18025         (print_reg): Likewise.
18026         (ix86_print_operand): Likewise.
18027         (ix86_expand_call): Likewise.
18028         (ix86_output_function_return): Likewise.
18029         (reg_encoded_number): Likewise.
18030         (BDESC_VERIFYS): Likewise.
18031         (ix86_init_mpx_builtins): Likewise.
18032         (ix86_init_builtins): Likewise.
18033         (ix86_emit_cmove): Likewise.
18034         (ix86_emit_move_max): Likewise.
18035         (ix86_expand_builtin): Likewise.
18036         (ix86_builtin_mpx_function): Likewise.
18037         (ix86_get_arg_address_for_bt): Likewise.
18038         (ix86_load_bounds): Likewise.
18039         (ix86_store_bounds): Likewise.
18040         (ix86_load_returned_bounds): Likewise.
18041         (ix86_store_returned_bounds): Likewise.
18042         (ix86_class_likely_spilled_p): Likewise.
18043         (ix86_hard_regno_mode_ok): Likewise.
18044         (x86_order_regs_for_local_alloc): Likewise.
18045         (ix86_mitigate_rop): Likewise.
18046         (ix86_bnd_prefixed_insn_p): Likewise.
18047         (ix86_mpx_bound_mode): Likewise.
18048         (ix86_make_bounds_constant): Likewise.
18049         (ix86_initialize_bounds): Likewise.
18050         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
18051         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
18052         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
18053         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
18054         (TARGET_CHKP_BOUND_MODE): Likewise.
18055         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
18056         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
18057         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
18058         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
18059         * config/i386/i386.h (TARGET_MPX): Likewise.
18060         (TARGET_MPX_P): Likewise.
18061         (VALID_BND_REG_MODE): Likewise.
18062         (FIRST_BND_REG): Likewise.
18063         (LAST_BND_REG): Likewise.
18064         (enum reg_class): Likewise.
18065         (BND_REG_P): Likewise.
18066         (BND_REGNO_P): Likewise.
18067         (BNDmode): Likewise.
18068         (ADJUST_INSN_LENGTH): Likewise.
18069         * config/i386/i386.md: Likewise.
18070         * config/i386/i386.opt: Likewise.
18071         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
18072         (defined): Likewise.
18073         (LINK_MPX): Likewise.
18074         (MPX_SPEC): Likewise.
18075         (LIBMPX_SPEC): Likewise.
18076         (LIBMPXWRAPPERS_SPEC): Likewise.
18077         (CHKP_SPEC): Likewise.
18078         * config/i386/predicates.md: Likewise.
18079         * dbxout.c (dbxout_type): Likewise.
18080         * doc/extend.texi: Likewise.
18081         * doc/invoke.texi: Likewise.
18082         * doc/md.texi: Likewise.
18083         * doc/tm.texi: Likewise.
18084         * doc/tm.texi.in: Likewise.
18085         * dwarf2out.c (is_base_type): Likewise.
18086         (gen_formal_types_die): Likewise.
18087         (gen_subprogram_die): Likewise.
18088         (gen_type_die_with_usage): Likewise.
18089         (gen_decl_die): Likewise.
18090         (dwarf2out_late_global_decl): Likewise.
18091         * expr.c (expand_assignment): Likewise.
18092         (emit_storent_insn): Likewise.
18093         (store_expr_with_bounds): Likewise.
18094         (store_expr): Likewise.
18095         (expand_expr_real_1): Likewise.
18096         * expr.h (store_expr_with_bounds): Likewise.
18097         * function.c (use_register_for_decl): Likewise.
18098         (struct bounds_parm_data): Likewise.
18099         (assign_parms_augmented_arg_list): Likewise.
18100         (assign_parm_find_entry_rtl): Likewise.
18101         (assign_parm_is_stack_parm): Likewise.
18102         (assign_parm_load_bounds): Likewise.
18103         (assign_bounds): Likewise.
18104         (assign_parms): Likewise.
18105         (expand_function_start): Likewise.
18106         * gcc.c (CHKP_SPEC): Likewise.
18107         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
18108         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
18109         (wrestrict_dom_walker::check_call): Likewise.
18110         * gimple.c (gimple_build_call_from_tree): Likewise.
18111         * gimple.h (enum gf_mask): Likewise.
18112         (gimple_call_with_bounds_p): Likewise.
18113         (gimple_call_set_with_bounds): Likewise.
18114         * gimplify.c (gimplify_init_constructor): Likewise.
18115         * ipa-cp.c (initialize_node_lattices): Likewise.
18116         (propagate_constants_across_call): Likewise.
18117         (find_more_scalar_values_for_callers_subset): Likewise.
18118         * ipa-hsa.c (process_hsa_functions): Likewise.
18119         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
18120         * ipa-icf.c (sem_function::merge): Likewise.
18121         * ipa-inline.c (early_inliner): Likewise.
18122         * ipa-pure-const.c (warn_function_noreturn): Likewise.
18123         (warn_function_cold): Likewise.
18124         (propagate_pure_const): Likewise.
18125         * ipa-ref.h (enum GTY): Likewise.
18126         * ipa-split.c (find_retbnd): Likewise.
18127         (consider_split): Likewise.
18128         (split_function): Likewise.
18129         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
18130         * ipa.c (walk_polymorphic_call_targets): Likewise.
18131         (symbol_table::remove_unreachable_nodes): Likewise.
18132         (process_references): Likewise.
18133         (cgraph_build_static_cdtor_1): Likewise.
18134         * lto-cgraph.c (lto_output_node): Likewise.
18135         (output_refs): Likewise.
18136         (compute_ltrans_boundary): Likewise.
18137         (input_overwrite_node): Likewise.
18138         (input_node): Likewise.
18139         (input_cgraph_1): Likewise.
18140         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
18141         * passes.c (pass_manager::execute_early_local_passes): Likewise.
18142         (class pass_chkp_instrumentation_passes): Likewise.
18143         (make_pass_chkp_instrumentation_passes): Likewise.
18144         * passes.def: Likewise.
18145         * rtl.h (struct GTY): Likewise.
18146         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
18147         * stor-layout.c (layout_type): Likewise.
18148         * symtab.c: Likewise.
18149         * target.def: Likewise.
18150         * targhooks.c (default_chkp_bound_type): Likewise.
18151         (default_chkp_bound_mode): Likewise.
18152         (default_builtin_chkp_function): Likewise.
18153         (default_chkp_function_value_bounds): Likewise.
18154         (default_chkp_make_bounds_constant): Likewise.
18155         (default_chkp_initialize_bounds): Likewise.
18156         * targhooks.h (default_chkp_bound_type): Likewise.
18157         (default_chkp_bound_mode): Likewise.
18158         (default_builtin_chkp_function): Likewise.
18159         (default_chkp_function_value_bounds): Likewise.
18160         (default_chkp_make_bounds_constant): Likewise.
18161         (default_chkp_initialize_bounds): Likewise.
18162         * toplev.c (compile_file): Likewise.
18163         (process_options): Likewise.
18164         * tree-core.h (DEF_BUILTIN): Likewise.
18165         (DEF_BUILTIN_CHKP): Likewise.
18166         * tree-inline.c (declare_return_variable): Likewise.
18167         (remap_gimple_stmt): Likewise.
18168         (copy_bb): Likewise.
18169         (initialize_inlined_parameters): Likewise.
18170         (expand_call_inline): Likewise.
18171         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
18172         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
18173         (make_pass_ipa_chkp_produce_thunks): Likewise.
18174         (make_pass_chkp): Likewise.
18175         (make_pass_chkp_opt): Likewise.
18176         (make_pass_chkp_instrumentation_passes): Likewise.
18177         * tree-pretty-print.c (dump_generic_node): Likewise.
18178         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
18179         * tree-ssa-dce.c (propagate_necessity): Likewise.
18180         (eliminate_unnecessary_stmts): Likewise.
18181         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
18182         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
18183         * tree-ssa-sccvn.h: Likewise.
18184         * tree-ssa-strlen.c (get_string_length): Likewise.
18185         (valid_builtin_call): Likewise.
18186         (adjust_last_stmt): Likewise.
18187         (handle_builtin_strchr): Likewise.
18188         (handle_builtin_strcpy): Likewise.
18189         (handle_builtin_stxncpy): Likewise.
18190         (handle_builtin_memcpy): Likewise.
18191         (handle_builtin_strcat): Likewise.
18192         (strlen_check_and_optimize_stmt): Likewise.
18193         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
18194         * tree-streamer-in.c: Likewise.
18195         * tree-streamer.c (record_common_node): Likewise.
18196         * tree.c (tree_code_size): Likewise.
18197         (wide_int_to_tree_1): Likewise.
18198         (type_contains_placeholder_1): Likewise.
18199         (build_common_tree_nodes): Likewise.
18200         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
18201         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
18202         (POINTER_BOUNDS_P): Likewise.
18203         (BOUNDED_TYPE_P): Likewise.
18204         (BOUNDED_P): Likewise.
18205         (CALL_WITH_BOUNDS_P): Likewise.
18206         (pointer_bounds_type_node): Likewise.
18207         * value-prof.c (gimple_ic): Likewise.
18208         * var-tracking.c (vt_add_function_parameters): Likewise.
18209         * varasm.c (make_decl_rtl): Likewise.
18210         (assemble_start_function): Likewise.
18211         (output_constant): Likewise.
18212         (maybe_assemble_visibility): Likewise.
18213         * varpool.c (ctor_for_folding): Likewise.
18214         * chkp-builtins.def: Remove.
18215         * ipa-chkp.c: Remove.
18216         * ipa-chkp.h: Remove.
18217         * rtl-chkp.c: Remove.
18218         * rtl-chkp.h: Remove.
18219         * tree-chkp-opt.c: Remove.
18220         * tree-chkp.c: Remove.
18221         * tree-chkp.h: Remove.
18223 2018-06-07  Carl Love  <cel@us.ibm.com>
18225         * config/rs6000/vsx.md (vextract_fp_from_shorth,
18226         vextract_fp_from_shortl): Add BE support.
18228 2018-06-07  Paul Koning  <ni1d@arrl.net>
18230         * compare-elim.c (try_merge_compare): Don't merge compare if
18231         address contains a side effect.
18232         (try_eliminate_compare): Likewise.
18234 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
18236         * config.gcc: Support "tremont".
18237         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
18238         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18239         PROCESSOR_TREMONT.
18240         * config/i386/i386.c (m_TREMONT): Define.
18241         (processor_target_table): Add "tremont".
18242         (PTA_TREMONT): Define.
18243         (ix86_lea_outperforms): Add TARGET_TREMONT.
18244         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
18245         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
18246         and M_INTEL_GOLDMONT_PLUS.
18247         (fold_builtin_cpu): Add "tremont".
18248         (ix86_add_stmt_cost): Add TARGET_TREMONT.
18249         (ix86_option_override_internal): Add "tremont".
18250         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
18251         (processor_type): Add PROCESSOR_TREMONT.
18252         * config/i386/x86-tune.def: Add m_TREMONT.
18253         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
18255 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18257         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
18258         symbol defined for msp430i* devices to be lower case.
18260 2018-06-07  Richard Biener  <rguenther@suse.de>
18262         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
18263         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
18264         Properly wrap signed arithmetic if overflow wraps.
18266 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
18268         PR tree-optimization/69615
18269         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
18270         of a cast from a same precision integral SSA_NAME in a bb dominated
18271         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
18272         cast to utype if rhs2 has already a compatible type.
18274 2018-06-07  Richard Biener  <rguenther@suse.de>
18276         PR tree-optimization/85935
18277         * graphite-scop-detection.c (find_params_in_bb): Analyze
18278         condition operands with respect to the correct loop.  Assert
18279         the analysis doesn't fail.
18281 2018-06-04  Carl Love  <cel@us.ibm.com>
18283         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
18284         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
18285         as it is slightly cheaper.
18286         (first_match_or_eos_index_<mode>):
18287         Calculate index using natural element order.
18288         (first_match_index_<mode>):
18289         Calculate index using natural element order.
18290         (first_match_or_eos_index_<mode>):
18291         Calculate index using natural order.
18292         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
18293         for BE and LE modes.
18294         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
18295         P9V_BUILTIN_VCLZLSBB_V16QI.
18296         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
18297         specific.
18299 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18301         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
18302         indentation and line wrap for many prototypes.  Add missing
18303         @smallexample directives around block of prototypes for vec_xl and
18304         vec_xst.
18306 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
18308         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
18309         track if we pass or return IEEE 128-bit floating point.
18310         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
18311         C++ mangling that is compatible with GCC 8.1.
18312         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
18313         (init_cumulative_args): Note if we pass or return IEEE 128-bit
18314         floating point types.
18315         (rs6000_function_arg_advance_1): Likewise.
18316         (rs6000_mangle_type): Optionally generate mangled names that match
18317         what GCC 8.1 generated for IEEE 128-bit floating point types.
18318         (rs6000_globalize_decl_name): If we have an external function that
18319         passes or returns IEEE 128-bit types, generate a weak reference
18320         from the mangled name used in GCC 8.1 to the current mangled
18321         name.
18322         (rs6000_init_builtins): Make __ibm128 use the long double type if
18323         long double is IBM extended double.  Make __float128 use the long
18324         double type if long double is IEEE 128-bit.
18326         PR target/85657
18327         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
18328         macro for __ibm128 built-in functions.
18329         (PACK_IF): Add __ibm128 pack/unpack functions.
18330         (UNPACK_IF): Likewise.
18331         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
18332         enable long double built-in functions if long double is IEEE
18333         128-bit floating point.
18334         (rs6000_invalid_builtin): Update long double built-in function
18335         error message.
18336         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
18337         functions, adjust the built-in function to use the long double
18338         built-in function if __ibm128 and long double are the same type.
18339         * doc/extend.texi (PowerPC builtins): Update documention for
18340         __builtin_{,un}pack_longdouble.  Add documentation for
18341         __builtin_{,un}pack_ibm128.
18343 2018-06-06  Jim Wilson  <jimw@sifive.com>
18345         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
18346         (struct machine_function): New field interrupt_mode.
18347         (riscv_handle_type_attribute): New function.  Add forward declaration.
18348         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
18349         (riscv_expand_epilogue): Check interrupt_mode field.
18350         (riscv_set_current_function): Check interrupt attribute args and
18351         set interrupt_mode field.
18352         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
18353         (riscv_sret, riscv_uret): New.
18354         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
18355         new arguments to interrupt attribute.
18357 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
18359         PR target/63177
18360         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
18361         Don't handle -mcpu=power8 if -mpower9-vector is also used.
18363 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18365         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
18366         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
18367         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
18368         several redundant entries.
18370 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
18372         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
18373         type from "rtx" to "rtx_insn *".
18374         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
18375         for local "call_insn", removing cast.
18376         (ix86_expand_call): Likewise, introducing a "call_insn" local.
18378 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
18380         PR tree-optimization/86066
18381         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
18382         for BIT_INSERT_EXPR stores.
18384 2018-06-06  Richard Biener  <rguenther@suse.de>
18386         PR tree-optimization/86062
18387         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
18388         component refs ontop
18389         of to be offsetted base.
18391 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18393         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
18394         to be static and remove check on interrupt attribute name.
18396 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18398         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
18399         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
18401 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
18403         PR target/79924
18404         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
18405         second argument.
18406         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
18407         Remove second argument, change how error is called.
18408         (aarch64_layout_arg): Remove second argument from
18409         aarch64_err_no_fpadvsimd call.
18410         (aarch64_init_cumulative_args): Ditto.
18411         (aarch64_gimplify_va_arg_expr): Ditto.
18412         * config/aarch64/aarch64.md (mov<mode>): Ditto.
18414 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
18416         * config/i386/i386.md (simple_return_indirect_internal): New expander.
18417         (*simple_return_indirect_internal<mode>): Rename from
18418         simple_return_indirect_internal.  Use W mode iterator.
18419         (rstorssp): New expander.
18420         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
18421         (clrssbsy): New expander.
18422         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
18424 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18426         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
18427         __typeof__.
18428         (cmse_check_pointed_object): Likewise.
18430 2018-06-05  Martin Liska  <mliska@suse.cz>
18432         PR gcov-profile/47618
18433         * doc/invoke.texi: Document how -fprofile-dir format
18434         is extended.
18436 2018-06-05  Richard Biener  <rguenther@suse.de>
18438         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
18439         removal pretend DOM info isn't available so we do not update
18440         it and only remove edges, not dominated blocks.  Actually free
18441         DOM info in case we removed something.  Remove unreachable blocks.
18442         (mfb_keep_latches): Work with either DOM info or marked backedges.
18443         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
18444         first.  Mark backedges if DOM info isn't available.
18445         (Re-)compute DOM info after cleanup_control_flow_pre.
18447 2018-06-05  Richard Biener  <rguenther@suse.de>
18449         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
18450         (locus_discrim_hasher::hash): Adjust.
18451         (locus_discrim_hasher::equal): Likewise.
18452         (next_discriminator_for_locus): Work on line directly.
18453         (same_line_p): Pass in expanded locus1 as well.
18454         (assign_discriminators): Avoid redundant location expansions.
18456 2018-06-05  Richard Biener  <rguenther@suse.de>
18458         PR tree-optimization/86046
18459         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
18460         if required after clearing TREE_ADDRESSABLE.
18462 2018-06-05  Richard Biener  <rguenther@suse.de>
18464         PR tree-optimization/86047
18465         * tree-ssa-loop.c (for_each_index): Glob handling of all
18466         decls and constants and really handle all of them.
18468 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18470         PR target/81497
18471         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
18472         qualifier_void_pointer and qualifier_const_void_pointer.
18473         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
18474         (arm_init_builtins): Handle the above.
18475         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
18476         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
18477         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
18478         void intrinsics.
18480 2018-06-05  Martin Liska  <mliska@suse.cz>
18482         * auto-profile.c (read_autofdo_file): Do not use
18483         gcov_ctr_summary struct.
18484         (afdo_callsite_hot_enough_for_early_inline): Likewise.
18485         * coverage.c (struct counts_entry): Likewise.
18486         (read_counts_file): Read just single summary entry.
18487         (get_coverage_counts): Use gcov_summary struct.
18488         * coverage.h (get_coverage_counts): Likewise.
18489         * gcov-dump.c (dump_working_sets): Likewise.
18490         (tag_summary): Dump just single summary.
18491         * gcov-io.c (gcov_write_summary): Write just histogram
18492         summary.
18493         (gcov_read_summary): Read just single summary.
18494         (compute_working_sets): Use gcov_summary struct.
18495         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
18496         of GCOV_COUNTERS_SUMMABLE.
18497         (GCOV_COUNTERS_SUMMABLE): Remove.
18498         (GCOV_FIRST_VALUE_COUNTER): Replace with
18499         GCOV_COUNTER_V_INTERVAL.
18500         (struct gcov_ctr_summary): Remove.
18501         (struct gcov_summary): Directly use fields of former
18502         gcov_ctr_summary.
18503         (compute_working_sets): Use gcov_summary struct.
18504         * gcov.c (read_count_file): Do not use ctrs fields.
18505         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
18506         struct.
18507         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
18508         struct.
18509         * profile.c: Likewise.
18510         * profile.h: Likewise.
18512 2018-06-05  Martin Liska  <mliska@suse.cz>
18514         PR gcov-profile/84846
18515         * gcov.c (output_lines): Print working directory only
18516         in intermediate format.
18518 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
18520         * config/s390/s390-builtin-types.def: Add void function type.
18521         * config/s390/s390-builtins.def: Use the function type for the
18522         tbeginc builtin.
18524 2018-06-04  Jim Wilson  <jimw@sifive.com>
18526         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
18527         to int.
18528         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
18529         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
18530         handle EH_RETURN_DATA_REGNO registers properly.
18531         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
18532         (riscv_expand_epilogue): Update comment.  Change argument name and
18533         type.  Update code to use new name and type.  Pass new args to
18534         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
18535         EXCEPTION_RETURN.
18536         * config/riscv/riscv.md (NORMAL_RETURN): New.
18537         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
18538         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
18539         (eh_return): Call gen_eh_return_internal and emit barrier.
18540         (eh_return_internal): Call riscv_expand_epilogue.
18542 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
18544         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
18545         bit_insertion field and declare can_be_merged_into method.
18546         (merged_store_group::can_be_merged_into): New method.
18547         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
18548         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
18549         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
18551 2018-06-04  Richard Biener  <rguenther@suse.de>
18553         PR tree-optimization/85955
18554         * builtins.c (fold_builtin_sincos): Convert pointers to
18555         destination to appropriate type before dereferencing.
18557 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
18559         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
18561 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
18563         * expr.c (expand_expr_real_1): Force the operand into memory if
18564         its TYPE_MODE is BLKmode and if there is no integer mode for
18565         the number of bits being extracted.
18567 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
18569         PR target/85832
18570         PR target/86036
18571         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
18572         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
18574 2018-06-04  Richard Biener  <rguenther@suse.de>
18576         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
18577         (cleanup_tree_cfg_noloop): ... single caller.  Do
18578         start_recording_case_labels later.
18580 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
18582         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
18583         to _IMMINTRIN_H_INCLUDED.
18584         * config/i386/pconfigintrin.h: Ditto.
18585         * config/i386/waitpkgintrin.h: Ditto.
18586         * config/i386/immintrin.h: Add includes for sgxintrin.h,
18587         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
18588         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
18589         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
18590         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
18591         waitpkgintrin.h and cldemoteintrin.h.
18593 2018-06-04  Richard Biener  <rguenther@suse.de>
18595         PR tree-optimization/86038
18596         * tracer.c (find_best_successor): Check probability for
18597         being initialized, bail out if not.
18599 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
18601         PR target/86003
18602         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
18603         of bits to ignore when comparing architectures.
18605 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
18607         PR tree-optimization/69615
18608         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
18609         maximum or minimum of the type, try to merge it also as if
18610         range1 is + [-, x - 1] or + [x + 1, -].
18612         PR c++/86025
18613         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
18615 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
18617         PR tree-optimization/86034
18618         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
18619         the unsigned bitfield type in a bit insertion sequence if it does not
18620         have a larger precision than the bitfield size.
18621         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
18623 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
18625         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
18627 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
18629         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
18630         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
18631         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
18632         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
18634 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
18636         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
18637         Disable -fdelete-null-pointer-checks for ELF toolchain.
18639 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
18640             Kito Cheng  <kito.cheng@gmail.com>
18642         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
18643         (nds32le-*-*, nds32be-*-*): Integrate checking process.
18644         (nds32*-*-*): Add glibc and uclibc conditions.
18645         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
18646         (TARGET_EXCEPT_UNWIND_INFO): Define.
18647         * config/nds32/elf.h: New file.
18648         * config/nds32/linux.h: New file.
18649         * config/nds32/nds32-elf.opt: New file.
18650         * config/nds32/nds32-linux.opt: New file.
18651         * config/nds32/nds32-fp-as-gp.c
18652         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
18653         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
18654         TARGET_LINUX_ABI.
18655         (nds32_asm_file_end): Ditto.
18656         (nds32_print_operand): Ditto.
18657         (nds32_insert_attributes): Ditto.
18658         (nds32_init_libfuncs): New function.
18659         (TARGET_HAVE_TLS): Define.
18660         (TARGET_INIT_LIBFUNCS): Define.
18661         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
18662         spec content.
18663         (TARGET_ELF): Apply different mcmodel setting.
18664         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
18665         been migrated into elf.h and linux.h files.
18666         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
18667         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
18668         (mcmodel): The content has been migrated into nds32-elf.opt and
18669         nds32-linux.opt files.
18670         * config/nds32/t-elf: New file.
18671         * config/nds32/t-linux: New file.
18673 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
18674             Shiva Chen  <shiva0217@gmail.com>
18676         * config/nds32/constants.md (unspec_volatile_element): Add
18677         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
18678         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
18679         optimization.
18680         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
18681         (make_pass_nds32_fp_as_gp): Declare.
18682         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
18683         optmization pass.
18684         (nds32_asm_function_end_prologue): Remove unused asm output.
18685         (nds32_asm_function_begin_epilogue): Remove unused asm output.
18686         (nds32_asm_file_start): Output necessary fp_as_gp information.
18687         (nds32_option_override): Adjust register usage.
18688         (nds32_expand_prologue): Consider fp_as_gp situation.
18689         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
18690         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
18691         (epilogue): Ditto.
18692         (return): Ditto.
18693         (simple_return): Ditto.
18694         (omit_fp_begin): Output special directive for fp_as_gp.
18695         (omit_fp_end): Output special directive for fp_as_gp.
18696         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
18697         mforbid-fp-as-gp): New options.
18699 2018-06-01  Mark Wielaard  <mark@klomp.org>
18701         * dwarf2out.c (dwarf2out_finish): Remove generation of
18702         DW_AT_loclists_base.
18704 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
18706         * gimple-ssa-store-merging.c: Include gimple-fold.h.
18707         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
18708         (struct merged_store_group): Add bit_insertion field.
18709         (dump_char_array): Use standard hexadecimal format.
18710         (merged_store_group::merged_store_group): Set bit_insertion to false.
18711         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
18712         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
18713         also print the mask in the dump file.
18714         (pass_store_merging::gate): Minor tweak.
18715         (imm_store_chain_info::coalesce_immediate): Fix wrong association
18716         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
18717         stores with INTEGER_CST stores.
18718         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
18719         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
18720         and use it throughout.  Generate bit insertion sequences if need be.
18721         (pass_store_merging::process_store): Remove redundant condition.
18722         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
18724 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
18726         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
18727         the 128-bit floating point types.  Fix function comment.
18729 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18731         * config/aarch64/aarch64-simd.md
18732         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
18733         mnemonics.
18734         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
18735         mnemonics.
18737 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
18739         PR tree-optimization/85989
18740         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
18741         variable.
18742         (backprop::intersect_uses): Check it when deciding whether this
18743         is a backedge reference.
18744         (backprop::process_block): Add each phi to m_visited_phis
18745         after visiting it, then clear it at the end.
18747 2018-06-01  Richard Biener  <rguenther@suse.de>
18749         * tree-vectorizer.h (vect_dr_stmt): New function.
18750         (vect_get_load_cost): Adjust.
18751         (vect_get_store_cost): Likewise.
18752         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
18753         Use vect_dr_stmt instead of DR_SMTT.
18754         (vect_record_base_alignments): Likewise.
18755         (vect_calculate_target_alignment): Likewise.
18756         (vect_compute_data_ref_alignment): Likewise and make static.
18757         (vect_update_misalignment_for_peel): Likewise.
18758         (vect_verify_datarefs_alignment): Likewise.
18759         (vector_alignment_reachable_p): Likewise.
18760         (vect_get_data_access_cost): Likewise.  Pass down
18761         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
18762         (vect_get_peeling_costs_all_drs): Likewise.
18763         (vect_peeling_hash_get_lowest_cost): Likewise.
18764         (vect_enhance_data_refs_alignment): Likewise.
18765         (vect_find_same_alignment_drs): Likewise.
18766         (vect_analyze_data_refs_alignment): Likewise.
18767         (vect_analyze_group_access_1): Likewise.
18768         (vect_analyze_group_access): Likewise.
18769         (vect_analyze_data_ref_access): Likewise.
18770         (vect_analyze_data_ref_accesses): Likewise.
18771         (vect_vfa_segment_size): Likewise.
18772         (vect_small_gap_p): Likewise.
18773         (vectorizable_with_step_bound_p): Likewise.
18774         (vect_prune_runtime_alias_test_list): Likewise.
18775         (vect_analyze_data_refs): Likewise.
18776         (vect_supportable_dr_alignment): Likewise.
18777         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
18778         (vect_gen_prolog_loop_niters): Likewise.
18779         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
18780         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
18781         modify DR_STMT.
18782         (vect_recog_mask_conversion_pattern): Likewise.
18783         (vect_try_gather_scatter_pattern): Likewise.
18784         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
18785         to vect_get_store_cost.
18786         (vect_get_store_cost): Get stmt_info instead of DR.
18787         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
18788         (vect_get_load_cost): Get stmt_info instead of DR.
18790 2018-06-01  Richard Biener  <rguenther@suse.de>
18792         PR middle-end/86017
18793         * gimple-fold.c (var_decl_component_p): Also allow offsetted
18794         vars wrapped in MEM_REFs.
18796 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
18798         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
18799         Fix subreg tests so that we only return a choice between
18800         GENERAL_REGS and FP_REGS if the original classes included both.
18802 2018-06-01  Richard Biener  <rguenther@suse.de>
18804         PR ipa/85960
18805         * tree-ssa-structalias.c (get_function_part_constraint):
18806         Handle NULL fi->decl.
18807         (find_func_aliases_for_call): Properly handle indirect
18808         fi from direct call.
18809         (find_func_clobbers): Likewise.
18810         (ipa_pta_execute): Likewise.
18811         (create_variable_info_for): For functions that are ifunc_resolver
18812         resolve to a varinfo that contains the result of the resolver call.
18813         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
18814         aliases.
18816 2018-05-31  Michael Collison  <michael.collison@arm.com>
18818         * config/aarch64/aarch64.md:
18819         (*fix_to_zero_extenddfdi2): New pattern.
18820         * gcc.target/aarch64/fix_extend1.c: New testcase.
18822 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
18824         PR middle-end/78809
18825         PR middle-end/83026
18826         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
18827         and BUILT_IN_STRNCMP_EQ.
18828         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
18829         BUILT_IN_STRNCMP_EQ.
18830         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
18831         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
18832         (gimple_fold_builtin): Likewise.
18833         * tree-ssa-strlen.c (compute_string_length): New function.
18834         (determine_min_obsize): New function.
18835         (handle_builtin_string_cmp): New function to handle calls to
18836         string compare functions.
18837         (strlen_optimize_stmt): Add handling to builtin string compare
18838         calls.
18839         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
18840         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
18841         * tree.c (build_common_builtin_nodes): Add new defines of
18842         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
18844 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
18846         PR target/85984
18847         * bb-reorder.c (pass_partition_blocks::gate): Return false for
18848         functions with naked attribute.
18850 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
18852         * config/i386/sse.md (avx_vec_concat<mode>):
18853         Substitute concat_tg_mode mode attribute with xtg_mode.
18854         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
18855         (concat_tg_mode): Remove mode attribute.
18857 2018-05-31  Martin Sebor  <msebor@redhat.com>
18859         PR c/82063
18860         * calls.c (alloc_max_size): Correct a logic error/typo.
18861         Treat excessive arguments as infinite.  Warn for invalid arguments.
18862         * doc/invoke.texi (-Walloc-size-larger-than): Update.
18864 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
18866         PR target/85829
18867         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
18868         and movx for Haswell.
18870 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
18871             Cesar Philippidis  <cesar@codesourcery.com>
18873         PR middle-end/85879
18874         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
18875         when emitting error on private/firstprivate reductions.
18876         * omp-low.c (lower_omp_target): Avoid reference-type processing
18877         on pointers for firstprivate clause.
18879 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
18881         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
18882         (st1x2): Likewise.
18883         (st1x3): Likewise.
18884         * config/aarch64/aarch64-simd.md
18885         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
18886         (aarch64_ld1_x3_<mode>): Likewise
18887         (aarch64_st1x2<VALLDIF:mode>): Likewise
18888         (aarch64_st1_x2_<mode>): Likewise
18889         (aarch64_st1x3<VALLDIF:mode>): Likewise
18890         (aarch64_st1_x3_<mode>): Likewise
18891         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
18892         (vld1_s8_x3): Likewise.
18893         (vld1_u16_x3): Likewise.
18894         (vld1_s16_x3): Likewise.
18895         (vld1_u32_x3): Likewise.
18896         (vld1_s32_x3): Likewise.
18897         (vld1_u64_x3): Likewise.
18898         (vld1_s64_x3): Likewise.
18899         (vld1_f16_x3): Likewise.
18900         (vld1_f32_x3): Likewise.
18901         (vld1_f64_x3): Likewise.
18902         (vld1_p8_x3): Likewise.
18903         (vld1_p16_x3): Likewise.
18904         (vld1_p64_x3): Likewise.
18905         (vld1q_u8_x3): Likewise.
18906         (vld1q_s8_x3): Likewise.
18907         (vld1q_u16_x3): Likewise.
18908         (vld1q_s16_x3): Likewise.
18909         (vld1q_u32_x3): Likewise.
18910         (vld1q_s32_x3): Likewise.
18911         (vld1q_u64_x3): Likewise.
18912         (vld1q_s64_x3): Likewise.
18913         (vld1q_f16_x3): Likewise.
18914         (vld1q_f32_x3): Likewise.
18915         (vld1q_f64_x3): Likewise.
18916         (vld1q_p8_x3): Likewise.
18917         (vld1q_p16_x3): Likewise.
18918         (vld1q_p64_x3): Likewise.
18919         (vst1_s64_x2): Likewise.
18920         (vst1_u64_x2): Likewise.
18921         (vst1_f64_x2): Likewise.
18922         (vst1_s8_x2): Likewise.
18923         (vst1_p8_x2): Likewise.
18924         (vst1_s16_x2): Likewise.
18925         (vst1_p16_x2): Likewise.
18926         (vst1_s32_x2): Likewise.
18927         (vst1_u8_x2): Likewise.
18928         (vst1_u16_x2): Likewise.
18929         (vst1_u32_x2): Likewise.
18930         (vst1_f16_x2): Likewise.
18931         (vst1_f32_x2): Likewise.
18932         (vst1_p64_x2): Likewise.
18933         (vst1q_s8_x2): Likewise.
18934         (vst1q_p8_x2): Likewise.
18935         (vst1q_s16_x2): Likewise.
18936         (vst1q_p16_x2): Likewise.
18937         (vst1q_s32_x2): Likewise.
18938         (vst1q_s64_x2): Likewise.
18939         (vst1q_u8_x2): Likewise.
18940         (vst1q_u16_x2): Likewise.
18941         (vst1q_u32_x2): Likewise.
18942         (vst1q_u64_x2): Likewise.
18943         (vst1q_f16_x2): Likewise.
18944         (vst1q_f32_x2): Likewise.
18945         (vst1q_f64_x2): Likewise.
18946         (vst1q_p64_x2): Likewise.
18947         (vst1_s64_x3): Likewise.
18948         (vst1_u64_x3): Likewise.
18949         (vst1_f64_x3): Likewise.
18950         (vst1_s8_x3): Likewise.
18951         (vst1_p8_x3): Likewise.
18952         (vst1_s16_x3): Likewise.
18953         (vst1_p16_x3): Likewise.
18954         (vst1_s32_x3): Likewise.
18955         (vst1_u8_x3): Likewise.
18956         (vst1_u16_x3): Likewise.
18957         (vst1_u32_x3): Likewise.
18958         (vst1_f16_x3): Likewise.
18959         (vst1_f32_x3): Likewise.
18960         (vst1_p64_x3): Likewise.
18961         (vst1q_s8_x3): Likewise.
18962         (vst1q_p8_x3): Likewise.
18963         (vst1q_s16_x3): Likewise.
18964         (vst1q_p16_x3): Likewise.
18965         (vst1q_s32_x3): Likewise.
18966         (vst1q_s64_x3): Likewise.
18967         (vst1q_u8_x3): Likewise.
18968         (vst1q_u16_x3): Likewise.
18969         (vst1q_u32_x3): Likewise.
18970         (vst1q_u64_x3): Likewise.
18971         (vst1q_f16_x3): Likewise.
18972         (vst1q_f32_x3): Likewise.
18973         (vst1q_f64_x3): Likewise.
18974         (vst1q_p64_x3): Likewise.
18976 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18978         * config/msp430/msp430.c (msp430_output_labelref): Prepend
18979         user_label_prefix to name.
18981         * tree-core.h: Update comment about the format of NAME string
18982         passed to handler in attribute_spec.
18984         * config/msp430/msp430.md: Remove erroneous subreg expression from
18985         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
18986         zero_extend{q,h}isi2.
18988 2018-05-30  Borislav Petkov  <bp@suse.de>
18990         * doc/extend.texi: Document some architecture specific
18991         constraints and sort entries.
18993 2018-05-30  Martin Sebor  <msebor@redhat.com>
18995         PR middle-end/85369
18996         * builtins.c (expand_builtin_stpcpy_1): New function.
18997         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
18998         only if the former succeeds.
19000 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
19002         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
19003         in saphira.
19005 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
19007         * doc/invoke.texi (-flinker-output): Document
19009 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
19011         * passes.c (ipa_write_summaries): Only modify statements if body
19012         is in memory.
19013         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
19014         incrementally linking.
19015         (ipa_passes): Likewise.
19016         * lto-cgraph.c (lto_output_node): When incrementally linking do not
19017         pass down resolution info.
19018         * common.opt (flag_incremental_link): Update info.
19019         * gcc.c (plugin specs): Turn flinker-output=* to
19020         -plugin-opt=-linker-output-known
19021         * toplev.c (compile_file): Also cut compilation when doing incremental
19022         link.
19023         * flag-types. (enum lto_partition_model): Add
19024         LTO_LINKER_OUTPUT_NOLTOREL.
19025         (invoke.texi): Add -flinker-output docs.
19026         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
19027         link same way as WPA; do not stream in dead initializers.
19029         * dwarf2out.c (dwarf2out_die_ref_for_decl,
19030         darf2out_register_external_decl): Support incremental link.
19032 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
19034         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
19036 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
19038         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
19039         it down to simple_object_copy_lto_debug_sections.
19040         (run_gcc): Determine incremental LTO link time and configure
19041         lto1 into non-wpa mode, disable renaming of debug sections.
19043 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19045         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
19046         descriptions of various incorrectly documented functions.
19048 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
19050         Revert:
19051         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
19052         address check not strict.
19054 2018-05-30  Richard Biener  <rguenther@suse.de>
19056         PR tree-optimization/85964
19057         * tracer.c (better_p): Drop initialized count check, we only
19058         call the function with initialized counts now.
19059         (find_best_successor): Do find a best edge if one
19060         has uninitialized count.
19061         (find_best_predecessor): Likewise.  Do BB frequency check only
19062         if count is initialized.
19064 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
19066         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
19067         (aarch64_ldrstr_offset_compare): New.
19068         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
19069         load/store orderings.
19070         (aarch64_gen_adjusted_ldpstp): Likewise.
19072 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
19074         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
19075         Check for subset of GENERAL_REGS and FP_REGS.
19076         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
19077         r=w alternative.
19079 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
19081         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
19082         and wi::to_poly_offset.  Add the current offset and then check
19083         whether the sum fits, rather than using an unchecked addition of
19084         a checked term.  Check for a shwi rather than a uhwi.
19085         * expr.c (get_bit_range): Use tree_to_poly_uint64.
19086         (store_constructor): Use poly_int_tree_p.
19087         (expand_expr_real_1): Likewise.
19088         * function.c (assign_temp): Likewise.
19089         * fold-const.c (const_binop): Use poly_int_tree_p and
19090         wi::to_poly_offset.
19091         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
19092         division.
19093         * ipa-icf-gimple.c (func_checker::compare_operand): Use
19094         to_poly_offset for MEM offsets.
19095         * ipa-icf.c (sem_variable::equals): Likewise.
19096         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
19097         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
19098         wi::to_poly_offset for BIT_FIELD_REF offsets.
19099         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
19100         wi::to_poly_offset.
19101         * var-tracking.c (emit_note_insn_var_location): Use
19102         tree_to_poly_uint64.
19104 2018-05-29  Jim Wilson  <jimw@sifive.com>
19106         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
19108 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
19110         PR target/85950
19111         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
19112         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
19113         sequence.
19114         (sse4_1_round<mode>2): Use nonimmediate_operand
19115         for operand 1 predicate.
19117 2018-05-29  Martin Sebor  <msebor@redhat.com>
19118             Richard Biener  <rguenther@suse.de>
19120         PR testsuite/85888
19121         * calls.c (get_size_range): Call determine_value_range instead
19122         of get_value_range..
19123         * tree-vrp.h (determine_value_range): Declared new function.
19124         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
19126 2018-05-29  Richard Biener  <rguenther@suse.de>
19128         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
19129         sure to use non-pattern stmts for get_earlier_stmt arguments.
19130         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
19131         called on pattern stmts.
19132         (get_later_stmt): Likewise.
19134 2018-05-29  Martin Liska  <mliska@suse.cz>
19136         PR gcov-profile/85759
19137         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
19138         env variables.
19140 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
19142         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
19143         VEC_UNPACK_*_EXPR.
19144         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
19145         VEC_PACK_*_EXPR.
19147         PR target/85918
19148         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
19149         VEC_PACK_FLOAT_EXPR): New tree codes.
19150         * tree-pretty-print.c (op_code_prio): Handle
19151         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
19152         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
19153         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
19154         * tree-inline.c (estimate_operator_cost): Likewise.
19155         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
19156         * fold-const.c (const_binop): Likewise.
19157         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
19158         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
19159         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
19160         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
19161         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
19162         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
19163         * expr.c (expand_expr_real_2): Likewise.
19164         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
19165         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
19166         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
19167         optabs.
19168         * optabs.c (expand_widen_pattern_expr): For
19169         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
19170         sign from result type rather than operand's type.
19171         (expand_binop_directly): For vec_packu_float_optab and
19172         vec_packs_float_optab allow result type to be different from operand's
19173         type.
19174         * optabs-tree.c (optab_for_tree_code): Handle
19175         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
19176         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
19177         * tree-vect-generic.c (expand_vector_operations_1):  Handle
19178         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
19179         VEC_PACK_FLOAT_EXPR.
19180         * tree-vect-stmts.c (supportable_widening_operation): Handle
19181         FIX_TRUNC_EXPR.
19182         (supportable_narrowing_operation): Handle FLOAT_EXPR.
19183         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
19184         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
19185         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
19186         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
19187         mode attributes.
19188         (vec_pack<floatprefix>_float_<mode>): New expander.
19189         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
19190         attributes.
19191         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
19192         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
19193         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
19194         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
19195         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
19196         Document.
19197         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
19198         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
19199         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
19200         VEC_PACK_FLOAT_EXPR): Document.
19202 2018-05-29  Richard Biener  <rguenther@suse.de>
19204         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
19205         member.
19206         (stmt_vec_info_vec): Make pointer.
19207         (init_stmt_vec_info_vec): Remove.
19208         (free_stmt_vec_info_vec): Likewise.
19209         (set_stmt_vec_info_vec): New function.
19210         (free_stmt_vec_infos): Likewise.
19211         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
19212         (set_vinfo_for_stmt): Likewise.
19213         (get_earlier_stmt): Likewise.
19214         (get_later_stmt): Likewise.
19215         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
19216         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
19217         (vec_info::~vec_info): Free stmt_vec_infos.
19218         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
19219         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
19220         (pass_slp_vectorize::execute): Likewise.
19221         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
19222         (free_stmt_vec_info_vec): Likewise.
19223         (set_stmt_vec_info_vec): New function.
19224         (free_stmt_vec_infos): Likewise.
19225         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
19226         the global stmt_vec_info_vec.
19227         * tree-parloops.c (gather_scalar_reductions): Use
19228         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
19229         vector.
19231 2018-05-29  Richard Biener  <rguenther@suse.de>
19233         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
19235 2018-05-29  Martin Liska  <mliska@suse.cz>
19236             David Malcolm  <dmalcolm@redhat.com>
19238         * vec.c (test_reverse): New.
19239         (vec_c_tests): Add new test.
19240         * vec.h (vl_ptr>::reverse): New function.
19242 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
19244         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
19246         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
19247         and later.
19249 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
19251         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
19253 2018-05-28  Richard Biener  <rguenther@suse.de>
19255         PR tree-optimization/85933
19256         * tree-vect-data-refs.c (vect_record_base_alignments): Only
19257         look at stmts marked as vectorizable.
19259 2018-05-28  Richard Biener  <rguenther@suse.de>
19261         PR tree-optimization/85934
19262         * tree-vect-generic.c (expand_vector_operations_1): Hoist
19263         vector boolean check before scalar optimization.
19265 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
19267         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
19268         for armv5te.
19270 2018-05-28  Mark Wielaard  <mark@klomp.org>
19272         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
19273         if it is an expression containing a minus sign.
19275 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
19277         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
19279 2018-05-27  Paul Koning  <ni1d@arrl.net>
19281         * config/pdp11/pdp11.md (truncsihi2): Remove.
19283 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
19284             Chung-Ju Wu  <jasonwucj@gmail.com>
19286         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
19287         implementation.
19288         (unaligned_store_dw): Ditto.
19289         * config/nds32/nds32-memory-manipulation.c
19290         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
19291         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
19292         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
19293         (emit_setmem_word_loop): Rename to ...
19294         (emit_setmem_doubleword_loop): ... this.
19295         (nds32_gen_dup_4_byte_to_word_value): New function.
19296         (nds32_gen_dup_8_byte_to_double_word_value): New function.
19297         (nds32_expand_setmem_loop): Refine implementation.
19298         (nds32_expand_setmem_loop_v3m): Ditto.
19299         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
19300         pattern.
19302 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
19304         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
19306 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
19308         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
19309         (nds32_init_machine_status): Initialize machine->attr_naked_p and
19310         machine->attr_no_prologue_p.
19311         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
19312         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
19313         (nds32_expand_epilogue): Consider attr_naked_p.
19314         (nds32_expand_epilogue_v3pop): Likewise.
19315         (nds32_can_use_return_insn): Likewise.
19316         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
19317         attr_no_prologue_p fields.
19318         * config/nds32/nds32.opt (mret-in-naked-func): New option.
19320 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
19322         PR target/85918
19323         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
19324         attributes.
19325         * config/i386/sse.md
19326         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
19327         Rename to ...
19328         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
19329         ... this.
19330         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
19331         Rename to ...
19332         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
19333         ... this.
19334         (*<floatsuffix>floatv2div2sf2): Rename to ...
19335         (*float<floatunssuffix>v2div2sf2): ... this.
19336         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
19337         (float<floatunssuffix>v2div2sf2_mask): ... this.
19338         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
19339         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
19340         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
19341         to ...
19342         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
19343         ... this.
19344         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
19345         Rename to ...
19346         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
19347         ... this.
19348         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
19349         Rename to ...
19350         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
19351         ... this.
19352         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
19353         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
19354         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
19355         gen_ufix_truncv8dfv8si2.
19356         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
19357         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
19358         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
19359         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
19360         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
19361         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
19362         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
19363         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
19365 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
19367         PR target/85900
19368         PR target/85345
19369         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
19371 2018-05-25  Jim Wilson  <jimw@sifive.com>
19373         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
19374         * config/riscv/riscv.c (struct machine_function): Add
19375         interrupt_handler_p and attribute_checked_p fields.
19376         (riscv_attribute_table): Add interrupt.
19377         (riscv_interrupt_type_p): New.
19378         (riscv_save_reg_p): Save extra regs for interrupt handler.
19379         (riscv_use_save_libcall): Return false  for interrupt handler.
19380         (riscv_first_stack_step): Add forward declaration.
19381         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
19382         for interrupt handler with large frame.  Use it for saved reg list.
19383         (riscv_expand_prologue): Move flag_stack_usage_info support to
19384         eliminate duplication.
19385         (riscv_expand_epilogue): Generate mret for interrupt handler.
19386         (riscv_epilogue_uses): New.
19387         (riscv_can_use_return_insn): Return false for interrupt handler.
19388         (riscv_function_ok_for_sibcall): Likewise.
19389         (riscv_set_current_function): Add interrupt handler support.
19390         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
19391         * config/riscv/riscv.md (UNSPECV_MRET): New.
19392         (GP_REGNUM): New.
19393         (riscv_frflags, riscv_fsflags): Use tab after opcode.
19394         (riscv_mret): New.
19395         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
19397 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
19399         PR tree-optimization/85712
19400         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
19401         this candidate has already been replaced in-situ by a copy.
19403 2018-05-25  Jason Merrill  <jason@redhat.com>
19405         PR c++/80485 - inline function non-zero address.
19406         * symtab.c (nonzero_address): Check DECL_COMDAT.
19408 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
19410         PR target/83628
19411         * config/alpha/alpha.md (ashlsi3): New insn pattern.
19412         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
19413         extension of SImode operation.  Use const123_operand predicate.
19414         (*saddsi_1): Remove.
19415         (*saddl_se_1): Ditto.
19416         (*ssubsi_1): Ditto.
19417         (*ssubl_se_1): Ditto.
19418         * config/alpha/predicates.md (const123_operand): New predicate.
19419         * config/alpha/constraints.md (P): Use IN_RANGE.
19421 2018-05-25  Richard Biener  <rguenther@suse.de>
19423         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
19424         defaulted to true.
19425         (ref_maybe_used_by_stmt_p): Likewise.
19426         (stmt_may_clobber_ref_p): Likewise.
19427         (stmt_may_clobber_ref_p_1): Likewise.
19428         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
19429         and pass it along.
19430         (ref_maybe_used_by_stmt_p): Likewise.
19431         (stmt_may_clobber_ref_p): Likewise.
19432         (stmt_may_clobber_ref_p_1): Likewise.
19433         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
19434         the alias oracle to disambiguate DRs with stmts DR analysis
19435         couldn't handle.
19436         (vect_analyze_data_refs): Do not give up on not analyzable
19437         DRs for BB vectorization.  Remove code truncating the dataref
19438         vector.
19440 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
19442         PR target/85832
19443         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
19444         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
19445         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
19447 2018-05-25  Richard Biener  <rguenther@suse.de>
19449         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
19450         function, combining stmt data ref gathering and fatal analysis
19451         parts.
19452         (vect_analyze_data_refs): Remove now redudnant code and simplify.
19453         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
19454         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
19455         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
19456         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
19458 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
19460         PR tree-optimization/85720
19461         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
19462         SCC if all partitions are builtins.
19463         (version_loop_by_alias_check): New parameter.  Generate cancelable
19464         runtime alias check if all partitions are builtins.
19465         (distribute_loop): Update call to above function.
19467 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
19469         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
19470         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
19471         (parm_default_def_partition_arg): Ditto.
19472         (set_parm_default_def_partition): Ditto.
19473         (get_parm_default_def_partitions): Ditto and make it static.
19474         (get_undefined_value_partitions): Ditto and make it static.
19475         (remove_ssa_form): Refactor call to init_var_map here.
19476         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
19477         computation for loop region.
19478         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
19479         (register_default_def): Delete.
19480         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
19481         (parm_default_def_partition_arg): Ditto.
19482         (set_parm_default_def_partition): Ditto.
19483         (get_parm_default_def_partitions): Ditto and make it static.
19484         (get_undefined_value_partitions): Ditto and make it static.
19485         (coalesce_with_default, coalesce_with_default): Update comment.
19486         (create_coalesce_list_for_region): New func factored out from
19487         create_outofssa_var_map.
19488         (populate_coalesce_list_for_outofssa): New func factored out from
19489         create_outofssa_var_map and coalesce_ssa_name.
19490         (create_outofssa_var_map): Delete.
19491         (coalesce_ssa_name): Refactor to support live range computation.
19492         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
19493         (get_parm_default_def_partitions): Delete.
19494         (get_undefined_value_partitions): Ditto.
19495         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
19496         computation for loop region.
19497         (new_tree_live_info, loe_visit_block): Ditto.
19498         (live_worklist, set_var_live_on_entry): Ditto.
19499         (calculate_live_on_exit, verify_live_on_entry): Ditto.
19500         * tree-ssa-live.h (struct _var_map): New fields.
19501         (init_var_map): Change decl.
19502         (region_contains_p): New.
19504 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
19506         * tree-ssa-live.h (live_merge_and_clear): Delete.
19508 2018-05-25  Richard Biener  <rguenther@suse.de>
19510         PR c++/85912
19511         * tree-dump.c (dequeue_and_dump): Remove access to removed
19512         operand 2 of a SWITCH_EXPR.
19514 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
19516         * doc/sourcebuild.texi (vect_double_cond_arith): Include
19517         multiplication and division.
19518         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
19519         (cond_udiv@var{m}, cond_umod@var{m}): Document.
19520         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
19521         (cond_udiv_optab, cond_umod_optab): New optabs.
19522         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
19523         (IFN_COND_RDIV): New internal functions.
19524         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
19525         TRUNC_MOD_EXPR and RDIV_EXPR.
19526         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
19527         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
19528         New unspecs.
19529         (SVE_INT_BINARY): Include mult.
19530         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
19531         (optab, sve_int_op): Handle mult.
19532         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
19533         UNSPEC_COND_DIV.
19534         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
19535         for SVE_INT_BINARY_SD.
19537 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
19539         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
19540         (optab, sve_int_op): Handle div and udiv.
19541         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
19542         for SVE_INT_BINARY_SD.
19543         (*<optab><mode>3): New insn for the same.
19545 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
19547         * tree-vect-patterns.c: Include predict.h.
19548         (vect_recog_divmod_pattern): Restrict check for division support
19549         to when optimizing for size.
19551 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
19553         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
19554         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
19555         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
19556         (gimple_match_op::set_op): Likewise.
19557         (gimple_resimplify4): Declare.
19558         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
19559         (expr::gen_transform): Likewise.
19560         (decision_tree::gen): Generate a simplification routine for 4 operands.
19561         * gimple-match-head.c (gimple_simplify): Add an overload for
19562         4 operands.  In the top-level function, handle up to 4 call
19563         arguments and call gimple_resimplify4.
19564         (gimple_resimplify4): New function.
19565         (build_call_internal): Pass a fourth operand.
19566         (maybe_push_to_seq): Likewise.
19567         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
19568         Fold VEC_COND_EXPRs of an operation and a default value into
19569         an IFN_COND_* function if possible.
19570         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
19571         New unspecs.
19572         (SVE_COND_FP_BINARY): Include them.
19573         (optab, sve_fp_op): Handle them.
19574         (SVE_INT_BINARY_REV): New code iterator.
19575         (SVE_COND_FP_BINARY_REV): New int iterator.
19576         (commutative): New int attribute.
19577         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
19578         Declare.
19579         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
19580         function.
19581         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
19582         (*cond_<optab><mode>): New patterns for reversed operands.
19584 2018-05-25  Richard Biener  <rguenther@suse.de>
19586         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
19587         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
19588         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
19589         (STMT_VINFO_GROUPED_ACCESS): Adjust.
19590         * tree-vect-data-refs.c (everywhere): Adjust users.
19591         * tree-vect-loop.c (everywhere): Likewise.
19592         * tree-vect-slp.c (everywhere): Likewise.
19593         * tree-vect-stmts.c (everywhere): Likewise.
19594         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
19596 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
19598         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
19599         Rename to...
19600         (gcc_cv_as_section_exclude): ... this.
19601         Try Solaris as #exclude syntax.
19602         * configure: Regenerate.
19603         * config.in: Regenerate.
19604         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
19605         SECTION_EXCLUDE.
19606         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
19607         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
19609         * varasm.c (default_elf_asm_named_section): Don't check if
19610         HAVE_GAS_SECTION_EXCLUDE is defined.
19612 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
19614         * doc/md.texi: Update the documentation of the cond_* optabs
19615         to mention the new final operand.  Fix GET_MODE_NUNITS call.
19616         Describe the scalar case too.
19617         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
19618         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
19619         instead of 2.
19620         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
19621         (get_conditional_internal_fn): Update comment.
19622         * tree-vect-loop.c (vectorizable_reduction): Pass the original
19623         accumulator value as a final argument to conditional functions.
19624         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
19625         a define_expand and add an "else" operand.  Assert for now that
19626         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
19627         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
19628         (*cond_<optab><mode>): New patterns.
19629         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
19630         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
19631         (UNSPEC_COND_EOR): Delete.
19632         (optab): Remove associated mappings.
19633         (SVE_INT_BINARY): New code iterator.
19634         (sve_int_op): Remove int attribute and add "minus" to the code
19635         attribute.
19636         (SVE_COND_INT_OP): Delete.
19637         (SVE_COND_FP_OP): Rename to...
19638         (SVE_COND_FP_BINARY): ...this.
19640 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
19642         * optabs.c (can_reuse_operands_p): New function.
19643         (maybe_legitimize_operands): Try to reuse the results for
19644         earlier operands.
19646 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
19648         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
19649         Add {q} suffix to insn mnemonic.
19651 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
19653         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
19654         (msp430_warn_func_return): New.
19656 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
19658         * fold-const.c (tree_nonzero_bits): New function.
19659         * fold-const.h (tree_nonzero_bits): Likewise.
19660         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
19661         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
19663 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
19665         PR target/85900
19666         PR target/85345
19667         * varasm.c (assemble_alias): Check ifunc_resolver only on
19668         FUNCTION_DECL.
19670 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
19672         PR target/85903
19673         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
19674         when memory input operand is handled.
19676 2018-05-24  Luis Machado  <luis.machado@linaro.org>
19678         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
19679         global.
19680         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
19682 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
19684         * match.pd: Delay FMA folds until after vectorization.
19686 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
19688         PR target/83009
19689         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
19690         address check not strict.
19692 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
19694         * gimple-match.h (gimple_match_op): New class.
19695         (mprts_hook): Replace parameters with a gimple_match_op *.
19696         (maybe_build_generic_op): Likewise.
19697         (gimple_simplified_result_is_gimple_val): Replace parameters with
19698         a const gimple_match_op *.
19699         (gimple_simplify): Replace code_helper * and tree * parameters with
19700         a gimple_match_op * parameter.
19701         (gimple_resimplify1): Replace code_helper *, tree and tree *
19702         parameters with a gimple_match_op * parameter.
19703         (gimple_resimplify2): Likewise.
19704         (gimple_resimplify3): Likewise.
19705         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
19706         parameters with a gimple_match_op * parameter.
19707         * gimple-match-head.c (gimple_simplify): Change prototypes of
19708         auto-generated functions to take a gimple_match_op * instead of
19709         separate code_helper * and tree * parameters.  Make the same
19710         change in the top-level overload and update calls to the
19711         gimple_resimplify routines.  Update calls to the auto-generated
19712         functions and to maybe_push_res_to_seq in the publicly-facing
19713         operation-specific gimple_simplify overloads.
19714         (gimple_match_op::MAX_NUM_OPS): Define.
19715         (gimple_resimplify1): Replace rcode and ops with a single res_op
19716         parameter.  Update call to gimple_simplify.
19717         (gimple_resimplify2): Likewise.
19718         (gimple_resimplify3): Likewise.
19719         (mprts_hook): Replace parameters with a gimple_match_op *.
19720         (maybe_build_generic_op): Likewise.
19721         (build_call_internal): Replace type, nargs and ops with
19722         a gimple_match_op *.
19723         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
19724         with a single gimple_match_op *.  Update calls to mprts_hook,
19725         build_call_internal and gimple_simplified_result_is_gimple_val.
19726         Factor out code that is common to the tree_code and combined_fn cases.
19727         * genmatch.c (expr::gen_transform): Replace tem_code and
19728         tem_ops with a gimple_match_op called tem_op.  Update calls
19729         to the gimple_resimplify functions and maybe_push_res_to_seq.
19730         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
19731         res_ops.  Update call to the gimple_resimplify functions.
19732         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
19733         (decision_tree::gen): Make the functions take a gimple_match_op *
19734         called res_op instead of separate res_code and res_ops parameters.
19735         Update call accordingly.
19736         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
19737         and ops with a single res_op parameter.  Update calls to
19738         maybe_build_generic_op and maybe_push_res_to_seq.
19739         (fold_stmt_1): Update calls to gimple_simplify and
19740         replace_stmt_with_simplification.
19741         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
19742         and gimple_simplified_result_is_gimple_val.
19743         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
19744         gimple_simplify.
19745         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
19746         with a gimple_match_op *.
19747         (vn_nary_build_or_lookup): Likewise.  Update call to
19748         vn_nary_build_or_lookup_1.
19749         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
19750         gimple_match_op *.  Update calls to the gimple_resimplify routines
19751         and to gimple_simplified_result_is_gimple_val.
19752         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
19753         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
19754         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
19755         (visit_nary_op): Likewise.
19756         (visit_reference_op_load): Likewise.
19758 2018-05-23  Luis Machado  <luis.machado@linaro.org>
19760         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
19761         modifier for printing the step amount.
19763 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
19765         PR target/78849
19766         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
19767         types.
19769 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
19771         * doc/sourcebuild.texi (Endianness): New subsubsection.
19773 2018-05-23  Luis Machado  <luis.machado@linaro.org>
19775         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19776         <prefetch_dynamic_strides>: New const bool field.
19777         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19778         prefetch_dynamic_strides.
19779         (exynosm1_prefetch_tune): Likewise.
19780         (thunderxt88_prefetch_tune): Likewise.
19781         (thunderx_prefetch_tune): Likewise.
19782         (thunderx2t99_prefetch_tune): Likewise.
19783         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
19784         false.
19785         (aarch64_override_options_internal): Update to set
19786         PARAM_PREFETCH_DYNAMIC_STRIDES.
19787         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19788         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19789         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19790         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19791         prefetch-dynamic-strides setting.
19793 2018-05-23  Luis Machado  <luis.machado@linaro.org>
19795         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19796         <minimum_stride>: New const int field.
19797         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19798         minimum_stride field defaulting to -1.
19799         (exynosm1_prefetch_tune): Likewise.
19800         (thunderxt88_prefetch_tune): Likewise.
19801         (thunderx_prefetch_tune): Likewise.
19802         (thunderx2t99_prefetch_tune): Likewise.
19803         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
19804         <default_opt_level>: Set to 3.
19805         (aarch64_override_options_internal): Update to set
19806         PARAM_PREFETCH_MINIMUM_STRIDE.
19807         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19808         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19809         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19810         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19811         stride is constant and is below the minimum stride threshold.
19813 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19815         * config/arm/arm-cpus.in (mode26): Delete.
19816         (armv4): Delete mode26 reference.
19817         * config/arm/arm.c (arm_configure_build_target): Delete use of
19818         isa_bit_mode26.
19820 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
19822         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
19823         New insn pattern.
19824         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
19825         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
19826         for non-SSE modes.
19827         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
19828         (floatunsdidf2): Ditto.
19830 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
19832         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
19833         (fixuns_trunc<mode>si2_avx512f): Ditto.
19834         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
19835         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
19836         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
19838 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
19840         PR rtl-optimization/79985
19841         * df-scan.c (df_insn_refs_collect): Remove special case for
19842         global registers and asm statements.
19844 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
19846         * extend.texi (Global Register Variables): Rewrite the bullet list.
19847         Note that the register is available for allocation. Note that access
19848         via inline asm must use constraints. Add note about async-signal
19849         handlers. Remove paragraph about automagic register selection.
19851 2018-05-23  Richard Biener  <rguenther@suse.de>
19853         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
19854         of fixed offset from memset VN.
19856 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
19858         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
19859         first_interp field.
19860         (alloc_cand_and_find_basis): Initialize first_interp field.
19861         (slsr_process_mul): Modify first_interp field.
19862         (slsr_process_add): Likewise.
19863         (slsr_process_cast): Modify first_interp field for each new
19864         interpretation.
19865         (slsr_process_copy): Likewise.
19866         (dump_candidate): Dump first_interp field.
19867         (replace_mult_candidate): Process all interpretations, not just
19868         subsequent ones.
19869         (replace_rhs_if_not_dup): Likewise.
19870         (replace_one_candidate): Likewise.
19872 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
19874         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
19875         Add new boolean.
19876         (aarch64_needs_frame_chain): New function.
19877         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
19879 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
19881         PR target/84882
19882         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
19883         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
19884         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
19885         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
19886         as true for strict-align.
19887         (aarch64_can_inline_p): Perform checks even when callee has no
19888         attributes to check for strict alignment.
19889         * doc/extend.texi (AArch64 Function Attributes): Document
19890         no-strict-align.
19891         * doc/invoke.texi: (AArch64 Options): Likewise.
19893 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
19895         PR tree-optimization/85853
19896         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
19897         the handling of the root of the node to...
19898         (vect_slp_analyze_node_operations_1): ...this new function,
19899         and run the whole thing with the child nodes' def types
19900         set according to their SLP node's def type.
19902 2018-05-23  Richard Biener  <rguenther@suse.de>
19904         PR middle-end/85874
19905         * tree-data-ref.c (create_runtime_alias_checks): Defer
19906         and ignore overflow warnings.
19908 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
19910         PR tree-optimization/85822
19911         * tree-vrp.c (is_masked_range_test): Fix handling of negative
19912         constants.
19914 2018-05-23  Richard Biener  <rguenther@suse.de>
19916         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
19917         memset constants via native_interpret_expr.
19919 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
19921         PR target/85345
19922         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
19923         attribute.
19924         (cgraph_node::create_alias): Likewise.
19925         (cgraph_node::get_availability): Check ifunc_resolver instead
19926         of looking up ifunc attribute.
19927         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
19928         * varasm.c (do_assemble_alias): Likewise.
19929         (assemble_alias): Likewise.
19930         (default_binds_local_p_3): Likewise.
19931         * cgraph.h (cgraph_node): Add ifunc_resolver.
19932         (cgraph_node::only_called_directly_or_aliased_p): Return false
19933         for IFUNC resolver.
19934         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
19935         attribute.
19936         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
19937         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
19938         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
19939         instead of looking up ifunc attribute.
19941 2018-05-22  Luis Machado  <luis.machado@linaro.org>
19943         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
19945 2018-05-22  Martin Sebor  <msebor@redhat.com>
19947         PR middle-end/85359
19948         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
19949         only when expasion succeeds.
19950         (expand_builtin_strcmp): Same.
19951         (expand_builtin_strncmp): Same.
19953 2018-05-22  Martin Sebor  <msebor@redhat.com>
19955         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
19957 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
19958             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19960         * config/aarch64/aarch64-ldpstp.md: Replace uses of
19961         aarch64_mem_pair_operand with memory_operand and delete operand swapping
19962         code.
19963         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
19964         Add check for legitimate_address.
19965         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
19966         (aarch64_swap_ldrstr_operands): New.
19967         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
19968         Define prototype.
19970 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
19971             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19973         * config/aarch64/aarch64.md: New patterns to generate stp
19974         and ldp.
19975         (store_pair_sw, store_pair_dw): New patterns to generate stp for
19976         single words and double words.
19977         (load_pair_sw, load_pair_dw): Likewise.
19978         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
19979         Delete.
19980         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
19981         Delete.
19982         * config/aarch64/aarch64-ldpstp.md: Modify peephole
19983         for different mode ldpstp and add peephole for merged zero stores.
19984         Likewise for loads.
19985         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
19986         Add size check.
19987         (aarch64_gen_store_pair): Rename calls to match new patterns.
19988         (aarch64_gen_load_pair): Rename calls to match new patterns.
19989         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
19990         (load_pair<DREG:mode><DREG2:mode>): ... This.
19991         (store_pair<mode>): Rename to...
19992         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
19993         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
19994         New mode iterators.
19995         (V_INT_EQUIV): Handle SImode.
19996         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
19997         New predicate.
19999 2018-05-22  Martin Sebor  <msebor@redhat.com>
20001         PR c/85623
20002         * calls.c (maybe_warn_nonstring_arg): Use string length to set
20003         or ajust the presumed bound on an operation to avoid unnecessary
20004         warnings.
20006 2018-05-22  Martin Sebor  <msebor@redhat.com>
20008         PR tree-optimization/85826
20009         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
20010         assuming that a DECL necesarily has a constant size.
20012 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
20014         PR middle-end/85862
20015         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
20017 2018-05-22  Richard Biener  <rguenther@suse.de>
20019         PR tree-optimization/85834
20020         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
20021         non-constant and non-zero memset arguments.
20023 2018-05-22  Martin Liska  <mliska@suse.cz>
20025         PR ipa/85607
20026         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
20028 2018-05-22  Richard Biener  <rguenther@suse.de>
20030         PR tree-optimization/85863
20031         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
20032         comparisons when vectype is specified.
20033         (vectorizable_condition): Do not specify vectype for
20034         vect_is_simple_cond when SLP vectorizing.
20036 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
20038         PR target/85657
20039         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
20040         define __ibm128 as long double.
20041         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
20042         as a distinct type when IEEE 128-bit support is enabled.
20043         (init_float128_ieee): Fix up conversions between IFmode and IEEE
20044         128-bit types to use the correct functions.
20045         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
20046         convert between 128-bit floating point types that have different
20047         modes but the same representation, instead of using gen_lowpart to
20048         makean alias.
20049         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
20050         KFmode.
20051         (IFKF_reg): New attributes to give the register constraints for
20052         IFmode and KFmode.
20053         (extend<mode>tf2_internal): New insns to mark an explicit
20054         conversion between 128-bit floating point types that have a
20055         different mode but share the same representation.
20057 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
20059         PR tree-optimization/85814
20060         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
20061         a null return from get_strinfo when unsharing the next
20062         strinfo in the chain.
20064 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
20066         PR gcc/84923
20067         * varasm.c (weak_finish): Clean up weak_decls.
20069 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20071         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
20072         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
20073         UNSPEC_UADALP values.
20074         * config/aarch64/iterators.md (ABAL): New int iterator.
20075         (ABDL2): Likewise.
20076         (ADALP): Likewise.
20077         (sur): Add mappings for the above.
20078         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
20079         New define_insn.
20080         (aarch64_<sur>abal<mode>_4): Likewise.
20081         (aarch64_<sur>adalp<mode>_3): Likewise.
20082         (<sur>sadv16qi): New define_expand.
20084 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
20086         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
20087         (*movdf_internal): Ditto.
20088         (*rcpsf2_sse): Ditto.
20089         (*rsqrtsf2_sse): Ditto.
20090         (*sqrt<mode>2_sse): Ditto.
20092 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
20094         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
20095         eor3q<mode>4.
20096         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
20097         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
20098         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
20099         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
20100         vbcaxq_s64): New.
20101         * config/aarch64/arm_neon.h: Likewise.
20102         * config/aarch64/iterators.md (VQ_I): New.
20104 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
20106         * config.gcc: Add arc/t-multilib-linux to tmake_file for
20107         arc*-*-linux*.
20108         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
20109         MULTILIB_DIRNAMES
20111 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
20113         * config/nds32/constraints.md (S): New constraint.
20114         * config/nds32/nds32.md (call_internal): Use constraint S.
20115         (call_value_internal): Likewise.
20116         (sibcall_internal): Likewise.
20117         (sibcall_value_internal): Likewise.
20119 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
20120             Chung-Ju Wu  <jasonwucj@gmail.com>
20122         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
20123         into consideration.
20125 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
20126             Chung-Ju Wu  <jasonwucj@gmail.com>
20128         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
20129         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
20130         (nds32_rtx_costs_impl): Simplify.
20131         (nds32_address_cost_impl): Simplify.
20132         (nds32_init_rtx_costs): New function.
20133         (nds32_rtx_costs_speed_prefer): Likewise.
20134         (nds32_rtx_costs_size_prefer): Likewise.
20135         (nds32_address_cost_speed_prefer): Likewise.
20136         (nds32_address_cost_speed_fwprop): Likewise.
20137         (nds32_address_cost_size_prefer): Likewise.
20138         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
20139         * config/nds32/nds32.c (nds32_option_override): Use
20140         nds32_init_rtx_costs function.
20142 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
20144         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
20145         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
20146         (TARGET_PIPELINE_N8): Likewise.
20147         (TARGET_PIPELINE_N10): Likewise.
20148         (TARGET_PIPELINE_N13): Likewise.
20149         (TARGET_PIPELINE_GRAYWOLF): Likewise.
20151 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
20153         * config/nds32/nds32-fpu.md: Update copyright year.
20155 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
20157         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
20159 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
20161         * config/nds32/nds32.c
20162         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
20163         * config/nds32/nds32.opt (minline-asm-r15): New option.
20165 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
20167         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
20168         MASK_HW_ABS.
20169         * config/nds32/nds32.md (abssi2): New pattern.
20171 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
20173         * config/i386/i386.md (rex64namesuffix): New mode attribute.
20174         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
20175         Merge insn pattern from sse_cvtsi2ss<round_name> and
20176         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
20177         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
20178         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
20179         using SWI48 mode iterator.
20180         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
20181         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
20182         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
20183         pattern from sse_cvttss2si<round_saeonly_name>
20184         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
20185         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
20186         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
20187         using SWI48 mode iterator.
20188         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
20189         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
20190         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
20191         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
20192         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
20193         using SWI48 mode iterator.
20194         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
20195         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
20196         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
20197         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
20198         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
20199         SWI48 mode iterator.
20200         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
20201         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
20202         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
20203         pattern from sse_cvttsd2si<round_saeonly_name>
20204         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
20206 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
20208         * config/nds32/nds32-md-auxiliary.c
20209         (nds32_valid_smw_lwm_base_p): Refine.
20210         (nds32_output_smw_single_word): Refine.
20211         (nds32_output_smw_double_word): New.
20212         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
20214 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
20216         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
20217         (nds32_output_stack_pop): Refine.
20218         (nds32_expand_unaligned_load): Refine.
20219         (nds32_expand_unaligned_store): Refine.
20221 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
20222             Chung-Ju Wu  <jasonwucj@gmail.com>
20224         * config/nds32/constants.md: Add TP_REGNUM constant.
20225         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
20226         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
20227         UNSPEC_ADD32.
20228         * config/nds32/nds32-doubleword.md: Consider flag_pic.
20229         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
20230         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
20231         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
20232         and PIC code generation.
20233         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
20234         code generation.
20235         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
20236         optimization.
20237         * config/nds32/nds32.md: Support TLS and PIC.
20238         * config/nds32/nds32.c: Support TLS and PIC.
20239         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
20240         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
20241         predicate.
20243 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
20245         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
20246         mode with E_ prefix.
20248 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
20249             Chung-Ju Wu  <jasonwucj@gmail.com>
20251         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
20252         * config/nds32/nds32-md-auxiliary.c
20253         (symbolic_reference_mentioned_p): New.
20254         (nds32_legitimize_ict_address): New.
20255         (nds32_expand_ict_move): New.
20256         (nds32_indirect_call_referenced_p): New.
20257         (nds32_symbol_binds_local_p): Delete.
20258         (nds32_long_call_p): Modify.
20259         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
20260         * config/nds32/nds32-protos.h
20261         (symbolic_reference_mentioned_p): Declare.
20262         (nds32_legitimize_ict_address): Declare.
20263         (nds32_expand_ict_move): Declare.
20264         (nds32_indirect_call_referenced_p): Declare.
20265         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
20266         (nds32_relax_group): Use nds32_ict_const_p as condition.
20267         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
20268         (nds32_asm_file_start): Output ict_model directive in asm code.
20269         (nds32_legitimate_address_p): Consider indirect call.
20270         (nds32_print_operand): Consider indirect call.
20271         (nds32_print_operand_address): Consider indirect call.
20272         (nds32_insert_attributes): Handle "indirect_call" attribute.
20273         (TARGET_LEGITIMATE_ADDRESS_P): Define.
20274         (TARGET_LEGITIMATE_CONSTANT_P): Define.
20275         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
20276         (TARGET_DELEGITIMIZE_ADDRESS): Define.
20277         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
20278         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
20279         (TARGET_ICT_MODEL_SMALL): Define.
20280         (TARGET_ICT_MODEL_LARGE): Define.
20281         * config/nds32/nds32.md (movsi): Consider ict model.
20282         (call, call_value): Consider ict model.
20283         (sibcall, sibcall_value): Consider ict model.
20284         * config/nds32/nds32.opt (mict-model): New option.
20285         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
20286         model.
20288 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
20289             Monk Chiang  <sh.chiang04@gmail.com>
20290             Jim Wilson <jimw@sifive.com>
20292         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
20293         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
20294         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
20295         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
20296         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
20297         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
20298         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
20299         compute save_libcall_adjustment properly.
20300         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
20301         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
20302         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
20303         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
20304         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
20305         (ABI_SPEC): Handle mabi=ilp32e.
20306         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
20307         (RVE): Add RVE mask.
20308         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
20309         <-march>: Add rv32e as an example.
20311 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
20313         PR c++/82899
20314         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
20315         (intra_create_variable_infos): Handle C++ constructors.
20317 2018-05-18  Martin Liska  <mliska@suse.cz>
20319         * passes.def: Remove a redundant pass.
20321 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
20323         PR bootstrap/85838
20324         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
20326 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20328         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
20329         (ARMv4): Update.
20330         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
20331         (ARMv6m): Update.
20332         (armv2, armv2a, armv3, armv3m): Delete architectures.
20333         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
20334         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
20335         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
20336         Delete cpus.
20337         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
20338         (*mulsidi3adddi): Likewise.
20339         (mulsidi3): Likewise.
20340         (*mulsidi3_nov6): Likewise.
20341         (umulsidi3): Likewise.
20342         (umulsidi3_nov6): Likewise.
20343         (umaddsidi4): Likewise.
20344         (*umulsidi3adddi): Likewise.
20345         (smulsi3_highpart): Likewise.
20346         (*smulsi3_highpart_nov6): Likewise.
20347         (umulsi3_highpart): Likewise.
20348         (*umulsi3_highpart_nov6): Likewise.
20349         * config/arm/arm.h (arm_arch3m): Delete.
20350         * config/arm/arm.c (arm_arch3m): Delete.
20351         (arm_option_override_internal): Update armv3-related comment.
20352         (arm_configure_build_target): Delete use of isa_bit_mode32.
20353         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
20354         (arm_rtx_costs_internal): Delete check of arm_arch3m.
20355         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
20356         (mulsa3): Likewise.
20357         (mulusa3): Likewise.
20358         * config/arm/arm-protos.h (arm_arch3m): Delete.
20359         * config/arm/arm-tables.opt: Regenerate.
20360         * config/arm/arm-tune.md: Likewise.
20361         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
20362         deleted architectures.
20364 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20366         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
20367         (armv5t, armv5te): New features.
20368         (ARMv5, ARMv5e): Delete fgroups.
20369         (ARMv5t, ARMv5te): Adjust for above changes.
20370         (ARMv6m): Likewise.
20371         (armv5, armv5e): Delete arches.
20372         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
20373         arm_arch5.
20374         (*call_reg_arm): Likewise.
20375         (*call_value_reg_armv5): Likewise.
20376         (*call_value_reg_arm): Likewise.
20377         (*call_symbol): Likewise.
20378         (*call_value_symbol): Likewise.
20379         (*sibcall_insn): Likewise.
20380         (*sibcall_value_insn): Likewise.
20381         (clzsi2): Likewise.
20382         (prefetch): Likewise.
20383         (define_split and define_peephole2 dependent on arm_arch5):
20384         Likewise.
20385         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
20386         arm_arch5e.
20387         (TARGET_ARM_QBIT): Likewise.
20388         (TARGET_DSP_MULTIPLY): Likewise.
20389         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
20390         (arm_arch5, arm_arch5e): Delete.
20391         (arm_arch5t, arm_arch5te): Declare.
20392         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
20393         (arm_arch5t): Declare.
20394         (arm_option_reconfigure_globals): Update for the above.
20395         (arm_options_perform_arch_sanity_checks): Update comment, replace
20396         use of arm_arch5 with arm_arch5t.
20397         (use_return_insn): Likewise.
20398         (arm_emit_call_insn): Likewise.
20399         (output_return_instruction): Likewise.
20400         (arm_final_prescan_insn): Likewise.
20401         (arm_coproc_builtin_available): Likewise.
20402         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
20403         arm_arch5e with arm_arch5t and arm_arch5te.
20404         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
20405         (arm_arch5t, arm_arch5te): Declare.
20406         * config/arm/arm-tables.opt: Regenerate.
20407         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
20408         * config/arm/t-multilib: Likewise.
20409         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
20410         instead of arm_arch5.
20411         (*call_reg_thumb1): Likewise.
20412         (*call_value_reg_thumb1_v5): Likewise.
20413         (*call_value_reg_thumb1): Likewise.
20414         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
20415         unreachable path.
20416         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
20418 2018-05-18  Martin Liska  <mliska@suse.cz>
20420         PR gcov-profile/84846
20421         * doc/gcov.texi: Document -t option of gcov tool.
20423 2018-05-18  Martin Liska  <mliska@suse.cz>
20425         PR gcov-profile/84846
20426         * gcov.c (print_usage): Add new -t option.
20427         (process_args): Handle the option.
20428         (generate_results): Use stdout as output when requested by
20429         the option.
20431 2018-05-18  Martin Liska  <mliska@suse.cz>
20433         PR gcov-profile/84846
20434         * coverage.c (coverage_init): Write PWD to .gcno file.
20435         * doc/gcov.texi: Document how working directory is printed.
20436         * gcov-dump.c (dump_gcov_file): Print PWD.
20437         * gcov.c (output_intermediate_file): Likewise.
20438         (read_graph_file): Read PWD string.
20439         (output_lines): Print PWD.
20441 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20443         PR middle-end/85817
20444         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
20445         for retval and return false if all args to phi are zero.
20447 2018-05-18  Richard Biener  <rguenther@suse.de>
20449         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
20450         method.
20451         (evrp_dom_walker::before_dom_children): Call it.
20453 2018-05-18  Richard Biener  <rguenther@suse.de>
20455         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
20456         results when processing array refs with variable index.
20458 2018-05-18  Toon Moene  <toon@moene.org>
20460         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
20461         directly after that of -floop-interchange. Indicate that both
20462         options are enabled by default when specifying -O3.
20464 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20466         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
20467         iterator.  Delete separate integer-mode vec_set<mode> expander.
20468         (aarch64_simd_vec_setv2di): Delete.
20469         (vec_setv2di): Delete.
20470         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
20471         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
20472         the "w, r" alternative.
20474 2018-05-18  Martin Liska  <mliska@suse.cz>
20476         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
20477         * tree-pass.h (make_pass_lower_switch_O0): New function.
20478         * tree-switch-conversion.c (node_has_low_bound): Remove.
20479         (node_has_high_bound): Likewise.
20480         (node_is_bounded): Likewise.
20481         (class pass_lower_switch): Make it a template type and create
20482         two instances.
20483         (pass_lower_switch::execute): Add template argument.
20484         (make_pass_lower_switch): New function.
20485         (make_pass_lower_switch_O0): New function.
20486         (do_jump_if_equal): Remove.
20487         (emit_case_nodes): Simplify to just handle all 3 cases and leave
20488         all the hard work to tree optimization passes.
20490 2018-05-18  Martin Liska  <mliska@suse.cz>
20492         * dbgcnt.c (limit_low): Renamed from limit.
20493         (limit_high): New variable.
20494         (dbg_cnt_is_enabled): Check for upper limit.
20495         (dbg_cnt): Adjust dumping.
20496         (dbg_cnt_set_limit_by_index): Add new argument for high
20497         value.
20498         (dbg_cnt_set_limit_by_name): Likewise.
20499         (dbg_cnt_process_single_pair): Parse new format.
20500         (dbg_cnt_process_opt): Use strtok.
20501         (dbg_cnt_list_all_counters): Remove 'value' and add
20502         'limit_high'.
20503         * doc/invoke.texi: Document changes.
20505 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
20507         * doc/sourcebuild.texi (scalar_all_fma): Document.
20508         * tree.def (FMA_EXPR): Delete.
20509         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
20510         * internal-fn.c (ternary_direct): New macro.
20511         (expand_ternary_optab_fn): Likewise.
20512         (direct_ternary_optab_supported_p): Likewise.
20513         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
20514         * builtins.c (fold_builtin_fma): Delete.
20515         (fold_builtin_3): Don't call it.
20516         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
20517         * expr.c (expand_expr_real_2): Likewise.
20518         * fold-const.c (operand_equal_p): Likewise.
20519         (fold_ternary_loc): Likewise.
20520         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
20521         * gimple.c (DEFTREECODE): Likewise.
20522         * gimplify.c (gimplify_expr): Likewise.
20523         * optabs-tree.c (optab_for_tree_code): Likewise.
20524         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
20525         * tree-eh.c (operation_could_trap_p): Likewise.
20526         (stmt_could_throw_1_p): Likewise.
20527         * tree-inline.c (estimate_operator_cost): Likewise.
20528         * tree-pretty-print.c (dump_generic_node): Likewise.
20529         (op_code_prio): Likewise.
20530         * tree-ssa-loop-im.c (stmt_cost): Likewise.
20531         * tree-ssa-operands.c (get_expr_operands): Likewise.
20532         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
20533         * fold-const-call.h (fold_fma): Delete.
20534         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
20535         CFN_FNMA and CFN_FNMS.
20536         (fold_fma): Delete.
20537         * genmatch.c (combined_fn): New enum.
20538         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
20539         (commutative_op): New function.
20540         (commutate): Use it.  Handle more than 2 operands.
20541         (dt_operand::gen_gimple_expr): Use commutative_op.
20542         (parser::parse_expr): Allow :c to be used with non-binary
20543         operators if the commutative operand is known.
20544         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
20545         CFN_FMS, CFN_FNMA and CFN_FNMS.
20546         (backprop::process_assign_use): Remove FMA_EXPR handling.
20547         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
20548         (gen_hsa_fma): New function.
20549         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
20550         IFN_FNMA and IFN_FNMS.
20551         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
20552         * gimple-fold.h (follow_all_ssa_edges): Declare.
20553         * gimple-fold.c (follow_all_ssa_edges): New function.
20554         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
20555         gimple_build interface and use follow_all_ssa_edges to fold the result.
20556         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
20557         instead of checking for optabs directly.
20558         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
20559         rather than FMA_EXPRs.
20560         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
20561         call to IFN_FMA instead of an FMA_EXPR.
20563 2018-05-17  Jim Wilson  <jimw@sifive.com>
20565         * expr.c (do_tablejump): When converting index to Pmode, if we have a
20566         sign extended promoted subreg, and the range does not have the sign bit
20567         set, then do a sign extend.
20569         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
20570         test, check for sign extended subreg and/or constant operands, and
20571         do a sign extend in that case.
20573 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
20575         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
20576         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
20577         Add untyped.
20578         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
20579         Change logics_shift_reg to logics_shift_imm.
20580         (thunderx2t99_fp_loadpair_basic): Delete.
20581         (thunderx2t99_fp_storepair_basic): Delete.
20582         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
20583         (thunderx2t99_asimd_polynomial): Delete.
20584         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
20585         and neon_fp_mul_d_scalar_q.
20586         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
20587         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
20588         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
20589         (thunderx2t99_asimd_lut): Add missing tbl types.
20590         (thunderx2t99_asimd_ext): Delete.
20591         (thunderx2t99_asimd_load1_1_mult): Delete.
20592         (thunderx2t99_asimd_load1_2_mult): Delete.
20593         (thunderx2t99_asimd_load1_ldp): New.
20594         (thunderx2t99_asimd_load1): New.
20595         (thunderx2t99_asimd_load2): Add missing *load2* types.
20596         (thunderx2t99_asimd_load3): New.
20597         (thunderx2t99_asimd_load4): New.
20598         (thunderx2t99_asimd_store1_1_mult): Delete.
20599         (thunderx2t99_asimd_store1_2_mult): Delete.
20600         (thunderx2t99_asimd_store2_mult): Delete.
20601         (thunderx2t99_asimd_store2_onelane): Delete.
20602         (thunderx2t99_asimd_store_stp): New.
20603         (thunderx2t99_asimd_store1): New.
20604         (thunderx2t99_asimd_store2): New.
20605         (thunderx2t99_asimd_store3): New.
20606         (thunderx2t99_asimd_store4): New.
20608 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
20610         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
20611         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
20613 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
20614             Segher Boessenkool  <segher@kernel.crashing.org>
20616         PR target/85698
20617         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
20618         operand.
20620 2018-05-17  Richard Biener  <rguenther@suse.de>
20622         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
20623         for pruning loop and prune defs feeding only already visited PHIs.
20625 2018-05-17  Richard Biener  <rguenther@suse.de>
20627         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
20629 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
20630             Richard Biener  <rguenther@suse.de>
20632         PR tree-optimization/85793
20633         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
20634         for VMAT_ELEMENTWISE.
20636 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
20638         * internal-fn.h (lookup_internal_fn): Declare
20639         * internal-fn.c (lookup_internal_fn): New function.
20640         * gimple.c (gimple_build_call_from_tree): Handle calls to
20641         internal functions.
20642         * gimple-pretty-print.c (dump_gimple_call): Print "." before
20643         internal function names.
20644         * tree-pretty-print.c (dump_generic_node): Likewise.
20645         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
20647 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
20649         * gimple-fold.h (gimple_build): Make the function forms take
20650         combined_fn rather than built_in_function.
20651         (gimple_simplify): Likewise.
20652         * gimple-match-head.c (gimple_simplify): Likewise.
20653         * gimple-fold.c (gimple_build): Likewise.
20654         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
20655         rather than gimple_build_call_internal.
20656         (get_initial_defs_for_reduction): Likewise.
20657         (vect_create_epilog_for_reduction): Likewise.
20658         (vectorizable_live_operation): Likewise.
20660 2018-05-17  Martin Liska  <mliska@suse.cz>
20662         * gimple-ssa-sprintf.c (format_directive): Do not use
20663         space in between 'G_' and '('.
20665 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
20667         PR target/85323
20668         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
20669         even if the mask is not all ones.
20671         PR target/85323
20672         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
20673         vector.
20674         (ix86_gimple_fold_builtin): Likewise.
20676         PR target/85323
20677         * config/i386/i386.c: Include tree-vector-builder.h.
20678         (ix86_vector_shift_count): New function.
20679         (ix86_fold_builtin): Fold shift builtins by scalar count.
20680         (ix86_gimple_fold_builtin): Likewise.
20682         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
20683         _mm512_setzero): New intrinsics.
20685 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
20686             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20688         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
20689         code generation for cases where splatting a value is not useful.
20690         * simplify-rtx.c (simplify_ternary_operation): Simplify
20691         vec_merge across a vec_duplicate and a paradoxical subreg forming
20692         a vector mode to a vec_concat.
20694 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
20696         * config.gcc: Support "goldmont-plus".
20697         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
20698         "goldmont-plus".
20699         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20700         PROCESSOR_GOLDMONT_PLUS.
20701         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
20702         (processor_target_table): Add "goldmont-plus".
20703         (PTA_GOLDMONT_PLUS): Define.
20704         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
20705         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
20706         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
20707         (fold_builtin_cpu): Add "goldmont-plus".
20708         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
20709         (ix86_option_override_internal): Add "goldmont-plus".
20710         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
20711         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
20712         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
20713         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
20715 2018-05-17  Richard Biener  <rguenther@suse.de>
20717         PR tree-optimization/85757
20718         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
20719         remove defs that only feed that PHI from further processing.
20721 2018-05-16  Jim Wilson  <jimw@sifive.com>
20723         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
20724         asterisk to name.
20725         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
20727 2018-05-16  Mark Wielaard  <mark@klomp.org>
20729         * dwarf2out.c (count_index_strings): New function.
20730         (output_indirect_strings): Call count_index_strings and generate
20731         header for dwarf_version >= 5.
20733 2018-05-16  Mark Wielaard  <mark@klomp.org>
20735         * dwarf2out.c (dwarf_FORM): New function.
20736         (set_indirect_string): Use dwarf_FORM.
20737         (reset_indirect_string): Likewise.
20738         (size_of_die): Likewise.
20739         (value_format): Likewise.
20740         (output_die): Likewise.
20741         (add_skeleton_AT_string): Likewise.
20742         (output_macinfo_op): Likewise.
20743         (index_string): Likewise.
20744         (output_index_string_offset): Likewise.
20745         (output_index_string): Likewise.
20746         (count_index_strings): Likewise.
20748 2018-05-16  Carl Love  <cel@us.ibm.com>
20750         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
20751         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
20753 2018-05-16  Martin Jambor  <mjambor@suse.cz>
20755         * ipa-prop.c (ipa_free_all_edge_args): Remove.
20756         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
20758 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
20760         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
20761         (fnma<mode>4): Likewise.
20762         (fms<mode>4): Likewise.
20763         (fnms<mode>4): Likewise.
20764         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
20765         (aarch64_fnma<mode>4): Likewise.
20766         (aarch64_fms<mode>4): Likewise.
20767         (aarch64_fnms<mode>4): Likewise.
20768         (aarch64_fnmadd<mode>4): Likewise.
20770 2018-05-16  Jason Merrill  <jason@redhat.com>
20772         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
20774 2018-05-16  Richard Biener  <rguenther@suse.de>
20776         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
20777         (dump_stmt_cost): Declare.
20778         (add_stmt_cost): Dump cost we add.
20779         (add_stmt_costs): New function.
20780         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
20781         No longer exported.
20782         (vect_analyze_stmt): Adjust prototype.
20783         (vectorizable_condition): Likewise.
20784         (vectorizable_live_operation): Likewise.
20785         (vectorizable_reduction): Likewise.
20786         (vectorizable_induction): Likewise.
20787         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
20788         cost vector to pass to vectorizable_ and record afterwards.
20789         (vect_model_reduction_cost): Take cost vector argument and adjust.
20790         (vect_model_induction_cost): Likewise.
20791         (vectorizable_reduction): Likewise.
20792         (vectorizable_induction): Likewise.
20793         (vectorizable_live_operation): Likewise.
20794         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
20795         SLP_TREE_NUMBER_OF_VEC_STMTS.
20796         (vect_analyze_slp_cost_1): Remove.
20797         (vect_analyze_slp_cost): Likewise.
20798         (vect_slp_analyze_node_operations): Take visited args and
20799         a target cost vector.  Avoid processing already visited stmt sets.
20800         (vect_slp_analyze_operations): Use a local cost vector to gather
20801         costs and register those of non-discarded instances.
20802         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
20803         (vect_schedule_slp_instance): Remove copying of
20804         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
20805         zero.
20806         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
20807         adding cost.  Record cost entry location.
20808         (vect_prologue_cost_for_slp_op): Function to compute cost of
20809         a constant or invariant generated for SLP vect in the prologue,
20810         split out from vect_analyze_slp_cost_1.
20811         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
20812         (vect_model_promotion_demotion_cost): Likewise.
20813         (vect_model_store_cost): Likewise, make static.
20814         (vect_model_load_cost): Likewise.
20815         (vectorizable_bswap): Add cost vector arg and adjust.
20816         (vectorizable_call): Likewise.
20817         (vectorizable_simd_clone_call): Likewise.
20818         (vectorizable_conversion): Likewise.
20819         (vectorizable_assignment): Likewise.
20820         (vectorizable_shift): Likewise.
20821         (vectorizable_operation): Likewise.
20822         (vectorizable_store): Likewise.
20823         (vectorizable_load): Likewise.
20824         (vectorizable_condition): Likewise.
20825         (vectorizable_comparison): Likewise.
20826         (can_vectorize_live_stmts): Likewise.
20827         (vect_analyze_stmt): Likewise.
20828         (vect_transform_stmt): Adjust calls to vectorizable_*.
20829         * tree-vectorizer.c: Include gimple-pretty-print.h.
20830         (dump_stmt_cost): New function.
20832 2018-05-16  Richard Biener  <rguenther@suse.de>
20834         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
20835         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
20836         * tree-ssa-dse.c: Include tree-ssa-loop.h.
20837         (check_name): New callback.
20838         (dse_classify_store): Track cycles via a visited bitmap of PHI
20839         defs and simplify handling of in-loop and across loop dead stores
20840         and properly fail for loop-variant refs.  Handle byte-tracking with
20841         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
20842         limiting the walk.
20844 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
20846         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
20847         (vect_get_mask_type_for_stmt): Likewise.
20848         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
20849         split out from...
20850         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
20851         to determine the statement's vector type and the vector type that
20852         should be used for calculating nunits.  Deal with cases in which
20853         the type has to be deferred.
20854         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
20855         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
20856         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
20857         (vect_determine_vf_for_stmt): New functions, split out from...
20858         (vect_determine_vectorization_factor): ...here.
20859         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
20860         (vect_get_mask_type_for_stmt): New functions, split out from
20861         vect_determine_vectorization_factor.
20863 2018-05-16  Richard Biener  <rguenther@suse.de>
20865         * tree-cfg.c (verify_gimple_assign_ternary): Properly
20866         verify the [VEC_]COND_EXPR embedded comparison.
20868 2018-05-15  Martin Sebor  <msebor@redhat.com>
20870         PR tree-optimization/85753
20871         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
20872         RECORD_TYPE in addition to ARRAY_TYPE.
20874 2018-05-15  Martin Sebor  <msebor@redhat.com>
20876         PR middle-end/85643
20877         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
20879 2018-05-15  Richard Biener  <rguenther@suse.de>
20881         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
20882         add by_clobber_p one.  Change algorithm to collect all defs
20883         representing uses we need to walk and try reducing them to
20884         a single one before failing.
20885         (dse_dom_walker::dse_optimize_stmt): Adjust.
20887 2018-05-13  Mark Wielaard  <mark@klomp.org>
20889         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
20890         (size_of_loc_descr): Likewise.
20891         (output_loc_operands): Likewise.
20892         (output_loc_operands_raw): Likewise.
20893         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
20894         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
20895         (hash_loc_operands): Likewise.
20896         (compare_loc_operands): Likewise.
20898 2018-05-14  Mark Wielaard  <mark@klomp.org>
20900         * dwarf2out.c (count_index_addrs): New function.
20901         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
20903 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20905         PR tree-optimization/83648
20906         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
20907         return value as malloc candidate.
20909 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20911         PR ipa/85734
20912         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
20913         param as true in call to suggest_attribute.
20915 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
20917         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
20918         -mreadonly-in-sdata.
20920 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20922         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
20923         New pattern.
20924         (aarch64_crypto_aesd_fused): Likewise.
20926 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
20928         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
20929         (movsi_aarch64): Likewise.
20930         (load_pairsi): Likewise.
20931         (load_pairdi): Likewise.
20932         (store_pairsi): Likewise.
20933         (store_pairdi): Likewise.
20934         (load_pairsf): Likewise.
20935         (load_pairdf): Likewise.
20936         (store_pairsf): Likewise.
20937         (store_pairdf): Likewise.
20938         (zero_extend): Likewise.
20939         (trunc): Swap alternatives.
20940         (fcvt_target): Add '?' to prefer w over r.
20942 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
20944         PR target/85756
20945         * config/i386/i386.md: Disallow non-commutative arithmetics in
20946         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
20947         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
20948         in the peephole2 before it.
20950 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
20952         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
20953         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
20954         (ix86_handle_option): Handle -mcldemote.
20955         * config.gcc: New header.
20956         * config/i386/cldemoteintrin.h: New file.
20957         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
20958         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
20959         -mcldemote.
20960         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20961         OPTION_MASK_ISA_CLDEMOTE.
20962         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
20963         (ix86_valid_target_attribute_inner_p): Ditto.
20964         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
20965         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
20966         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
20967         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
20968         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
20969         (cldemote): New.
20970         * config/i386/i386.opt: Add -mcldemote.
20971         * config/i386/x86intrin.h: New header.
20972         * doc/invoke.texi: Add -mcldemote.
20974 2018-05-14  Richard Biener  <rguenther@suse.de>
20976         * doc/match-and-simplify.texi: Adjust :s documentation.
20978 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
20980         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
20981         intended memcpy size.
20982         (REORDER_45): Likewise.
20984 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
20986         * sort.cc: New file.
20987         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
20988         * vec.c (qsort_chk): Use gcc_qsort.
20989         * Makefile.in (OBJS-libcommon): Add sort.o.
20990         (build/sort.o): New target.  Use it...
20991         (BUILD_RTL): ... here, and...
20992         (build/gencfn-macros): ... here, and...
20993         (build/genmatch): ... here.
20995 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
20996             Chung-Ju Wu  <jasonwucj@gmail.com>
20998         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
20999         * config/nds32/nds32-graywolf.md: New file.
21000         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
21001         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
21002         pipeline.
21003         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
21004         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
21005         * config/nds32/nds32.md (pipeline_model): Add graywolf.
21006         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
21007         * config/nds32/pipelines.md: Include n15 settings.
21009 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
21010             Chung-Ju Wu  <jasonwucj@gmail.com>
21012         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
21013         * config/nds32/nds32-n13.md: New file.
21014         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
21015         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
21016         pipeline.
21017         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
21018         * config/nds32/nds32.md (pipeline_model): Add n13.
21019         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
21020         * config/nds32/pipelines.md: Include n13 settings.
21022 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
21023             Chung-Ju Wu  <jasonwucj@gmail.com>
21025         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
21026         * config/nds32/nds32-n10.md: New file.
21027         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
21028         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
21029         pipeline.
21030         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
21031         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
21032         * config/nds32/nds32.md (pipeline_model): Add n10.
21033         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
21034         * config/nds32/pipelines.md: Include n10 settings.
21036 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
21037             Kito Cheng  <kito.cheng@gmail.com>
21038             Chung-Ju Wu  <jasonwucj@gmail.com>
21040         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
21041         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
21042         Add enum values for DSP extension instructions.
21043         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
21044         New constraints.
21045         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
21046         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
21047         New code iterators.
21048         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
21049         * config/nds32/nds32-dspext.md: New file for DSP implementation.
21050         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
21051         * config/nds32/nds32-intrinsic.md: Likewise.
21052         * config/nds32/nds32_intrinsic.h: Likewise.
21053         * config/nds32/nds32-md-auxiliary.c: Likewise.
21054         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
21055         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
21056         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
21057         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
21058         * config/nds32/nds32-protos.h: New declarations for DSP extension.
21059         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
21060         TYPE_DMAC in switch statement.
21061         * config/nds32/nds32.c: New checking and implementation for DSP
21062         extension instructions.
21063         * config/nds32/nds32.h: Likewise.
21064         * config/nds32/nds32.md: Likewise.
21065         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
21066         * config/nds32/predicates.md: Implement new predicates for DSP
21067         extension.
21069 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
21071         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
21072         Reformat alternatives and attributes so it is easier to identify
21073         which constraints/attributes go with which instruction.
21074         (mov<mode>_hardfloat32, FMOVE64): Likewise.
21075         (mov<mode>_softfloat32, FMOVE64): Likewise.
21076         (mov<mode>_hardfloat64, FMOVE64): Likewise.
21077         (mov<mode>_softfloat64, FMOVE64): Likewise.
21079 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21081         * doc/extend.texi (PowerPC Built-in Functions): Rename this
21082         subsection.
21083         (Basic PowerPC Built-in Functions): The new name of the
21084         subsection previously known as "PowerPC Built-in Functions".
21085         (Basic PowerPC Built-in Functions Available on all Configurations):
21086         New subsubsection.
21087         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
21088         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
21089         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
21090         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
21092 2018-05-11  Martin Jambor  <mjambor@suse.cz>
21094         PR ipa/85655
21095         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
21096         single const.
21098 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
21100         PR target/85733
21101         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
21103 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
21105         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
21106         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
21107         (ix86_handle_option): Handle -mwaitpkg.
21108         * config.gcc: New header.
21109         * config/i386/cpuid.h (bit_WAITPKG): New bit.
21110         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
21111         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
21112         function type.
21113         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21114         OPTION_MASK_ISA_WAITPKG.
21115         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
21116         (ix86_option_override_internal): Add PTA_WAITPKG.
21117         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
21118         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
21119         IX86_BUILTIN_TPAUSE.
21120         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
21121         __builtin_ia32_umwait and __builtin_ia32_tpause.
21122         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
21123         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
21124         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
21125         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
21126         UNSPECV_TPAUSE): New.
21127         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
21128         * config/i386/i386.opt: Add -mwaitpkg.
21129         * config/i386/waitpkgintrin.h: New file.
21130         * config/i386/x86intrin.h: New header.
21131         * doc/invoke.texi: Add -mwaitpkg.
21133 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
21135         PR target/85606
21136         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
21137         equivalent.
21138         (cortex-m0): Use armv6s-m isa.
21139         (cortex-m0plus): Likewise.
21140         (cortex-m1): Likewise.
21141         (cortex-m0.small-multiply): Likewise.
21142         (cortex-m0plus.small-multiply): Likewise.
21143         (cortex-m1.small-multiply): Likewise.
21145 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
21146             Jakub Jelinek  <jakub@redhat.com>
21148         PR tree-optimization/85692
21149         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
21150         source permute as well.
21152 2018-05-11  Martin Liska  <mliska@suse.cz>
21154         PR sanitizer/85556
21155         * doc/extend.texi: Document LLVM style format for no_sanitize
21156         attribute.
21158 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
21160         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
21161         mode_supports_vsx_dform_quad to mode_supports_dq_form.
21162         (mode_supports_vsx_dform_quad): Likewise.
21163         (mode_supports_vmx_dform): Move these functions to be next to the
21164         other mode_supports functions.
21165         (mode_supports_dq_form): Likewise.
21166         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
21167         mode_supports_dq_form.
21168         (reg_offset_addressing_ok_p): Likewise.
21169         (offsettable_ok_by_alignment): Likewise.
21170         (rs6000_legitimate_offset_address_p): Likewise.
21171         (legitimate_lo_sum_address_p): Likewise.
21172         (rs6000_legitimize_address): Likewise.
21173         (rs6000_legitimize_reload_address): Likewise.
21174         (rs6000_secondary_reload_inner): Likewise.
21175         (rs6000_preferred_reload_class): Likewise.
21176         (rs6000_output_move_128bit): Likewise.
21178 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
21180         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
21181         Generate SImode target register for null target.
21182         <case IX86_BUILTIN_XGETBV>: Ditto.
21183         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
21184         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
21186 2018-05-10  Carl Love  <cel@us.ibm.com>
21188         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
21189         dcbtt and dcbtstt if operands[2] is 0.
21191 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
21193         PR target/85693
21194         * config/i386/sse.md (usadv64qi): New expander.
21196 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
21198         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
21199         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
21200         -maltivec=be support.
21201         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
21202         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
21203         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
21204         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
21205         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
21206         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
21207         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
21208         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
21209         altivec_vsumsws): Adjust.
21210         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
21211         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
21212         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
21213         support.
21214         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
21215         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
21216         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
21217         (altivec_lve<VI_char>x): Delete expand.
21218         (*altivec_lve<VI_char>x_internal): Rename to...
21219         (altivec_lve<VI_char>x): ... this.
21220         (altivec_lvxl_<mode>): Delete expand.
21221         (*altivec_lvxl_<mode>_internal): Rename to ...
21222         (altivec_lvxl_<mode>): ... this.
21223         (altivec_stvxl_<mode>): Delete expand.
21224         (*altivec_stvxl_<mode>_internal): Rename to ...
21225         (altivec_stvxl_<mode>): ... this.
21226         (altivec_stve<VI_char>x): Delete expand.
21227         (*altivec_stve<VI_char>x_internal): Rename to ...
21228         (altivec_stve<VI_char>x): ... this.
21229         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
21230         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
21231         reduc_plus_scal_<mode>): Adjust.
21232         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
21233         comment.
21234         (rs6000_cpu_cpp_builtins): Adjust.
21235         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
21236         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
21237         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
21238         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
21239         -maltivec=be support.
21240         (rs6000_split_vec_extract_var): Adjust.
21241         (rs6000_split_v4si_init): Adjust.
21242         (swap_selector_for_mode): Delete.
21243         (altivec_expand_lvx_be, altivec_expand_stvx_be,
21244         altivec_expand_stvex_be): Delete.
21245         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
21246         -maltivec=be support.
21247         (rs6000_gimple_fold_builtin): Ditto.
21248         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
21249         Adjust.
21250         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
21251         (TARGET_DIRECT_MOVE_64BIT): Adjust.
21252         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
21253         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
21254         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
21255         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
21256         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
21257         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
21258         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
21259         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
21260         anonymous split): Adjust.
21261         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
21262         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
21264 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
21266         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
21267         when --with-gxx-include-dir is also specified.
21268         * configure: Regenerate.
21270 2018-05-09  Jim Wilson  <jimw@sifive.com>
21272         PR target/84797
21273         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
21274         * config/riscv/t-withmultilib: New.
21275         * config/riscv/withmultilib.h: New.
21276         * doc/install.texi: Document RISC-V --with-multilib-list support.
21278 2018-05-09  Richard Biener  <rguenther@suse.de>
21280         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
21281         vector.
21282         (vect_bb_vectorization_profitable_p): Adjust.  Compute
21283         actual scalar cost using the cost vector and the add_stmt_cost
21284         machinery.
21286 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
21288         PR rtl-optimization/85645
21289         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
21290         in the REG_CFA_REGISTER note for LR, don't leave it empty.
21292 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
21294         PR rtl-optimization/85645
21295         * shrink-wrap.c (spread_components): Return a boolean saying if
21296         anything was changed.
21297         (try_shrink_wrapping_separate): Iterate spread_components until
21298         nothing changes anymore.
21300 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
21302         PR rtl-optimization/85645
21303         * regrename.c (build_def_use): Also kill the chains that include the
21304         destination of a REG_CFA_REGISTER note.
21306 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
21308         PR rtl-optimization/85645
21309         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
21310         insn that has a REG_CFA_REGISTER note.
21312 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
21314         * cfgexpand.c (expand_clobber): New function.
21315         (expand_gimple_stmt_1): Use it.
21316         * tree-vect-stmts.c (vect_clobber_variable): New function,
21317         split out from...
21318         (vectorizable_simd_clone_call): ...here.
21319         (vectorizable_store): Emit a clobber either side of an
21320         IFN_STORE_LANES sequence.
21321         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
21323 2018-05-09  Tom de Vries  <tom@codesourcery.com>
21325         PR target/85626
21326         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
21327         (define_insn "trap_if_false"): Add exit after trap.
21329 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
21331         PR rtl-optimization/85638
21332         * bb-reorder.c: Include common/common-target.h.
21333         (create_forwarder_block): New function extracted from...
21334         (fix_up_crossing_landing_pad): ...here.  Rename into...
21335         (dw2_fix_up_crossing_landing_pad): ...this.
21336         (sjlj_fix_up_crossing_landing_pad): New function.
21337         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
21338         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
21339         from both partitions and exit the loop after one iteration.
21341 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21343         Revert:
21344         * doc/extend.texi (PowerPC Built-in Functions): Rename this
21345         subsection.
21346         (Basic PowerPC Built-in Functions): The new name of the
21347         subsection previously known as "PowerPC Built-in Functions".
21348         (Basic PowerPC Built-in Functions Available on all Configurations):
21349         New subsubsection.
21350         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
21351         subsubsection.
21352         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
21353         subsubsection.
21354         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
21355         subsubsection.
21356         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
21357         subsubsection.
21359 2018-05-08  Jim Wilson  <jimw@sifive.com>
21361         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
21362         (LD_EMUL_SUFFIX): New.
21363         (LINK_SPEC): Use it.
21365 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21367         * doc/extend.texi (PowerPC Built-in Functions): Rename this
21368         subsection.
21369         (Basic PowerPC Built-in Functions): The new name of the
21370         subsection previously known as "PowerPC Built-in Functions".
21371         (Basic PowerPC Built-in Functions Available on all Configurations):
21372         New subsubsection.
21373         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
21374         subsubsection.
21375         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
21376         subsubsection.
21377         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
21378         subsubsection.
21379         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
21380         subsubsection.
21382 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
21384         PR target/85683
21385         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
21386         after cmpelim optimization.
21388 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
21390         * config.gcc: Support "goldmont".
21391         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
21392         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21393         PROCESSOR_GOLDMONT.
21394         * config/i386/i386.c (m_GOLDMONT): Define.
21395         (processor_target_table): Add "goldmont".
21396         (PTA_GOLDMONT): Define.
21397         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
21398         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
21399         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
21400         (fold_builtin_cpu): Add "goldmont".
21401         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
21402         (ix86_option_override_internal): Add "goldmont".
21403         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
21404         (processor_type): Add PROCESSOR_GOLDMONT.
21405         * config/i386/i386.md: Add CPU "glm".
21406         * config/i386/glm.md: New file.
21407         * config/i386/x86-tune.def: Add m_GOLDMONT.
21408         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
21410 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
21412         PR target/85572
21413         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
21414         E_V4DImode.
21415         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
21416         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
21417         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
21419         PR target/85317
21420         * config/i386/i386.c (ix86_fold_builtin): Handle
21421         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
21423         PR target/85480
21424         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
21425         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
21427 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
21429         PR target/85658
21430         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
21431         (check_arch): Likewise.
21432         (check_fpu): Return the result rather than printing it.
21433         (end arch): Fix operator precedence.
21434         (end cpu): Likewise.
21435         (END): Print the result from check_fpu.
21437 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
21438             Alan Hayward  <alan.hayward@arm.com>
21439             David Sherwood  <david.sherwood@arm.com>
21441         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
21442         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
21443         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
21444         (*fcmuo<mode>_and): New patterns.
21446 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
21448         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
21449         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
21450         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
21451         (cmp_op, sve_imm_con): New code attributes.
21452         (SVE_COND_INT_CMP, imm_con): Delete.
21453         (cmp_op): Remove above unspecs from int attribute.
21454         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
21455         to...
21456         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
21457         comparison-specific unspecs.
21458         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
21459         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
21460         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
21461         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
21462         (*vec_fcm<cmp_op><mode>): Rename to...
21463         (*fcm<cmp_op><mode>): ...this and adjust likewise.
21464         (*vec_fcmuo<mode>): Rename to...
21465         (*fcmuo<mode>): ...this and adjust likewise.
21466         (*pred_fcm<cmp_op><mode>): New pattern.
21467         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
21468         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
21469         functions.
21470         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
21471         and UNORDERED.
21472         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
21473         (aarch64_emit_sve_predicated_cond): New function.
21474         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
21475         (aarch64_emit_unspec_cond_or): Replace with...
21476         (aarch64_emit_sve_or_conds): ...this new function.  Use
21477         aarch64_emit_sve_ptrue_op for the individual comparisons and
21478         aarch64_emit_binop to OR them together.
21479         (aarch64_emit_inverted_unspec_cond): Replace with...
21480         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
21481         aarch64_emit_sve_ptrue_op for the comparison and
21482         aarch64_emit_unop to invert the result.
21483         (aarch64_expand_sve_vec_cmp_float): Update after the above
21484         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
21486 2018-05-07  Nathan Sidwell  <nathan@acm.org>
21488         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
21489         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
21490         (Backwards Compatibility): Likewise.
21492 2018-05-07  Luis Machado  <luis.machado@linaro.org>
21494         PR bootstrap/85681
21495         Revert:
21496         2018-05-07  Luis Machado  <luis.machado@linaro.org>
21498         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21499         <prefetch_dynamic_strides>: New const bool field.
21500         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21501         prefetch_dynamic_strides.
21502         (exynosm1_prefetch_tune): Likewise.
21503         (thunderxt88_prefetch_tune): Likewise.
21504         (thunderx_prefetch_tune): Likewise.
21505         (thunderx2t99_prefetch_tune): Likewise.
21506         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
21507         to false.
21508         (aarch64_override_options_internal): Update to set
21509         PARAM_PREFETCH_DYNAMIC_STRIDES.
21510         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
21511         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
21512         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
21513         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
21514         prefetch-dynamic-strides setting.
21516         2018-05-07  Luis Machado  <luis.machado@linaro.org>
21518         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21519         <minimum_stride>: New const int field.
21520         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21521         minimum_stride field.
21522         (exynosm1_prefetch_tune): Likewise.
21523         (thunderxt88_prefetch_tune): Likewise.
21524         (thunderx_prefetch_tune): Likewise.
21525         (thunderx2t99_prefetch_tune): Likewise.
21526         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
21527         (aarch64_override_options_internal): Update to set
21528         PARAM_PREFETCH_MINIMUM_STRIDE.
21529         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
21530         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
21531         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
21532         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
21533         stride is constant and is below the minimum stride threshold.
21535 2018-05-07  Luis Machado  <luis.machado@linaro.org>
21537         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
21538         to 512.
21540 2018-05-07  Luis Machado  <luis.machado@linaro.org>
21542         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21543         <prefetch_dynamic_strides>: New const bool field.
21544         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21545         prefetch_dynamic_strides.
21546         (exynosm1_prefetch_tune): Likewise.
21547         (thunderxt88_prefetch_tune): Likewise.
21548         (thunderx_prefetch_tune): Likewise.
21549         (thunderx2t99_prefetch_tune): Likewise.
21550         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
21551         to false.
21552         (aarch64_override_options_internal): Update to set
21553         PARAM_PREFETCH_DYNAMIC_STRIDES.
21554         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
21555         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
21556         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
21557         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
21558         prefetch-dynamic-strides setting.
21560 2018-05-07  Luis Machado  <luis.machado@linaro.org>
21562         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21563         <minimum_stride>: New const int field.
21564         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21565         minimum_stride field.
21566         (exynosm1_prefetch_tune): Likewise.
21567         (thunderxt88_prefetch_tune): Likewise.
21568         (thunderx_prefetch_tune): Likewise.
21569         (thunderx2t99_prefetch_tune): Likewise.
21570         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
21571         (aarch64_override_options_internal): Update to set
21572         PARAM_PREFETCH_MINIMUM_STRIDE.
21573         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
21574         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
21575         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
21576         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
21577         stride is constant and is below the minimum stride threshold.
21579 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
21581         PR c++/85659
21582         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
21583         the type is addressable.  Don't force op into register if it has
21584         BLKmode.
21586 2018-05-05  Roland McGrath  <mcgrathr@google.com>
21588         PR other/77609
21589         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
21590         any section for which we don't know a specific type it should have,
21591         regardless of name.  Previously this was done only for the exact
21592         names ".init_array", ".fini_array", and ".preinit_array".
21593         (default_elf_asm_named_section): Add comment about
21594         relationship with default_section_type_flags and SECTION_NOTYPE.
21595         (get_section): Don't consider it a type conflict if one side has
21596         SECTION_NOTYPE and the other doesn't, as long as neither has the
21597         SECTION_BSS et al used in the default_section_type_flags logic.
21599 2018-05-05  Tom de Vries  <tom@codesourcery.com>
21601         PR target/85653
21602         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
21603         (workaround_barsyncs): New function.
21604         (nvptx_reorg): Use workaround_barsyncs.
21605         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
21606         (define_expand "nvptx_membar_cta"): New define_expand.
21607         (define_insn "*nvptx_membar_cta"): New insn.
21609 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
21611         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
21612         To improve optimization opportunities.
21613         * builtin-types.def: The new needed builtin types for the above.
21615 2018-05-04  Richard Biener  <rguenther@suse.de>
21617         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
21618         * gimple-ssa-store-merging.c
21619         (imm_store_chain_info::output_merged_store): Remove redundant create,
21620         release split_store vector contents on failure.
21621         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
21622         scalar stmt vector on cache hit.
21624 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
21626         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
21627         Xilinx FP support.
21628         * config.gcc (powerpc-xilinx-eabi*): Remove.
21629         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
21630         support.
21631         (fusion_addis_mem_combo_load): Ditto.
21632         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
21633         FP support.
21634         (rs6000_cpu_cpp_builtins): Ditto.
21635         * config/rs6000/rs6000-linux.c
21636         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
21637         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
21638         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
21639         support.
21640         (rs6000_setup_reg_addr_masks): Ditto.
21641         (rs6000_init_hard_regno_mode_ok): Ditto.
21642         (rs6000_option_override_internal): Ditto.
21643         (legitimate_lo_sum_address_p): Ditto.
21644         (rs6000_legitimize_address): Ditto.
21645         (rs6000_legitimize_reload_address): Ditto.
21646         (rs6000_legitimate_address_p): Ditto.
21647         (abi_v4_pass_in_fpr): Ditto.
21648         (setup_incoming_varargs): Ditto.
21649         (rs6000_gimplify_va_arg): Ditto.
21650         (rs6000_split_multireg_move): Ditto.
21651         (rs6000_savres_strategy): Ditto.
21652         (rs6000_emit_prologue_components): Ditto.
21653         (rs6000_emit_epilogue_components): Ditto.
21654         (rs6000_emit_prologue): Ditto.
21655         (rs6000_emit_epilogue): Ditto.
21656         (rs6000_elf_file_end): Ditto.
21657         (rs6000_function_value): Ditto.
21658         (rs6000_libcall_value): Ditto.
21659         * config/rs6000/rs6000.h: Ditto.
21660         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
21661         (TARGET_MINMAX): ... this.  New.
21662         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
21663         * config/rs6000/rs6000.md: Remove Xilinx FP support.
21664         (*movsi_internal1_single): Delete.
21665         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
21666         mfpu=, mxilinx-fpu): Delete.
21667         * config/rs6000/singlefp.h: Delete.
21668         * config/rs6000/sysv4.h: Remove Xilinx FP support.
21669         * config/rs6000/t-rs6000: Ditto.
21670         * config/rs6000/t-xilinx: Delete.
21671         * config/rs6000/titan.md: Adjust for fp_type removal.
21672         * config/rs6000/vsx.md: Remove Xilinx FP support.
21673         (VStype_simple): Delete.
21674         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
21675         * config/rs6000/xfpu.h: Delete.
21676         * config/rs6000/xfpu.md: Delete.
21677         * config/rs6000/xilinx.h: Delete.
21678         * config/rs6000/xilinx.opt: Delete.
21679         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
21680         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
21682 2018-05-04  Tom de Vries  <tom@codesourcery.com>
21684         PR libgomp/85639
21685         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
21686         if ignore == 0.
21688 2018-05-04  Richard Biener  <rguenther@suse.de>
21690         PR middle-end/85627
21691         * tree-complex.c (update_complex_assignment): We are always in SSA form.
21692         (expand_complex_div_wide): Likewise.
21693         (expand_complex_operations_1): Likewise.
21694         (expand_complex_libcall): Preserve EH info of the original stmt.
21695         (tree_lower_complex): Handle removed blocks.
21696         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
21697         on complex multiplication and division libcall builtins.
21699 2018-05-04  Richard Biener  <rguenther@suse.de>
21701         PR middle-end/85574
21702         * fold-const.c (negate_expr_p): Restrict negation of operand
21703         zero of a division to when we know that can happen without
21704         overflow.
21705         (fold_negate_expr_1): Likewise.
21707 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
21709         PR libstdc++/85466
21710         * real.h (real_nextafter): Declare.
21711         * real.c (real_nextafter): New function.
21712         * fold-const-call.c (fold_const_nextafter): New function.
21713         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
21714         CASE_CFN_NEXTTOWARD.
21715         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
21716         even when arg1_mode is different from arg0_mode.
21718 2018-05-03  Nathan Sidwell  <nathan@acm.org>
21720         * doc/extend.texi (Deprecated Features): Remove
21721         -ffriend-injection.
21722         (Backwards Compatibility): Likewise.
21723         * doc/invoke.texi (C++ Language Options): Likewise.
21724         (C++ Dialect Options): Likewise.
21726 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
21728         PR target/85530
21729         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
21730         _mm512_mask_mullox_epi64): New intrinsics.
21732 2018-05-03  Tom de Vries  <tom@codesourcery.com>
21734         PR testsuite/85106
21735         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
21736         dump files): Add offload-tree.
21738 2018-05-03  Richard Biener  <rguenther@suse.de>
21740         PR tree-optimization/85615
21741         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
21742         to loops not nested in BBs loop father to avoid creating multi-entry
21743         loops.
21745 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21747         PR tree-optimization/70291
21748         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
21749         arguments.  Change return type to tree.  Emit libcall as a new
21750         statement rather than replacing existing one when inplace_p is true.
21751         (expand_complex_multiplication_components): New function.
21752         (expand_complex_multiplication): Expand floating-point complex
21753         multiplication using the above.
21754         (expand_complex_division): Rename inner_type parameter to type.
21755         Update expand_complex_libcall call-site.
21756         (expand_complex_operations_1): Update expand_complex_multiplication
21757         and expand_complex_division call-sites.
21759 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
21761         PR target/85582
21762         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
21763         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
21764         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
21765         the highest significant bit of the shift count mask is clear.  In
21766         check whether and[sq]i3 is needed verify that all significant bits
21767         of the shift count other than the highest are set.
21769 2018-05-02  Tom de Vries  <tom@codesourcery.com>
21771         PR libgomp/82428
21772         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
21773         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
21774         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
21775         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
21776         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
21777         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
21778         __builtin_goacc_parlevel_size.
21780 2018-05-02  Richard Biener  <rguenther@suse.de>
21782         PR tree-optimization/85597
21783         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
21784         do not use split vect_get_vec_defs call but call vect_get_slp_defs
21785         directly.
21787 2018-05-02  Tom de Vries  <tom@codesourcery.com>
21789         PR testsuite/85106
21790         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
21791         dump files): Add ltrans-tree.
21793 2018-05-02  Tom de Vries  <tom@codesourcery.com>
21795         PR testsuite/85106
21796         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
21797         dump files): Add wpa-ipa.
21799 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
21801         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
21802         powerpc*-*-linux*paired* target.
21803         * config/rs6000/750cl.h: Delete.
21804         * config/rs6000/paired.h: Delete.
21805         * config/rs6000/paired.md: Delete.
21806         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
21807         float support.
21808         * config/rs6000/rs6000-builtin.def: Remove paired float support.
21809         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
21810         comment.  Remove paired float support.
21811         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
21812         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
21813         VECTOR_PAIRED.
21814         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
21815         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
21816         declarations.
21817         * config/rs6000/rs6000.c: Remove paired float support.
21818         (paired_expand_vector_init, paired_expand_vector_move,
21819         paired_emit_vector_compare, paired_emit_vector_cond_expr,
21820         (paired_expand_lv_builtin, paired_expand_stv_builtin,
21821         paired_expand_builtin, paired_expand_predicate_builtin,
21822         paired_init_builtins): Delete.
21823         * config/rs6000/rs6000.h: Remove paired float support.
21824         * config/rs6000/rs6000.md: Remove paired float support.
21825         (move_from_CR_ov_bit): Delete.
21826         * config/rs6000/rs6000.opt (mpaired): Delete.
21827         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
21828         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
21830 2018-05-02  Richard Biener  <rguenther@suse.de>
21832         PR middle-end/85567
21833         * gimplify.c (gimplify_save_expr): When in SSA form allow
21834         SAVE_EXPRs to compute to SSA vars.
21836 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
21838         PR target/85582
21839         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
21840         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
21841         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
21842         clobber operands[2], instead use a new pseudo.  Formatting fixes.
21844 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
21846         PR tree-optimization/85586
21847         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
21848         exit early for statements in the same group if the accesses are
21849         not strided.
21851 2018-05-02  Tom de Vries  <tom@codesourcery.com>
21853         PR lto/85451
21854         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
21855         error message.
21857 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
21859         PR tree-optimization/85143
21860         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
21862 2018-05-01  Tom de Vries  <tom@codesourcery.com>
21864         PR lto/85451
21865         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
21866         not found" error message.
21868 2018-05-01  Tom de Vries  <tom@codesourcery.com>
21870         PR other/83786
21871         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
21872         * vec.c (test_ordered_remove_if): New function.
21873         (vec_c_tests): Call test_ordered_remove_if.
21874         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
21875         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
21876         * tree-vect-patterns.c (vect_pattern_recog_1): Use
21877         VEC_ORDERED_REMOVE_IF.
21879 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
21881         PR tree-optimization/82665
21882         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
21883         pointer subtraction where arguments come from a memchr call.
21885 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
21887         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
21888         --push-state --as-needed and --pop-state instead of --as-needed and
21889         --no-as-needed if ld supports it.
21890         * configure: Regenerated.
21892         PR web/85578
21893         * doc/install.texi2html: Replace _002d with - and _002a with * in
21894         generated html files using sed.
21896 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
21898         PR c++/85523
21899         * gcc-rich-location.c (blank_line_before_p): New function.
21900         (use_new_line): New function.
21901         (gcc_rich_location::add_fixit_insert_formatted): New function.
21902         * gcc-rich-location.h
21903         (gcc_rich_location::add_fixit_insert_formatted): New function.
21905 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
21907         * selftest.c (assert_streq): Rename "expected" and "actual" to
21908         "val1" and "val2".  Extend NULL-handling to cover both inputs
21909         symmetrically, while still requiring both to be non-NULL for a pass.
21910         * selftest.h (assert_streq): Rename "expected" and "actual" to
21911         "val1" and "val2".
21912         (ASSERT_EQ): Likewise.
21913         (ASSERT_EQ_AT): Likewise.
21914         (ASSERT_KNOWN_EQ): Likewise.
21915         (ASSERT_KNOWN_EQ_AT): Likewise.
21916         (ASSERT_NE): Likewise.
21917         (ASSERT_MAYBE_NE): Likewise.
21918         (ASSERT_MAYBE_NE_AT): Likewise.
21919         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
21920         the assertion to pass.
21921         (ASSERT_STREQ_AT): Likewise.
21923 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
21925         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
21926         interaction with -pie.
21928 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
21930         * selftest.h: Fix alphabetization of per-source-file selftest
21931         declarations.
21933 2018-04-30  Jason Merrill  <jason@redhat.com>
21935         PR c++/61982 - dead stores to destroyed objects.
21936         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
21937         of clobber.
21939 2018-04-30  Jason Merrill  <jason@redhat.com>
21941         * tree.c (build_clobber): New.
21942         * tree.h: Declare it.
21943         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
21945 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
21947         * diagnostic-show-locus.c (layout::layout): Update for
21948         location_get_source_line returning a char_span.
21949         (struct char_span): Move to input.h.
21950         (struct correction): Update for fields in char_span becoming
21951         private.
21952         (struct source_line): Update for location_get_source_line
21953         returning a char_span.
21954         (layout::print_line): Likewise.
21955         * edit-context.c (edited_file::print_content): Likewise.
21956         (edited_file::print_diff_hunk): Likewise.
21957         (edited_file::print_run_of_changed_lines): Likewise.
21958         (edited_file::get_num_lines): Likewise.
21959         (edited_line::edited_line): Likewise.
21960         * final.c (asm_show_source): Likewise.
21961         * input.c (location_get_source_line): Convert return type
21962         from const char * to char_span, losing the final "line_len"
21963         param.
21964         (dump_location_info): Update for the above.
21965         (get_substring_ranges_for_loc): Likewise.  Use a char_span
21966         when handling the literal within the line.
21967         (test_reading_source_line): Update for location_get_source_line
21968         returning a char_span.
21969         * input.h (class char_span): Move here from
21970         diagnostic-show-locus.c, converting from a struct to a class.
21971         Make data members private.
21972         (char_span::operator bool): New.
21973         (char_span::length): New.
21974         (char_span::get_buffer): New.
21975         (char_span::operator[]): New.
21976         (char_span::subspan): Make const.
21977         (char_span::xstrdup): New.
21978         (location_get_source_line): Convert return type from const char *
21979         to char_span, losing the final "line_size" param.
21981 2018-04-30  Jan Hubicka  <jh@suse.cz>
21983         * lto-wrapper.c (ltrans_priorities): New static var.
21984         (cmp_priority): New.
21985         (run_gcc): Read priorities and if doing parallel build order
21986         the Makefile by them.
21988 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
21990         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
21992 2018-04-30  Richard Biener  <rguenther@suse.de>
21994         * tree-cfg.c (verify_address): Remove base argument, add
21995         flag whether to check TREE_ADDRESSABLE and do that.
21996         (verify_expr): Remove.
21997         (verify_types_in_gimple_reference): Add pieces from verify_expr.
21998         (verify_gimple_assign_single): Likewise.
21999         (verify_gimple_switch): Likewise.
22000         (verify_expr_location_1): Dereference tp once.  Add (disabled)
22001         piece from verify_expr.
22002         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
22004 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
22006         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
22008 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
22010         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
22011         (small_data_pattern): Likewise.
22012         (arc_rewrite_small_data): Likewise.
22013         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
22014         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
22015         (get_symbol_alignment): New function.
22016         (legitimate_small_data_address_p): Likewise.
22017         (legitimate_scaled_address): Update, call
22018         legitimate_small_data_address_p.
22019         (output_sdata): New static variable.
22020         (arc_print_operand): Update how we handle small data operands.
22021         (arc_print_operand_address): Likewise.
22022         (arc_legitimate_address_p): Update, use
22023         legitimate_small_data_address_p.
22024         (arc_rewrite_small_data_p): Remove.
22025         (arc_rewrite_small_data_1): Likewise.
22026         (arc_rewrite_small_data): Likewise.
22027         (small_data_pattern): Likewise.
22028         (compact_sda_memory_operand): Update to use
22029         legitimate_small_data_address_p and get_symbol_alignment.
22030         (prepare_move_operands): Don't rewite sdata pattern.
22031         (prepare_extend_operands): Remove.
22032         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
22033         pattern.
22034         (zero_extendqisi2): Likewise.
22035         (zero_extendhisi2): Likewise.
22036         (extendqihi2): Likewise.
22037         (extendqisi2): Likewise.
22038         (extendhisi2): Likewise.
22039         (addsi3): Likewise.
22040         (subsi3): Likewise.
22041         (andsi3): Likewise.
22042         * config/arc/constraints.md (Usd): Change it to memory constraint.
22044 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
22046         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
22047         as source of std instructions.
22048         * config/arc/arc.md (movsi_insn): Update pattern predicate to
22049         allow 6-bit constants as source for store instructions.
22050         (movdi_insn): Update instruction pattern to allow 6-bit constants
22051         as source for store instructions.
22053 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
22055         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
22057 2018-04-30  Nathan Sidwell  <nathan@acm.org>
22058             Sandra Loosemore <sandra@codesourcery.com>
22060         * dumpfile.c (dump_open): Allow '-' for stdout.
22061         * doc/invoke.texi (Developer Options): Document dump filename
22062         determination early.  Document stdin/stdout selection.
22064 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
22066         Microblaze Target: PIC data text relative
22068         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
22069         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
22070         Add declaration.
22071         * config/microblaze/microblaze.h (microblaze_constant_address_p):
22072         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
22073         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
22074         New addressing mode for data-text relative position indepenedent code.
22075         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
22076         'ADDRESS_SYMBOLIC_TXT_REL'.
22077         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
22078         (microblaze_legitimate_pic_operand): Exclude function calls from
22079         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
22080         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
22081         addresses cases.
22082         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
22083         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
22084         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
22085         for 'address + offset'.
22086         (microblaze_expand_prologue): Add new function prologue call for
22087         'r20' assignation.
22088         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
22089         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
22090         table in case of TARGET_PIC_DATA_TEXT_REL.
22091         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
22092         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
22093         Add new macros 'UNSPEC_TEXT',
22094         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
22095         + exclude function calls from 'UNSPEC_PLT' in case of data text
22096         relative mode.
22097         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
22098         new target hook for generating address diff vector tables in case of
22099         flag_pic.
22100         * doc/tm.texi : Regenerate.
22101         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
22102         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
22103         of addr diff vector generation.
22104         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
22105         target hook definition.
22106         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
22107         Add default function for generate_pic_addr_diff_vec -> flag_pic.
22108         * doc/invoke.texi (Add new pic option): Add new microblaze pic
22109         option for data text relative.
22111 2018-04-30  Richard Biener  <rguenther@suse.de>
22113         * tree-chrec.h (evolution_function_is_constant_p): Remove
22114         redundant check.
22115         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
22117 2018-04-30  Richard Biener  <rguenther@suse.de>
22119         PR bootstrap/85571
22120         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
22122 2018-04-30  Richard Biener  <rguenther@suse.de>
22124         PR tree-optimization/28364
22125         PR tree-optimization/85275
22126         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
22127         copying first exit test.
22129 2018-04-28  Mark Wielaard  <mark@klomp.org>
22131         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
22132         dwarf_version >= 5.
22133         (dwarf_AT): Handle DW_AT_addr_base.
22134         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
22136 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
22138         PR target/84431
22139         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
22140         (*ashl<dwi>3_doubleword_mask_1): Ditto.
22141         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
22142         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
22144 2018-04-28  Richard Biener  <rguenther@suse.de>
22146         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
22147         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
22148         to reflect use.  Only add interesting stmts.
22150 2018-04-27  Martin Jambor  <mjambor@suse.cz>
22152         PR ipa/85549
22153         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
22154         the jump function allows for passing through aggregate values.
22156 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
22158         * input.h (in_system_header_at): Convert from macro to inline
22159         function.
22160         (from_macro_expansion_at): Likewise.
22161         (from_macro_definition_at): Likewise.
22163 2018-04-27  Jeff Law  <law@redhat.com>
22165         * config.gcc: Mark tile* targets as deprecated/obsolete.
22167 2018-04-27  Richard Biener  <rguenther@suse.de>
22169         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
22170         fix for ILP32.
22172 2018-04-27  Richard Biener  <rguenther@suse.de>
22174         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
22176 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
22178         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
22179         with Yd constraint. Set "preferred_for_speed" attribute from
22180         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
22181         with Yd constraint.
22182         (*movdi_internal): Ditto.
22183         (movti_interunit splitters): Remove
22184         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
22185         (movdi_interunit splitters): Ditto.
22186         * config/i386/constraints.md (Ye): Remove.
22187         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
22189 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22191         PR target/85512
22192         * config/aarch64/constraints.md (Usg): Limit to 31.
22193         (Usj): Limit to 63.
22195 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
22197         PR tree-optimization/85529
22198         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
22199         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
22200         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
22201         zero extension or masking of the MSB bit.
22202         (optimize_range_tests): Add FIRST_BB argument, pass it through
22203         to optimize_range_tests_var_bound.
22204         (maybe_optimize_range_tests, reassociate_bb): Adjust
22205         optimize_range_tests callers.
22207 2018-04-26  Richard Biener  <rguenther@suse.de>
22208             Jakub Jelinek  <jakub@redhat.com>
22210         * cgraph.h (symbol_table): Just declare debug method here.
22211         * symtab.c (symbol_table::debug): Define.
22213 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
22215         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
22217 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
22219         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
22220         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
22221         (*movdi_internal): Substitute Yi and Yj constraint with x
22222         and Ym and Yn constraint with y constraint.  Update "isa"
22223         attribute and set "preferred_for_speed" attribute from
22224         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
22225         (*movsi_internal): Ditto.
22226         (*movdf_internal): Ditto.
22227         (*movsf_internal): Ditto.
22228         (*zero_extendsidi2): Ditto.
22229         * config/i386/sse.md (vec_set<mode>_0): Ditto.
22230         (sse2_loadld): Ditto.
22231         (*vec_extract<ssevecmodelower>_0): Ditto.
22232         (*vec_extractv4si_0_zext_sse4): Ditto.
22233         (vec_concatv2di): Ditto.
22234         (*vec_dup<mode>): Ditto.
22235         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
22236         * config/i386/constraints.md (Yi): Remove.
22237         (Yj): Remove.
22238         (Ym): Remove.
22239         (Yn): Remove.
22241 2018-04-26  Nathan Sidwell  <nathan@acm.org>
22243         * dumpfile.c (dump_open): New.
22244         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
22245         (dump_finish): Detect stdio/stderr by value not name.
22247 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
22249         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
22251 2018-04-26  Tom de Vries  <tom@codesourcery.com>
22253         PR target/84952
22254         * config/nvptx/nvptx.c (verify_neutering_jumps)
22255         (verify_neutering_labels): New function
22256         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
22258 2018-04-26  Tom de Vries  <tom@codesourcery.com>
22260         PR target/84025
22261         * config/nvptx/nvptx.c (needs_neutering_p): New function.
22262         (nvptx_single): Use needs_neutering_p to skip over insns that do not
22263         need neutering.
22265 2018-04-26  Richard Biener <rguenther@suse.de>
22266             Tom de Vries  <tom@codesourcery.com>
22268         PR lto/85422
22269         * lto-streamer-out.c (output_function): Fixup loops if required to match
22270         discovery done in the reader.
22272 2018-04-26  Richard Biener  <rguenther@suse.de>
22274         PR tree-optimization/85116
22275         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
22276         have a loop exit from the single latch predecessor.  Remove
22277         case of header with just condition.
22278         (ch_base::copy_headers): Exclude infinite loops from any
22279         processing.
22280         (pass_ch::execute): Record exits.
22282 2018-04-26  Richard Biener  <rguenther@suse.de>
22284         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
22285         prologue cost vector and pass it to vect_get_load_cost.
22286         (vect_get_peeling_costs_all_drs): Likewise.
22287         (vect_peeling_hash_get_lowest_cost): Likewise.
22288         (vect_enhance_data_refs_alignment): Likewise.
22290 2018-04-26  Richard Biener  <rguenther@suse.de>
22292         PR middle-end/85450
22293         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
22294         checking of integer<->pointer conversions.
22295         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
22296         sign-/zero-extending pointer types.
22297         (expand_omp_for_static_chunk): Likewise.
22299 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
22300             Jean Lee  <xiaoyur347@gmail.com>
22302         * config/mips/mips.c (mips_asan_shadow_offset): New function.
22303         (TARGET_ASAN_SHADOW_OFFSET): Define.
22304         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
22305         true for -fsanitize=address.
22307 2018-04-25  Mark Wielaard  <mark@klomp.org>
22309         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
22310         shorter ones.
22312 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
22314         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
22315         than "alu", remove explicit "memory" and "imm_disp" attributes.
22316         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
22318         PR middle-end/85414
22319         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
22320         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
22321         gen_lowpart_no_emit.
22323 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
22325         PR target/85473
22326         * config/i386/i386.c (ix86_expand_builtin): Change memory
22327         operand to XI, extend p0 to Pmode.
22328         * config/i386/i386.md: Change unspec volatile and operand
22329         1 mode to XI, change operand 0 mode to P.
22331 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
22333         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
22334         GET_MODE_MASK before any checking.
22335         (nds32_can_use_bset_p): Likewise.
22336         (nds32_can_use_btgl_p): Likewise.
22338 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
22340         * config/nds32/nds32-doubleword.md: New define_split pattern for
22341         illegal register number.
22343 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
22345         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
22347 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
22349         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
22351 2018-04-25  Richard Biener  <rguenther@suse.de>
22353         * lto-streamer.h (LTO_major_version): Bump to 8.
22355 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
22357         * BASE-VER: Set to 9.0.0.
22359 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
22361         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
22362         in __abskf2 and __powikf2.
22364 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22366         PR target/85512
22367         * config/aarch64/constraints.md (Usg, Usj): New constraints.
22368         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
22369         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
22370         Use the above on operand 2.  Reindent.
22371         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
22373 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
22375         PR target/85485
22376         * common/config/i386/i386-common.c (ix86_handle_option): Don't
22377         handle OPT_mcet.
22378         * config/i386/i386.opt (mcet): Removed.
22379         * doc/install.texi: Remove -mcet documentation.
22380         * doc/invoke.texi: Likewise.
22382 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
22384         PR target/85485
22385         * doc/install.texi: Remove -mcet from bootstrap-cet.
22387 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
22389         PR target/85511
22390         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
22391         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
22392         if TARGET_64BIT.
22394         PR target/85503
22395         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
22396         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
22397         containing a CONST_VECTOR.
22399 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
22401         * doc/install.texi: Update newlib dependency for nvptx.
22403 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
22405         PR target/85508
22406         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
22407         instead of INTVAL when shifting x left.
22409 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
22411         PR tree-optimization/85478
22412         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
22413         vect_grouped_store_supported for single element vectors.
22415 2018-04-24  Richard Biener  <rguenther@suse.de>
22417         PR target/85491
22418         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
22419         load cost increase to the case of non-constant step.
22421 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
22423         PR target/84828
22424         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
22425         destination if any_malformed_asm.
22427 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
22429         PR middle-end/85496
22430         * expr.c (store_field): In the bitfield case, if the value comes from
22431         a function call and is returned in registers by means of a PARALLEL,
22432         do not change the mode of the temporary unless BLKmode and VOIDmode.
22434 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
22436         PR rtl-optimization/85423
22437         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
22438         dependencies to debug insns when the previous insn is non-debug.
22440 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
22442         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
22443         enums into a single definition.
22444         (fls): Fix predicates and printing.
22445         (seti): Likewise.
22447 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
22449         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
22450         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
22451         and short u6 immediate.
22452         (check_if_valid_sleep_operand): Remove.
22453         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
22455 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
22457         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
22458         flag_always_save_lp condition.
22459         * config/nds32/nds32.opt (malways-save-lp): New option.
22461 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
22463         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
22464         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
22465         * config/nds32/nds32.h
22466         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
22467         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
22469 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
22471         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
22472         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
22474 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
22475             Chung-Ju Wu  <jasonwucj@gmail.com>
22477         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
22478         Declare.
22479         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
22480         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
22482 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
22484         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
22486 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
22488         * config/nds32/nds32-protos.h (nds32_data_alignment,
22489         nds32_local_alignment): Declare.
22490         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
22491         nds32_local_alignment): New functions.
22492         (TARGET_CONSTANT_ALIGNMENT): Define.
22493         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
22495 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
22497         * config/nds32/nds32.c
22498         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
22499         (TARGET_MODES_TIEABLE_P): Likewise.
22501 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
22503         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
22504         level Ofast and Og.
22506 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
22507             Chung-Ju Wu  <jasonwucj@gmail.com>
22509         * config/nds32/constants.md (unspec_volatile_element): Add enum values
22510         for unaligned access.
22511         * config/nds32/nds32-intrinsic.c: Implementation of expanding
22512         unaligned access.
22513         * config/nds32/nds32-intrinsic.md: Likewise.
22514         * config/nds32/nds32_intrinsic.h: Likewise.
22515         * config/nds32/nds32.h (nds32_builtins): Likewise.
22516         * config/nds32/nds32.opt (munaligned-access): New option.
22517         * config/nds32/nds32.c (nds32_asm_file_start): Display
22518         flag_unaligned_access status.
22520 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
22522         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
22523         -mno-relax is present.
22524         * config/riscv/linux.h (LINK_SPEC): Ditto.
22526 2018-04-20  Martin Sebor  <msebor@redhat.com>
22528         PR c/85365
22529         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
22530         for null pointers.
22531         (gimple_fold_builtin_stxcpy_chk): Same.
22532         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
22534 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
22536         PR target/85456
22537         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
22538         __powikf2 when long double is IEEE 128-bit.
22540 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
22542         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
22543         step to make sure stack always aligned.
22545 2018-04-20  Carl Love  <cel@us.ibm.com>
22547         PR target/83402
22548         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
22549         size check for arg0.
22551 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
22552             Tom de Vries  <tom@codesourcery.com>
22554         PR target/85445
22555         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
22556         Emit insns for calls too.
22557         (nvptx_find_par): Always look for worker-level predecessor insn.
22558         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
22559         calls.
22560         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
22561         (nvptx_process_pars): Propagate frames for calls.
22563 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
22565         PR target/85469
22566         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
22567         Removed.
22568         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
22569         (ix86_handle_option): Don't handle OPT_mibt.
22570         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
22571         __SHSTK__.
22572         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
22573         has_ibt and ibt.
22574         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
22575         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
22576         (ix86_target_macros): Define __CET__ with flag_cf_protection
22577         for -fcf-protection.
22578         * config/i386/i386.c (isa2_opts): Remove -mibt.
22579         * config/i386/i386.h (TARGET_IBT): Removed.
22580         (TARGET_IBT_P): Likewise.
22581         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
22582         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
22583         * config/i386/i386.opt (mcet): Update help message.
22584         (mshstk): Likewise.
22585         (mibt): Removed.
22586         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
22587         -mcet as an alias for -mshstk.
22589 2018-04-20  Richard Biener <rguenther@suse.de>
22591         PR middle-end/85475
22592         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
22593         complexity by forcing a single use of the multiply operand.
22595 2018-04-20  Martin Jambor  <mjambor@suse.cz>
22597         ipa/85449
22598         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
22599         recursion dependency to only apply to non-clones.
22601 2018-04-20  Martin Jambor  <mjambor@suse.cz>
22603         ipa/85447
22604         * ipa-cp.c (create_specialized_node): Check that clones of
22605         self-recursive edges exist during IPA-CP.
22607 2018-04-19  Toon Moene  <toon@moene.org>
22609         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
22610         by -O3.
22612 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
22614         PR tree-optimization/85467
22615         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
22616         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
22617         VECTOR_CST element to type.
22619 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
22621         PR target/85397
22622         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
22623         * config/i386/i386.md (builtin_setjmp_setup): Removed.
22624         (builtin_longjmp): Likewise.
22625         (save_stack_nonlocal): New pattern.
22626         (restore_stack_nonlocal): Likewise.
22628 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
22630         PR target/85404
22631         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
22632         Replace ASM_OUTPUT_LABEL with fprintf.
22634 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
22636         PR target/85417
22637         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
22638         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
22639         * config/i386/i386-c.c (ix86_target_macros_internal): Also
22640         define __IBT__ and __SHSTK__ for -fcf-protection.
22641         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
22642         TARGET_IBT.
22643         (ix86_trampoline_init): Likewise.
22644         (x86_output_mi_thunk): Likewise.
22645         (ix86_notrack_prefixed_insn_p): Likewise.
22646         (ix86_option_override_internal): Don't disallow -fcf-protection.
22647         * config/i386/i386.md (rdssp<mode>): Also enable for
22648         -fcf-protection.
22649         (incssp<mode>): Likewise.
22650         (nop_endbr): Likewise.
22651         * config/i386/i386.opt (mcet): Change help message to built-in
22652         functions only.
22653         (mibt): Likewise.
22654         (mshstk): Likewise.
22655         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
22656         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
22657         enable CET built-in functions.
22659 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
22661         * common/config/i386/i386-common.c
22662         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
22663         OPTION_MASK_ISA_MOVDIRI_UNSET,
22664         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
22665         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
22666         * config.gcc (movdirintrin.h): New header.
22667         * config/i386/cpuid.h (bit_MOVDIRI,
22668         bit_MOVDIR64B): New bits.
22669         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
22670         and -mmvodir64b.
22671         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
22672         (VOID, PVOID, PCVOID)): New function types.
22673         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
22674         __builtin_ia32_directstoreu_u64,
22675         __builtin_ia32_movdir64b): New builtins.
22676         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
22677         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
22678         and -mmovdiri.
22679         (ix86_valid_target_attribute_inner_p): Ditto.
22680         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
22681         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
22682         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
22683         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
22684         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
22685         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
22686         (movdiri<mode>, movdir64b_<mode>): New.
22687         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
22688         * config/i386/immintrin.h: Include movdirintrin.h.
22689         * config/i386/movdirintrin.h: New file.
22690         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
22692 2018-04-19  Richard Biener  <rguenther@suse.de>
22694         PR middle-end/85455
22695         * cfg.c (clear_bb_flags): When loop state says we have
22696         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
22698 2018-04-19  Richard Biener  <rguenther@suse.de>
22700         PR tree-optimization/84737
22701         * tree-vect-data-refs.c (vect_copy_ref_info): New function
22702         copying restrict info.
22703         (vect_setup_realignment): Use it.
22704         * tree-vectorizer.h (vect_copy_ref_info): Declare.
22705         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
22706         the first DR to all generated stores.
22707         (vectorizable_load): Likewise for loads.
22709 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
22711         PR tree-optimization/85446
22712         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
22713         the integral and pointer types to have the same precision.
22715         * doc/install.texi: Document --disable-cet being the default and
22716         --enable-cet=auto.
22718 2018-04-18  Martin Liska  <mliska@suse.cz>
22720         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
22721         style.
22723 2018-04-18  Martin Liska  <mliska@suse.cz>
22725         Revert
22726         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
22728         PR ipa/83983
22729         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
22730         arguments if they are comparable.
22732 2018-04-18  Martin Liska  <mliska@suse.cz>
22734         Revert
22735         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
22737         PR lto/84805
22738         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
22739         incomplete types.
22741 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
22743         PR target/85388
22744         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
22745         ENDBR after calling __morestack.
22747 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
22749         PR jit/85384
22750         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
22751         by using gcc_base_ver to generate a gcc_driver_version, and use
22752         it when generating GCC_DRIVER_NAME.
22753         * configure: Regenerate.
22755 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
22757         PR target/81084
22758         * config.gcc: Obsolete powerpc*-*-*spe*.
22760 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
22762         PR debug/84637
22763         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
22764         (stabstr_D): Change type of unum from unsigned int to
22765         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
22766         type.
22768 2018-04-17  Jim Wilson  <jimw@sifive.com>
22770         PR 84856
22771         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
22772         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
22773         Set arg_pointer_offset after using pretend_args_size.
22775 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
22777         PR rtl-optimization/85431
22778         * dse.c (record_store): Ignore zero width stores.
22780         PR sanitizer/85230
22781         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
22782         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
22783         __builtin_stack_restore rather than after it.
22784         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
22785         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
22786         argument instead of virtual_dynamic_stack_rtx.
22788 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22790         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
22791         New prototype.
22792         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22793         Add note to error message to explain internal mapping of overloaded
22794         built-in function name to non-overloaded built-in function name.
22795         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
22796         function.
22798 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
22800         PR target/85424
22801         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
22802         where the inputs overlap with the output.
22804 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
22806         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
22807         (=v, v) alternative and explicit "memory" attribute.
22808         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
22809         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
22810         attributes.
22811         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
22812         "sselog1" type instead of "sselog".
22813         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
22814         "sselog".  Remove explicit "memory" attribute.
22815         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
22816         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
22817         attributes.
22818         (vec_extract_hi_v32hi): Merge all alternatives into one, use
22819         "sselog1" type instead of "sselog".  Remove explicit "memory"
22820         attribute.
22821         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
22822         use "sselog1" type instead of "sselog".  Remove explicit "memory"
22823         attribute.
22824         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
22825         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
22826         attributes.
22827         (vec_extract_hi_v64qi): Merge all alternatives into one, use
22828         "sselog1" type instead of "sselog".  Remove explicit "memory"
22829         attribute.
22830         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
22831         use "sselog1" type instead of "sselog".  Remove explicit "memory"
22832         attribute.
22834         PR target/85430
22835         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
22837         PR middle-end/85414
22838         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
22839         on a SUBREG.
22841 2018-04-17  Martin Jambor  <mjambor@suse.cz>
22843         PR ipa/85421
22844         * ipa-cp.c (create_specialized_node): Call
22845         expand_all_artificial_thunks if necessary.
22847 2018-04-17  Martin Liska  <mliska@suse.cz>
22849         PR lto/85405
22850         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
22851         in message, remote space in between '_G' and '('.
22853 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
22855         PR target/85281
22856         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
22857         avx512f_vmcmp<mode>3<round_saeonly_name>,
22858         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
22859         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
22860         avx512f_rndscale<mode><round_saeonly_name>,
22861         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
22862         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
22863         Use %<iptr>2 instead of %2 for -masm=intel.
22864         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
22865         avx512f_vcvttss2usi<round_saeonly_name>,
22866         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
22867         -masm=intel.
22868         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
22869         avx512f_vcvttsd2usi<round_saeonly_name>,
22870         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
22871         Use %q1 instead of %1 for -masm=intel.
22872         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
22873         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
22874         of %3 for -masm=intel.
22875         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
22876         -masm=intel.
22877         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
22878         -masm=intel.
22879         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
22880         -masm=intel.
22881         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
22882         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
22883         %g1.
22884         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
22885         -masm=intel.
22886         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
22887         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
22888         %g1 and one with %0 and %1.
22889         (avx512er_vmrcp28<mode><round_saeonly_name>,
22890         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
22891         %1 for -masm=intel.
22892         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
22893         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
22894         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
22895         of %0 and %{%4%} for -masm=intel.
22896         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
22897         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
22898         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
22899         order of %0 and %{%5%}%{z%} for -masm=intel.
22901 2018-04-17  Jan Hubicka  <jh@suse.cz>
22903         PR lto/85405
22904         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
22906 2018-04-17  Martin Liska  <mliska@suse.cz>
22908         PR ipa/85329
22909         * multiple_target.c (create_dispatcher_calls): Set apostrophes
22910         for target_clone error message.  Make default implementation
22911         clone to be a local declaration.
22912         (separate_attrs): Add new argument and check for an empty
22913         string.
22914         (expand_target_clones): Handle it.
22915         (ipa_target_clone): Make redirection just for target_clones
22916         functions.
22918 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
22919             Tom de Vries  <tom@codesourcery.com>
22921         PR middle-end/84955
22922         * omp-expand.c (expand_oacc_for): Add dummy false branch for
22923         tiled basic blocks without omp continue statements.
22925 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
22927         PR target/83660
22928         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
22929         vec_extract expression as having side effects to make sure it gets
22930         a cleanup point.
22932 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
22934         PR target/85403
22935         * config/i386/i386.c (get_builtin_code_for_version): Check
22936         error_mark_node.
22938 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
22940         PR target/84331
22941         * config.gcc: Support "skylake".
22942         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22943         PROCESSOR_SKYLAKE.
22944         * config/i386/i386.c (m_SKYLAKE): Define.
22945         (processor_target_table): Add "skylake".
22946         (ix86_option_override_internal): Add "skylake".
22947         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
22948         PROCESSOR_CANNONLAKE.
22949         (get_builtin_code_for_version): Fix priority for
22950         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
22951         PROCESSOR_SKYLAKE-AVX512.
22952         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
22953         (processor_type): Add PROCESSOR_SKYLAKE.
22955 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
22956             Jason Merrill  <jason@redhat.com>
22958         PR c++/85112
22959         * convert.c (convert_to_integer_1): Use direct recursion for
22960         enumeral types and types with a precision less than the number
22961         of bits in their mode.
22963 2018-04-16  Julia Koval  <julia.koval@intel.com>
22965         PR target/84413
22966         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
22967         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
22969 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
22971         PR target/85293
22972         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
22973         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
22974         and -mno-direct-move.
22976 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
22978         PR target/83402
22979         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
22980         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
22981         Ensure negative shifts result in {0}.
22983 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
22985         PR rtl-optimization/79916
22986         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
22987         regs (if any) to define how to gnerate SD moves when LRA is in
22988         progress.
22990 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
22992         PR rtl-optimization/85393
22993         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
22994         * except.c (expand_dw2_landing_pad_for_region): Make static.
22995         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
22996         a label and unconditional jump to old_bb, rather than
22997         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
22998         basic block.
23000         PR rtl-optimization/85376
23001         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
23002         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
23003         instead of a specific value.
23005 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
23006             Bin Cheng  <bin.cheng@arm.com>
23008         PR tree-optimization/82965
23009         PR tree-optimization/83991
23010         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
23011         by_profile_only parameter.
23012         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
23013         information if the loop was predicted to iterate too many times.
23014         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
23016 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
23018         PR lto/71991
23019         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
23020         always inline.
23022 2018-04-13  Martin Liska  <mliska@suse.cz>
23023             Jakub Jelinek  <jakub@redhat.com>
23025         PR middle-end/81657
23026         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
23027         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
23028         * builtins.c (expand_builtin_memory_copy_args): Use
23029         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
23030         handle dest_addr == pc_rtx.
23032 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
23034         PR target/85291
23035         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
23036         asked to not generate direct moves.
23037         (fix_trunc<mode>si2_stfiwx): Similar.
23038         (fix_trunc<mode>si2_internal): Similar.
23040 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
23042         PR debug/83157
23043         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
23044         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
23045         lookup if dest in some wider mode is known to be const0_rtx and
23046         if so, record permanent equivalence for it to be ZERO_EXTEND of
23047         the narrower mode destination.
23049 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
23051         * lto-streamer-out.c (output_function): Revert 259346.
23052         * omp-expand.c (expand_oacc_for): Likewise.
23054 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
23056         PR rtl-optimization/85354
23057         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
23058         * sel-sched.c (sel_global_init): ... here.
23060 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
23062         PR target/85238
23063         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
23064         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
23065         mode for PE-COFF targets.
23066         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
23067         (i386_pe_asm_lto_end): Likewise.
23068         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
23069         (TARGET_ASM_LTO_END): Likewise.
23070         * config/i386/winnt.c (saved_debug_info_level): New static variable.
23071         (i386_pe_asm_lto_start): New function.
23072         (i386_pe_asm_lto_end): Likewise.
23074 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
23075             Richard Biener  <rguenther@suse.de>
23077         PR middle-end/84955
23078         * lto-streamer-out.c (output_function): Fix CFG loop state before
23079         streaming out.
23080         * omp-expand.c (expand_oacc_for): Handle calls to internal
23081         functions like regular functions.
23083 2018-04-12  Richard Biener  <rguenther@suse.de>
23085         PR lto/85371
23086         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
23087         for the early LTO debug to properly generate references to it
23088         during DIE emission.  Do not re-use that for the skeleton for
23089         split-dwarf.
23090         (dwarf2out_early_finish): Likewise.
23092 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
23094         PR target/85328
23095         * config/i386/sse.md
23096         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
23097         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
23098         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
23099         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
23100         and output is a reg, avoid creating invalid lowpart subreg, but
23101         instead split into a 512-bit move.  Don't split if not AVX512VL,
23102         input is xmm16+ reg and output is a mem.
23103         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
23104         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
23105         xmm16+ reg and output is a mem.
23107 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23109         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
23110         also for flag_dwarf2_cfi_asm.
23112 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
23114         PR rtl-optimization/85342
23115         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
23116         a bool scalar var inside of the loop instead.  Don't try to update
23117         recog_data.operand after failed apply_change_group.
23119 2018-04-12  Tom de Vries  <tom@codesourcery.com>
23121         PR target/85296
23122         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
23123         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
23124         array with flexible array member as array without given dimension.
23125         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
23126         argument for undefined param to true.
23128 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
23130         PR target/85321
23131         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
23132         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
23133         from PowerPC section.
23134         * config/rs6000/sysv4.opt (mcall-): Improve help text.
23135         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
23136         help text that is too long.
23137         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
23138         help text that is too long.
23139         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
23140         help text that is too long.
23142 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
23144         * config/alpha/alpha.md (stack_probe_internal): Rename
23145         from "probe_stack".  Update all callers.
23147 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
23149         PR rtl-optimization/84566
23150         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
23151         sched_macro_fuse_insns.
23153 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
23155         PR target/84301
23156         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
23157         (compute_block_dependences): ... from here.
23159 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
23161         PR tree-optimization/85331
23162         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
23163         from int to HOST_WIDE_INT.
23165 2018-04-11  Martin Jambor  <mjambor@suse.cz>
23167         PR ipa/84149
23168         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
23169         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
23170         not the same as the source val.
23171         (cgraph_edge_brings_value_p): New parameter.
23172         (gather_edges_for_value): Pass destination value to
23173         cgraph_edge_brings_value_p.
23174         (perhaps_add_new_callers): Likewise.
23175         (get_info_about_necessary_edges): Likewise and exclude values brought
23176         only by self-recursive edges.
23177         (create_specialized_node): Redirect only clones of self-calling edges.
23178         (+self_recursive_pass_through_p): New function.
23179         (find_more_scalar_values_for_callers_subset): Use it.
23180         (find_aggregate_values_for_callers_subset): Likewise.
23181         (known_aggs_to_agg_replacement_list): Removed.
23182         (decide_whether_version_node): Re-calculate known constants for all
23183         remaining context clones.
23185 2018-04-11  Richard Biener  <rguenther@suse.de>
23187         PR lto/85339
23188         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
23189         from early DWARF output.
23190         (dwarf2out_early_finish): Output line info unconditionally into
23191         early DWARF and add reference to it.
23193 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
23195         PR target/85281
23196         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
23197         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
23198         other than V2DFmode using iptr mode attribute.
23199         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
23201 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
23203         PR rtl-optimization/84659
23204         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
23206 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
23208         PR debug/85302
23209         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
23210         SIZEP is NULL.
23211         (output_loc_list): Pass address of a dummy size variable even in the
23212         locview handling loop.
23213         (index_location_lists): Add comment on why skip_loc_list_entry can't
23214         call size_of_locs.
23216 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
23218         PR target/85261
23219         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
23220         into register.
23222 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
23224         PR target/85321
23225         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
23226         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
23227         and -mstring-compare-inline-limit.
23229 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
23231         PR target/85287
23232         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
23233         for stack clash protection in a register whenever we need it to be in
23234         a register.
23236 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
23238         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
23239         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
23241 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
23243         PR target/85321
23244         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
23245         the help text.
23246         (mlong-double-): Ditto.
23247         * config/rs6000/sysv4.opt (msdata=): Ditto.
23248         (mtls-size=): Ditto.
23250 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23252         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
23253         erroneous entries for
23254         "vector int vec_ldl (int, long int *)", and
23255         "vector unsigned int vec_ldl (int, unsigned long int *)".
23256         Add comments and entries for
23257         "vector bool char vec_ldl (int, bool char *)",
23258         "vector bool short vec_ldl (int, bool short *)",
23259         "vector bool int vec_ldl (int, bool int *)",
23260         "vector bool long long vec_ldl (int, bool long long *)",
23261         "vector pixel vec_ldl (int, pixel *)",
23262         "vector long long vec_ldl (int, long long *)",
23263         "vector unsigned long long vec_ldl (int, unsigned long long *)".
23264         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
23265         type tree bool_long_long_type_node and correct definition of
23266         bool_V2DI_type_node to make reference to this new type tree.
23267         (rs6000_mangle_type): Replace erroneous reference to
23268         bool_long_type_node with bool_long_long_type_node.
23269         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
23270         comments to emphasize sign distinctions for char and int types and
23271         replace RS6000_BTI_bool_long constant with
23272         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
23273         use of RS6000_BTI_pixel.
23274         (bool_long_type_node): Remove this macro definition.
23275         (bool_long_long_type_node): New macro definition
23277 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
23279         PR rtl-optimization/85300
23280         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
23281         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
23282         simplify_unary_operation fails.
23284 2018-04-10  Martin Liska  <mliska@suse.cz>
23286         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
23287         cgraph_edge and ipa_ref.
23289 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
23291         PR target/85177
23292         PR target/85255
23293         * config/i386/sse.md
23294         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
23295         computation of the VEC_MERGE selector from mask.
23296         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
23297         Fix decoding of the VEC_MERGE selector into mask.
23299 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
23301         PR tree-optimization/85286
23302         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
23304 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
23306         * final.c (final_1): Set insn_last_address as well as
23307         insn_current_address.
23309 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23311         PR target/85173
23312         * explow.c (emit_stack_probe): Call validize_mem on memory location
23313         before passing it to gen_probe_stack.  Create address operand and
23314         legitimize it for the probe_stack_address case.
23316 2018-04-09  Jan Hubicka  <jh@suse.cz>
23318         PR lto/85078
23319         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
23320         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
23321         * tree.c (free_lang_data_in_type): Fix handling of binfos;
23322         walk basetypes.
23323         (free_lang_data): Rebuild type inheritance graph.
23325 2018-04-09  Martin Sebor  <msebor@redhat.com>
23327         * invoke.texi (-finline-small-functions): Mention other optimization
23328         options.
23329         (-findirect-inlining, -fpartial-inlining): Same.
23330         (-finline-functions-called-once): Same.
23331         (-freorder-blocks-and-partition): Same.
23333 2018-04-09  Jan Hubicka  <jh@suse.cz>
23335         PR rtl/84058
23336         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
23337         jumps; choose last target that matches the criteria (i.e.
23338         no partition changes for non-crossing jumps).
23339         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
23340         support for redirecting crossing jumps to non-crossing.
23342 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
23344         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
23345         also for naked functions.
23347 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
23349         * config/arc/arc.md (add_shift): New pattern.
23350         (add_shift2): Likewise.
23351         (sub_shift): Likewise.
23352         (sub_shift_cmp0_noout): Likewise.
23353         (compare_si_ashiftsi): Likewise.
23354         (xbfu_cmp0_noout): New combine pattern.
23355         (xbfu_cmp0"): Likewise.
23356         (movsi_set_cc_insn): Place the predicable variant first.
23357         (commutative_binary_cmp0_noout): Remove clobber.
23358         (commutative_binary_cmp0): New pattern.
23359         (noncommutative_binary_cmp0): Likewise.
23360         (noncommutative_binary_cmp0_noout): Likewise.
23361         (noncommutative_binary_comparison_result_used): Removed.
23362         (rsub_cmp0): New pattern.
23363         (rsub_cmp0_noout): Likewise.
23364         (extzvsi): Changed, keep only meaningful variants.
23365         (SQH, SEZ): New iterators.
23366         (SQH_postfix): New mode attribute.
23367         (SEZ_prefix): New code attribute.
23368         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
23369         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
23370         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
23371         of numerical value.
23372         (noncommutative_operator): Check the availability of barrel
23373         shifter option.
23375 2018-04-09  Richard Biener  <rguenther@suse.de>
23377         PR tree-optimization/85284
23378         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
23379         Only use the niter constraining form of simple_iv when the exit
23380         is always executed.
23382 2018-04-09  Tom de Vries  <tom@codesourcery.com>
23384         PR target/84041
23385         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
23386         (define_expand "*memory_barrier"): New define_expand.
23387         (define_insn "memory_barrier"): New insn.
23389 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
23391         PR rtl-optimization/80463
23392         PR rtl-optimization/83972
23393         PR rtl-optimization/83480
23395         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
23396         correct producer for the insn.
23397         (tidy_control_flow): Fixup seqnos in case of debug insns.
23399 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
23401         PR rtl-optimization/83913
23403         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
23404         different sched-times when merging exprs.
23406 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
23408         PR rtl-optimization/83962
23410         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
23411         tidy_fallthru_edge and tidy_control_flow.
23413 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
23415         PR rtl-optimization/83530
23417         * sel-sched.c (force_next_insn): New global variable.
23418         (remove_insn_for_debug): When force_next_insn is true, also leave only
23419         next insn in the ready list.
23420         (sel_sched_region): When the region wasn't scheduled, make another pass
23421         over it with force_next_insn set to 1.
23423 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
23425         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
23426         into tm_file.
23427         * config/nds32/constants.md (unspec_volatile_element): Add enum values
23428         for interrupt control.
23429         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
23430         functions for interrupt control.
23431         * config/nds32/nds32-intrinsic.md: Likewise.
23432         * config/nds32/nds32_intrinsic.h: Likewise.
23433         * config/nds32/nds32.h (nds32_builtins): Likewise.
23435 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
23437         * config/nds32/nds32.c (nds32_init_machine_status,
23438         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
23439         strict_aligned_p field.
23440         (nds32_expand_to_rtl_hook): New function.
23441         (TARGET_EXPAND_TO_RTL_HOOK): Define.
23442         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
23444 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
23445             Chung-Ju Wu  <jasonwucj@gmail.com>
23447         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
23448         * config/nds32/nds32-n7.md: New file.
23449         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
23450         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
23451         pipeline.
23452         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
23453         * config/nds32/nds32.md (pipeline_model): Add n7.
23454         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
23455         * config/nds32/pipelines.md: Include n7 settings.
23457 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
23458             Chung-Ju Wu  <jasonwucj@gmail.com>
23460         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
23461         * config/nds32/nds32-e8.md: New file.
23462         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
23463         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
23464         pipeline.
23465         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
23466         * config/nds32/nds32.md (pipeline_model): Add e8.
23467         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
23468         * config/nds32/pipelines.md: Include e8 settings.
23470 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
23471             Chung-Ju Wu  <jasonwucj@gmail.com>
23473         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
23474         * config/nds32/nds32-n8.md: New file.
23475         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
23476         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
23477         pipeline.
23478         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
23479         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
23480         * config/nds32/nds32.md (pipeline_model): Add n8.
23481         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
23482         * config/nds32/pipelines.md: Include n8 settings.
23484 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
23485             Chung-Ju Wu  <jasonwucj@gmail.com>
23487         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
23488         * config/nds32/nds32-n9-2r1w.md: New file.
23489         * config/nds32/nds32-n9-3r2w.md: New file.
23490         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
23491         nds32_register_ports): New or modify for cpu n9.
23492         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
23493         pipeline.
23494         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
23495         * config/nds32/nds32-utils.c: New file.
23496         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
23497         TARGET_MUL_SLOW): Define.
23498         * config/nds32/nds32.md (pipeline_model): New attribute.
23499         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
23500         New options that support cpu n9.
23501         * config/nds32/pipelines.md: Include n9 settings.
23502         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
23504 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
23506         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
23507         information if necessary.
23508         (output_cond_branch_compare_zero): Likewise.
23509         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
23510         (nds32_target_alignment): Refine for alignment.
23511         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
23512         (FUNCTION_BOUNDARY): Modify.
23513         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
23514         align case.
23515         * config/nds32/nds32.opt (malways-align, malign-functions): New.
23517 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
23519         * config/nds32/constants.md (unspec_volatile_element): Add values for
23520         TLB operation and data prefetch.
23521         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
23522         functions for TLB operation and data prefetch.
23523         * config/nds32/nds32-intrinsic.md: Likewise.
23524         * config/nds32/nds32_intrinsic.h: Likewise.
23525         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
23526         (nds32_print_operand): Likewise.
23527         * config/nds32/nds32.h (nds32_builtins): Likewise.
23529 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
23530         Andrew Pinski <pinsika@gcc.gnu.org>
23532         PR middle-end/82976
23533         * match.pd: Use constant_boolean_node of correct type instead of
23534         boolean_true_node or boolean_false_node for simplifying
23535         pointer comparisons to zero.
23537 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
23539         PR tree-optimization/80021
23540         * tree.c (verify_type_variant): Make error call in verify_variant_match
23541         translatable and remove final full stop.
23543 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
23545         * config/nds32/constants.md (unspec_volatile_element): Add
23546         UNSPEC_VOLATILE_EH_RETURN.
23547         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
23548         nds32_output_stack_pop): Support dwarf exception handling process.
23549         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
23550         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
23551         exception handling process.
23552         (nds32_compute_stack_frame): Likewise.
23553         (nds32_return_addr_rtx): Likewise.
23554         (nds32_initial_elimination_offset): Likewise.
23555         (nds32_expand_prologue): Likewise.
23556         (nds32_expand_epilogue): Likewise.
23557         (nds32_dynamic_chain_address): New function.
23558         * config/nds32/nds32.h (machine_function): Add fields for dwarf
23559         exception handling.
23560         (DYNAMIC_CHAIN_ADDRESS): Define.
23561         (EH_RETURN_DATA_REGNO): Define.
23562         (EH_RETURN_STACKADJ_RTX): Define.
23563         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
23564         patterns for dwarf exception handling.
23566 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
23568         * config/nds32/nds32.h: Clean up obsolete macros.
23570 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
23572         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
23573         Add enum values for particular instructions.
23574         * config/nds32/nds32-intrinsic.c: Implementation of expanding
23575         particular intrinsic functions.
23576         * config/nds32/nds32-intrinsic.md: Likewise.
23577         * config/nds32/nds32_intrinsic.h: Likewise.
23578         * config/nds32/nds32.h (nds32_builtins): Likewise.
23579         * config/nds32/nds32.md (type): Add pbsad and pbsada.
23580         (btst, ave): New patterns for particular instructions.
23582 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
23584         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
23585         Add enum values for atomic load/store and memory sync.
23586         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
23587         and memory sync.
23588         * config/nds32/nds32-intrinsic.md: Likewise.
23589         * config/nds32/nds32_intrinsic.h: Likewise.
23590         * config/nds32/nds32.h (nds32_builtins): Likewise.
23592 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
23594         PR tree-optimization/85257
23595         * fold-const.c (native_encode_vector): If not all elts could fit
23596         and off is -1, return 0 rather than offset.
23597         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
23598         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
23599         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
23600         adjust buffer in native_interpret_expr call.
23602 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
23604         * config/nds32/constants.md (unspec_volatile_element): Add cache
23605         control enum values.
23606         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
23607         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
23608         * config/nds32/nds32.c (nds32_cctl_names): New.
23609         (nds32_print_operand): Handle cache control register names.
23610         * config/nds32/nds32.h (nds32_builtins): New enum values.
23611         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
23612         macros.
23613         * config/nds32/nds32.md (type): Add mmu.
23614         * config/nds32/pipelines.md (simple_insn): Add mmu.
23616 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
23618         * config/nds32/nds32.md (type): Remove call.
23619         * config/nds32/pipelines.md (simple_insn): Likewise.
23621 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
23623         * config/nds32/constants.md (unspec_volatile_element): Add
23624         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
23625         UNSPEC_VOLATILE_FMFCFG.
23626         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
23627         description for fmfcfg and fmfcsr.
23628         (bdesc_1arg): Add fmtcsr.
23629         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
23630         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
23631         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
23632         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
23633         unspec_fmfcfg): New patterns.
23634         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
23635         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
23636         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
23637         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
23638         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
23639         __nds32__fmfcfg): Define.
23641 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
23643         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
23644         intrinsic register names.
23645         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
23646         intrinsic register enum values and macros.
23648 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
23650         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
23651         for load/store addressing form.
23652         (nds32_print_operand_address): Likewise.
23654 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
23656         PR target/85196
23657         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
23658         based on LABEL_REF.  Remove useless assertion.
23659         (pic_address_needs_scratch): Fix formatting.
23660         (sparc_legitimize_pic_address): Minor tweaks.
23661         (sparc_delegitimize_address): Adjust assertion accordingly.
23662         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
23663         into symbolic_operand.
23664         (movsi_high_pic_label_ref): Likewise.
23665         (movsi_lo_sum_pic_label_ref): Likewise.
23666         (movdi_pic_label_ref): Likewise.
23667         (movdi_high_pic_label_ref): Likewise.
23668         (movdi_lo_sum_pic_label_ref): Likewise.
23670 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
23672         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
23673         custom LIB_SPEC setup.
23675 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
23676             Kito Cheng  <kito.cheng@gmail.com>
23678         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
23679         * config/riscv/freebsd.h: New.
23681 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
23683         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
23684         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
23685         file.
23687 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
23688             Kito Cheng  <kito.cheng@gmail.com>
23690         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
23691         nds32_output_call, nds32_symbol_binds_local_p): New functions.
23692         * config/nds32/nds32-protos.h (nds32_output_call,
23693         nds32_output_return): Declare.
23694         * config/nds32/nds32.md: Refine all the call and return patterns.
23696 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
23698         PR debug/85252
23699         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
23700         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
23702         PR rtl-optimization/84872
23703         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
23704         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
23705         EDGE_CROSSING edge.
23707 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
23709         * expr.c (copy_blkmode_to_reg): Revert 254862.
23710         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
23712 2018-04-06  Richard Biener  <rguenther@suse.de>
23714         PR middle-end/85244
23715         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
23716         after seeing a component reference with an adjacent field.  Treat
23717         refs to arrays at struct end of external decls similar to
23718         refs to unconstrained commons.
23720 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
23722         PR sanitizer/85213
23723         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
23724         look through SAVE_EXPRs with non-side-effects argument.  Adjust
23725         recursive calls.
23726         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
23727         save_p here.
23729 2018-04-06  Richard Biener  <rguenther@suse.de>
23731         PR middle-end/85180
23732         * alias.c (find_base_term): New wrapper around find_base_term
23733         unwinding CSELIB_VAL_PTR changes.
23734         (find_base_term): Do not restore CSELIB_VAL_PTR during the
23735         recursion.
23737 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23739         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
23740         instructions.
23741         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
23742         constant definitions.
23743         ("nop"): lr 0,0 -> nopr r0
23744         ("nop_lr0", "nop_lr1"): New insn definitions.
23746 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
23748         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
23749         NDS32_V3PUSH_AVAILABLE_P macro.
23751 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
23752             Chung-Ju Wu  <jasonwucj@gmail.com>
23754         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
23755         (nds32*-*-*): Add float and fpu_config into supported_defaults.
23756         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
23757         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
23758         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
23759         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
23760         * config/nds32/constraints.md: New constraints and checking for hard
23761         float configuration.
23762         * config/nds32/iterators.md: New mode iterator and attribute for hard
23763         float configuration.
23764         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
23765         patterns.
23766         * config/nds32/nds32-fpu.md: New file.
23767         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
23768         deal with hard float code generation.
23769         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
23770         ARCH_V3S.
23771         (abi_type, float_reg_number): New enum type.
23772         * config/nds32/nds32-predicates.c: New predicates for hard float.
23773         * config/nds32/nds32-protos.h: Declare functions for hard float.
23774         * config/nds32/nds32.c: Implementation for hard float configuration.
23775         * config/nds32/nds32.h: Definitions for hard float configuration.
23776         * config/nds32/nds32.md: Include hard float machine description and
23777         modify patterns for hard float configuration.
23778         * config/nds32/nds32.opt: New options for hard float configuration.
23779         * config/nds32/predicates.md: New predicates for hard float
23780         configuration.
23782 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
23784         * common/config/nds32/nds32-common.c
23785         (nds32_option_optimization_table): Enable -mreleax-hint by default.
23787 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
23789         PR middle-end/85195
23790         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
23791         CONSTRUCTOR_ELT (ctor, ...)->value.
23793 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
23795         PR target/85193
23796         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
23798 2018-04-05  Tom de Vries  <tom@codesourcery.com>
23800         PR target/85204
23801         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
23802         cond jump.
23804 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
23805             Kito Cheng  <kito.cheng@gmail.com>
23807         * config/nds32/constraints.md (U33): Fine-tune checking condition.
23808         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
23809         * config/nds32/nds32.h (nds32_16bit_address_type): Add
23810         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
23812 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
23813             Kito Cheng  <kito.cheng@gmail.com>
23815         * config/nds32/constraints.md (Ufe): New memory constraint.
23816         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
23817         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
23818         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
23819         operands.
23820         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
23821         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
23823 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23825         * config/nds32/nds32.md: Use optimize_size in the condition for
23826         alu-shift instructions.
23828 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23830         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
23832 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23834         * config/nds32/nds32.md (negsi2): Refine pattern.
23836 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
23837             Chung-Ju Wu  <jasonwucj@gmail.com>
23839         * config/nds32/iterators.md (shift_rotate): New code iterator.
23840         (shift): New code attribute.
23841         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
23842         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
23843         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
23844         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
23845         bit-wise operations.
23846         (andsi3, *andsi3): Ditto.
23847         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
23848         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
23849         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
23850         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
23851         nds32_ior_operand, nds32_xor_operand): New predicates.
23853 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23855         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
23856         (addsi3, subsi3): ... this.
23858 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23860         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
23862 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23864         * config/nds32/nds32.md: Adjust indention.
23866 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
23868         * config/nds32/nds32.md (feature): New attribute.
23870 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
23872         * config/nds32/nds32.md (subtype): New attribute.
23874 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
23876         PR target/85203
23877         * config/arm/arm-builtins.c (arm_expand_builtin): Change
23878         expansion to perform a bitwise AND of the argument followed by a
23879         boolean negation of the result.
23881 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
23883         PR rtl-optimization/84878
23884         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
23885         the basic block.  Assert the use reference is not artificial and that
23886         it has an associated insn.
23888 2018-04-04  Michael Matz  <matz@suse.de>
23890         * builtins.c (compute_objsize): Pass correct operand
23891         to array_at_struct_end_p.
23893 2018-04-04  Richard Biener  <rguenther@suse.de>
23895         PR lto/85176
23896         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
23897         from contexts for DINFO_LEVEL_TERSE and below.
23899 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
23901         * config/nds32/nds32-doubleword.md (move_<mode>): Require
23902         resiter_operand condition.
23903         * config/nds32/nds32.md (*move<mode>): Ditto.
23905 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
23906             Monk Chiang  <sh.chiang04@gmail.com>
23908         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
23910 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
23912         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
23914 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
23915             Kito Cheng  <kito.cheng@gmail.com>
23917         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
23918         nds32_cond_code_str, output_cond_branch,
23919         output_cond_branch_compare_zero, nds32_expand_cbranch,
23920         nds32_expand_cstore, nds32_expand_movcc,
23921         nds32_output_cbranchsi4_equality_zero,
23922         nds32_output_cbranchsi4_equality_reg,
23923         nds32_output_cbranchsi4_equality_reg_or_const_int,
23924         nds32_output_cbranchsi4_greater_less_zero: New functions.
23925         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
23926         nds32_expand_cstore, nds32_expand_movcc,
23927         nds32_output_cbranchsi4_equality_zero,
23928         nds32_output_cbranchsi4_equality_reg,
23929         nds32_output_cbranchsi4_equality_reg_or_const_int,
23930         nds32_output_cbranchsi4_greater_less_zero): Declare.
23931         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
23932         nds32_rimm11s_operand): New predicates.
23933         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
23934         * config/nds32/nds32.md: Rewrite all the branch and conditional move
23935         patterns.
23937 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
23939         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
23940         * config/nds32/nds32.md: Ditto.
23941         * config/nds32/pipelines.md: Ditto.
23943 2018-04-04  Richard Biener  <rguenther@suse.de>
23945         PR tree-optimization/85168
23946         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
23947         propagating abnormals.
23949 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
23951         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
23953 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
23954             Kito Cheng  <kito.cheng@gmail.com>
23956         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
23957         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
23958         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
23959         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
23960         * config/nds32/nds32.md (sibcall_internal): New.
23961         (sibcall_register): Remove.
23962         (sibcall_immediate): Remove.
23963         (sibcall_value_internal): New.
23964         (sibcall_value_register): Remove.
23965         (sibcall_value_immediate): Remove.
23966         * config/nds32/predicates.md (nds32_general_register_operand): New.
23967         (nds32_call_address_operand): New.
23969 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
23971         PR rtl-optimization/85167
23972         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
23973         bb_defs if *split_p, instead preinitialize it to NULL.
23975         PR tree-optimization/85156
23976         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
23977         evaluating the argument multiple times.
23979 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
23981         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
23982         than vector.
23983         (_mm_cvtpd_ps): Likewise.
23984         (_mm_cvttpd_epi32): Likewise.
23985         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
23986         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
23987         vector, pixel, and bool following altivec.h include.
23989 2018-04-03  Martin Sebor  <msebor@redhat.com>
23991         * doc/extend.texi (Common Function Attributes): Clarify.
23992         (const attribute): Likewise.
23993         (pure attribute): Likewise.
23995 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
23997         PR target/85169
23998         * config/i386/i386.c (ix86_expand_vector_set): Use
23999         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
24001 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
24003         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
24004         instructions when changing rounding bits to preserve precision bits
24005         in the x87 control word.
24007 2018-04-03  Martin Liska  <mliska@suse.cz>
24009         PR tree-optimization/82491
24010         * rtl.h (strip_offset_and_add): Replace += suboffset with
24011         poly_uint64 () + suboffset.
24013 2018-03-29  Martin Liska  <mliska@suse.cz>
24014             Martin Jambor  <mjambor@suse.cz>
24016         PR ipa/84947
24017         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
24018         param_type is not an integral or pointer type.
24020 2018-04-03  Richard Biener  <rguenther@suse.de>
24022         * sese.h (recompute_all_dominators): Remove.
24024 2018-04-02  Martin Sebor  <msebor@redhat.com>
24026         * doc/invoke.texi (-Wrestrict): Fix typos.
24028 2018-04-02  Jim Wilson  <jimw@sifive.com>
24030         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
24031         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
24032         (<optab>di3, <optab>si3_extend): Likewise.
24033         (<optab>si3_mask, <optab>si3_mask_1): New.
24034         (<optab>di3_mask, <optab>di3_mask_1): New.
24035         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
24036         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
24037         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
24039 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
24041         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
24042         example.
24044 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
24046         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
24047         (nds32_canonicalize_comparison): New function.
24049 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
24050             Kito Cheng  <kito.cheng@gmail.com>
24051             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
24053         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
24054         * config/nds32/constants.md (unspec_volatile_element): Add
24055         UNSPEC_VOLATILE_RELAX_GROUP.
24056         * config/nds32/nds32-relax-opt.c: New file.
24057         * config/nds32/nds32-predicates.c
24058         (nds32_symbol_load_store_p): New function.
24059         * config/nds32/nds32-protos.h
24060         (nds32_symbol_load_store_p): Declare function.
24061         (make_pass_nds32_relax_opt): Declare new rtl pass function.
24062         * config/nds32/nds32.c
24063         (nds32_register_pass): New function to register pass.
24064         (nds32_register_passes): New function to register passes.
24065         * config/nds32/nds32.md (relax_group): New pattern.
24066         * config/nds32/nds32.opt (mrelax-hint): New option.
24067         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
24069 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
24071         * config/nds32/t-nds32: Modify files dependency.
24073 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
24075         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
24076         (PROFILE_HOOK): Define its implementation.
24078 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
24080         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
24081         type and 32-bit size.
24083 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
24085         PR middle-end/85090
24086         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
24087         (V_128_256): New mode iterator.
24088         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
24089         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
24090         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
24091         of V.
24092         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
24093         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
24095 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
24097         PR target/83315
24098         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
24099         NaN inputs correctly.
24101 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
24103         PR target/80546
24104         * config/rs6000/vsx.md (??r): New mode attribute.
24105         (*vsx_mov<mode>_64bit): Use it.
24106         (*vsx_mov<mode>_32bit): Likewise.
24108 2018-03-30  Martin Sebor  <msebor@redhat.com>
24110         PR tree-optimization/84818
24111         * builtins.c (check_access): Use warning_n.
24113 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24115         PR target/83822
24116         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
24117         condition.
24118         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
24119         condition.
24121 2018-03-30  Julia Koval  <julia.koval@intel.com>
24123         PR target/84413
24124         * x86-tune.def (movx, partial_reg_dependency): Enable for
24125         m_SKYLAKE_AVX512.
24127 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
24129         PR inline-asm/84985
24130         * lra-constraints.c (process_alt_operands): Move setting
24131         this_alternative_matches below.
24133 2018-03-29  Martin Liska  <mliska@suse.cz>
24135         PR lto/84995.
24136         * doc/invoke.texi: Document how LTO works with debug info.
24137         Describe auto-load support of binutils.  Mention 'x86-64'
24138         as valid option value of -march option.
24140 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
24142         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
24144         PR c/85094
24145         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
24146         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
24147         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
24148         checking.
24150 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
24152         PR target/84912
24153         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
24154         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
24155         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
24156         for RS6000_BTM_POWERPC64.
24157         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
24158         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
24159         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
24160         definition.
24161         (DIVDE): Use it.
24162         (DIVDEU): Likewise.
24164 2018-03-28  Carl Love  <cel@us.ibm.com>
24166         Revert
24167         2017-09-27  Carl Love  <cel@us.ibm.com>
24169         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
24170         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
24171         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
24172         fctiw instruction.
24174 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24176         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
24177         instead of __vector bool.
24178         (_mm_max_pu8): Likewise.
24179         (_mm_min_pi16): Likewise.
24181 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
24183         PR target/84912
24184         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
24185         (DIVWEUO): Likewise.
24186         (DIVDEO): Likewise.
24187         (DIVDEUO): Likewise.
24188         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
24189         DIVWEUO and DIVDEUO.
24190         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
24191         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
24192         (div_extend): Likewise.
24193         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
24194         builtin function.
24195         (__builtin_divweuo): Likewise.
24196         (__builtin_divdeo): Likewise.
24197         (__builtin_divdeuo): Likewise.
24199 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
24201         PR target/85095
24202         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
24203         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
24205         PR tree-optimization/82004
24206         * gimple-match-head.c (optimize_pow_to_exp): New function.
24207         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
24208         Don't fold to exp if optimize_pow_to_exp is false.
24210 2018-03-28  Martin Liska  <mliska@suse.cz>
24212         PR other/84819
24213         * calls.c (initialize_argument_information): Fix trailing space.
24214         * common.opt: Fix typo and provide better explanation for
24215         -fsanitize-coverage option.
24216         * config/i386/i386.opt: Fix typo.
24218 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
24219             Martin Liska  <mliska@suse.cz>
24221         PR sanitizer/85081
24222         * gimplify.c (asan_poison_variable): Don't do the check for
24223         gimplify_omp_ctxp here.
24224         (gimplify_decl_expr): Do it here.
24225         (gimplify_target_expr): Likewise.
24227 2018-03-28  Martin Liska  <mliska@suse.cz>
24229         PR target/84988
24230         * config/i386/i386.c (ix86_function_arg_advance): Do not call
24231         chkp_type_bounds_count if MPX is not enabled.
24233 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
24235         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
24237 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
24239         PR target/84914
24240         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
24241         function to create the function decl for complex long double
24242         multiply and divide for -mabi=ieeelongdouble.
24243         (init_float128_ieee): Call it.
24245 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
24247         PR target/85044
24248         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
24249         -fcf-protection=branch -mibt.
24250         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
24252 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
24254         PR target/81863
24255         * config/arm/arm.c (arm_valid_symbolic_address): Handle
24256         arm_word_relocations.
24258 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
24260         PR target/85056
24261         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
24262         extern array declarations.
24264 2018-03-27  Richard Biener  <rguenther@suse.de>
24266         PR middle-end/84067
24267         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
24268         explicit single_use checks.
24270 2018-03-27  Richard Biener  <rguenther@suse.de>
24272         PR tree-optimization/85082
24273         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
24274         Valueize the VUSE.
24276 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
24278         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
24279         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
24280         Turn on fasynchronous-unwind-tables and funwind-tables.
24282 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
24284         PR target/85073
24285         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
24286         (*bmi_blsr_<mode>_ccz): Ditto.
24288 2018-03-26  Tom de Vries  <tom@codesourcery.com>
24290         PR tree-optimization/85063
24291         * omp-general.c (offloading_function_p): New function.  Factor out
24292         of ...
24293         * omp-offload.c (pass_omp_target_link::gate): ... here.
24294         * omp-general.h (offloading_function_p): Declare.
24295         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
24296         with attribute omp declare target for offloading functions.
24298 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
24300         PR tree-optimization/84005
24301         * tree-data-ref.h (get_base_for_alignment): Declare.
24302         * tree-data-ref.c (get_base_for_alignment_1): New function.
24303         (get_base_for_alignment): Likewise.
24304         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
24305         get_base_for_alignment to find a suitable base object, instead
24306         of always using drb->base_address.
24308 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
24310         PR inline-asm/85022
24311         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
24312         known size by default.
24314 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
24316         PR inline-asm/85030
24317         * lra-constraints.c (process_alt_operands): Don't match BLKmode
24318         and non BLKmode operands.
24320 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24322         PR target/85026
24323         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
24324         Clean up attributes.
24326 2018-03-23  Richard Biener  <rguenther@suse.de>
24328         PR debug/85020
24329         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
24330         we are going to emit early debug for LTO.
24332 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
24334         PR inline-asm/85034
24335         * function.c (match_asm_constraints_1): Don't optimize if input
24336         doesn't satisfy general_operand predicate for output's mode.
24338         PR inline-asm/85022
24339         * alias.c (write_dependence_p): Don't require for x_canonicalized
24340         non-VOIDmode if x has VOIDmode.
24342         PR sanitizer/85029
24343         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
24344         just don't try to optimize it rather than assert it never happens.
24346 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24348         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
24349         macro expansions for definition of ST_INTERNAL_<mode> and
24350         LD_INTERNAL_<mode> builtins.
24351         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
24352         Remove prototype.
24353         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
24354         function.
24355         (altivec_expand_st_builtin): Likewise.
24356         (altivec_expand_builtin): Remove calls to deleted functions.
24357         (rs6000_address_for_altivec): Delete this function.
24358         * config/rs6000/vector.md: Remove expands for
24359         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
24361 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
24363         PR target/84826
24364         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
24365         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
24366         re-computing once computed.
24367         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
24368         (arm_init_machine_status): Initialize
24369         machine->static_chain_stack_bytes.
24371 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24373         PR target/84760
24374         * doc/extend.texi: Add four new prototypes for vec_ld.
24375         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
24376         definitions for more logical presentation.
24377         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
24378         entries for V1TI variants of __builtin_altivec_ld builtin.
24379         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
24380         handling of V1TI variant of LVX icode pattern.
24381         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
24382         (rs6000_gimple_fold_builtin): Likewise.
24383         (altivec_init_builtins): Add code to define
24384         __builtin_altivec_lvx_v1ti function.
24386 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
24388         PR inline-asm/84941
24389         * function.c (match_asm_constraints_1): Don't do the optimization
24390         if input isn't a REG, SUBREG, MEM or constant.
24392 2018-03-22  Tom de Vries  <tom@codesourcery.com>
24394         PR tree-optimization/84956
24395         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
24396         bb_has_abnormal_pred.
24398 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
24400         PR sanitizer/85018
24401         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
24402         DECL_INITIAL (decl) to decl at the end.
24403         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
24404         adjust the comment.
24406 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
24408         * doc/extend.texi (__builtin_tgmath): Document when complex
24409         integer types are treated as _Complex _Float64.
24411 2018-03-21  Tom de Vries  <tom@codesourcery.com>
24413         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
24415 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
24417         PR tree-optimization/84960
24418         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
24419         if it is ENTRY block, move them into single succ of ENTRY in that case.
24421 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
24423         PR tree-optimization/84811
24424         * poly-int.h (poly_span_traits): Remove the T3 parameter and
24425         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
24426         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
24427         (known_subrange_p): Update accordingly.  Cast each value involved
24428         in the size comparison, rather than casting the result of the
24429         subtraction.
24431 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
24433         PR tree-optimization/84982
24434         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
24435         by flipping the least significant bit rather than all bits from
24436         bitpos to bitpos + bitsize - 1.
24438 2018-03-21  Nathan Sidwell  <nathan@acm.org>
24440         * doc/extend.texi (Deprecated Features): Remove mention of
24441         long-deleted deprecations.
24443 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24445         PR jit/84288
24446         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
24447         * configure: Regenerate.
24449 2018-03-21  Tom de Vries  <tom@codesourcery.com>
24451         PR tree-optimization/83126
24452         * tree-parloops.c (num_phis): New function.
24453         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
24455 2018-03-21  Nathan Sidwell  <nathan@acm.org>
24457         * doc/extend.texi (Deprecated Features): Update deprecated flags,
24458         mention anon-struct/union members and trailing attributes.
24460 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
24462         PR tree-optimization/84969
24463         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
24464         builtin memset partitions if they set different rhs values.
24466 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
24468         PR rtl-optimization/84989
24469         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
24470         VEC_DUPLICATE with scalar result mode.
24472 2018-03-21  Martin Liska  <mliska@suse.cz>
24474         PR ipa/84963
24475         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
24476         not intended return statement.
24478 2018-03-21  Martin Liska  <mliska@suse.cz>
24480         PR target/84988
24481         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
24482         (chkp_find_bound_slots_1): Limit number of iterations.
24484 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
24486         PR target/84838
24487         * Minor grammar fixes for x86 options.
24489 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
24491         PR debug/84875
24492         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
24493         holding REG_CFA_RESTORE notes, instead turn them into a USE.
24495 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
24497         PR target/83789
24498         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
24499         (altivec_lvx_<mode>_1op): Likewise.
24500         (altivec_stvx_<mode>_2op): Likewise.
24501         (altivec_stvx_<mode>_1op): Likewise.
24502         (altivec_lvx_<VM2:mode>): New define_expand.
24503         (altivec_stvx_<VM2:mode>): Likewise.
24504         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
24505         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
24506         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
24507         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
24508         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
24509         (rs6000_gen_lvx): Likewise.
24510         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
24511         (altivec_expand_stv_builtin): Likewise.
24512         (altivec_expand_builtin): Likewise.
24513         * config/rs6000/vector.md: Likewise.
24515 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24517         PR target/82518
24518         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
24519         BYTES_BIG_ENDIAN.
24521 2018-03-20  Richard Biener  <rguenther@suse.de>
24523         PR target/84986
24524         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
24525         sign-conversions as zero, fall back to standard scalar_stmt
24526         cost for the rest.
24528 2018-03-20  Martin Liska  <mliska@suse.cz>
24530         PR ipa/84825
24531         * predict.c (rebuild_frequencies): Handle case when we have
24532         PROFILE_ABSENT, but flag_guess_branch_prob is false.
24534 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
24536         PR target/84990
24537         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
24538         flag_section_anchors.
24539         * varasm.c (use_blocks_for_decl_p): Remove hack for
24540         dw2_force_const_mem.
24542         PR target/84845
24543         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
24544         to ...
24545         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
24546         be created, use lowpart_subreg of operands[0] rather than operands[0]
24547         itself.
24548         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
24549         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
24550         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
24551         and n constraint instead of aarch64_shift_imm_di and Usd.
24552         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
24553         (*aarch64_<optab>_reg_minus<mode>3): ... this.
24555 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
24557         PR target/82989
24558         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
24559         to favor GPR over NEON registers.
24560         (<shift>di3_neon): Likewise.
24562 2018-03-20  Tom de Vries  <tom@codesourcery.com>
24564         PR target/84952
24565         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
24566         (nvptx_process_pars): Emit bar.sync asap and alap.
24568 2018-03-20  Tom de Vries  <tom@codesourcery.com>
24570         PR target/84954
24571         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
24572         seen_label if seen_label is already set.
24574 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
24576         PR target/84945
24577         * config/i386/i386.c (fold_builtin_cpu): For features above 31
24578         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
24579         Use 1U instead of 1.  Formatting fixes.
24581         PR c/84953
24582         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
24583         instead of TREE_TYPE (s1) for the return value.
24585 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
24587         PR tree-optimization/84946
24588         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
24589         bitsize + bitsize in poly_uint64 rather than poly_int64.
24591         PR sanitizer/78651
24592         * dwarf2asm.c: Include fold-const.c.
24593         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
24594         of decl rather than decl itself.
24596         PR rtl-optimization/84643
24597         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
24599 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
24601         PR sanitizer/78651
24602         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
24603         calling assemble_variable.
24605 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
24607         PR target/81647
24608         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
24609         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
24611 2018-03-19  Jim Wilson  <jimw@sifive.com>
24613         PR bootstrap/84856
24614         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
24615         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
24616         (riscv_first_stack_step): Likewise.
24617         (riscv_option_override): Use STACK_BOUNDARY instead of
24618         MIN_STACK_BOUNDARY.
24619         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
24620         MIN_STACK_BOUNDARY.
24621         (BIGGEST_ALIGNMENT): Set to 128.
24622         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
24623         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
24624         STACK_BOUNDARY.
24626 2018-03-19  Richard Biener  <rguenther@suse.de>
24628         PR tree-optimization/84933
24629         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
24630         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
24632 2018-03-19  Richard Biener  <rguenther@suse.de>
24634         PR tree-optimization/84859
24635         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
24636         (cond_if_else_store_replacement): Perform sinking operation on
24637         single-store BBs regardless of MAX_STORES_TO_SINK setting.
24638         Generalize what a BB with a single eligible store is.
24640 2018-03-19  Richard Biener  <rguenther@suse.de>
24642         PR tree-optimization/84929
24643         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
24644         chrec_is_positive against non-chrec arg.
24646 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
24648         PR target/84711
24649         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
24651 2018-03-18  Martin Liska  <mliska@suse.cz>
24653         PR rtl-optimization/84635
24654         * regrename.c (build_def_use): Use matches_mode only when
24655         matches >= 0.
24657 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
24659         PR tree-optimization/84913
24660         * tree-vect-loop.c (vectorizable_reduction): Don't try to
24661         vectorize chains of COND_EXPRs.
24663 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
24665         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
24667 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
24669         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
24671 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
24673         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
24675 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
24676             Kito Cheng  <kito.cheng@gmail.com>
24678         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
24679         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
24680         (nds32_adjust_reg_alloc_order): New function.
24681         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
24683 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
24685         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
24686         nds32_print_operand, nds32_print_operand_address): Use
24687         HOST_WIDE_INT_PRINT_DEC instead.
24689 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
24691         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
24693 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
24695         PR target/84902
24696         * config/i386/i386.c (initial_ix86_tune_features,
24697         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
24698         unsigned long long.
24699         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
24700         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
24701         rather than 1u << ix86_tune.  Formatting fix.
24702         (ix86_option_override_internal): Change ix86_arch_mask from
24703         unsigned int to unsigned HOST_WIDE_INT, initialize to
24704         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
24705         (ix86_function_specific_restore): Likewise.
24707 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
24709         PR target/84899
24710         * postreload.c (reload_combine_recognize_pattern): Perform
24711         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
24712         truncate_int_for_mode the result for the destination's mode.
24714         PR c/84909
24715         * hsa-gen.c (mem_type_for_type): Fix comment typo.
24716         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
24717         Likewise.
24718         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
24719         Likewise.
24721 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
24723         PR target/84876
24724         * lra-assigns.c (lra_split_hard_reg_for): Don't use
24725         regno_allocno_class_array and sorted_pseudos.
24726         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
24727         insns where regno is used.
24729 2018-03-16  Martin Liska  <mliska@suse.cz>
24731         PR ipa/84833
24732         * multiple_target.c (create_dispatcher_calls): Redirect
24733         reference in the symbol table.
24735 2018-03-16  Martin Liska  <mliska@suse.cz>
24737         PR ipa/84722
24738         * multiple_target.c (create_dispatcher_calls): Redirect also
24739         an alias.
24741 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
24743         PR c++/79937
24744         PR c++/82410
24745         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
24746         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
24747         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
24749 2018-03-16  Julia Koval  <julia.koval@intel.com>
24751         * doc/invoke.texi (Skylake Server): Add CLWB.
24752         Cannonlake): Remove CLWB.
24754 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
24756         PR tree-optimization/84841
24757         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
24758         1 << 3.
24759         (FLOAT_ONE_CONST_TYPE): Define.
24760         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
24761         (sort_by_operand_rank): Put entries with higher constant_type last
24762         rather than first to match comments.
24764 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
24766         * config/nios2/nios2.md (movsi_internal): Fix thinko in
24767         split predicate.
24769 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
24771         PR c++/79085
24772         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
24773         check and use address of target always.
24775 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
24777         PR target/84574
24778         * config/i386/i386.c (indirect_thunk_needed): Update comments.
24779         (indirect_thunk_bnd_needed): Likewise.
24780         (indirect_thunks_used): Likewise.
24781         (indirect_thunks_bnd_used): Likewise.
24782         (indirect_return_needed): New.
24783         (indirect_return_bnd_needed): Likewise.
24784         (output_indirect_thunk_function): Add a bool argument for
24785         function return.
24786         (output_indirect_thunk_function): Don't generate alias for
24787         function return thunk.
24788         (ix86_code_end): Call output_indirect_thunk_function to generate
24789         function return thunks.
24790         (ix86_output_function_return): Set indirect_return_bnd_needed
24791         and indirect_return_needed instead of indirect_thunk_bnd_needed
24792         and indirect_thunk_needed.
24794 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
24796         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
24797         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
24798         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
24800 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
24801             Paul Hua <paul.hua.gm@gmail.com>
24803         PR c/84852
24804         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
24806 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
24808         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
24809         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
24810         resp. SFmode cases.
24812 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
24814         PR target/84711
24815         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
24816         instead of GET_MODE_SIZE when comparing Units.
24818 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
24820         PR target/68256
24821         * varasm.c (hash_section): Return an unchangeble hash value
24822         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
24823         Return !aarch64_can_use_per_function_literal_pools_p ().
24825 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
24827         PR target/84860
24828         * optabs.c (emit_conditional_move): Pass address of cmode's copy
24829         rather than address of cmode as last argument to prepare_cmp_insn.
24831 2018-03-15  Julia Koval  <julia.koval@intel.com>
24833         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
24834         F_AVX512VNNI, F_AVX512BITALG): New.
24836 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
24838         PR target/83451
24839         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
24840         insn for floating-point loads and stores.
24842 2018-03-14  Carl Love  <cel@us.ibm.com>
24844         * config/rs6000/rs6000-c.c: Add macro definitions for
24845         ALTIVEC_BUILTIN_VEC_PERMXOR.
24846         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
24847         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
24848         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
24849         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
24850         UNSPEC_VPERMXOR.
24851         * config/doc/extend.texi: Add prototypes for vec_permxor.
24853 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
24855         PR c/84852
24856         * diagnostic-show-locus.c (class layout_point): Convert m_line
24857         from int to linenum_type.
24858         (line_span::comparator): Use linenum "compare" function when
24859         comparing line numbers.
24860         (test_line_span): New function.
24861         (layout_range::contains_point): Convert param "row" from int to
24862         linenum_type.
24863         (layout_range::intersects_line_p): Likewise.
24864         (layout::will_show_line_p): Likewise.
24865         (layout::print_source_line): Likewise.
24866         (layout::should_print_annotation_line_p): Likewise.
24867         (layout::print_annotation_line): Likewise.
24868         (layout::print_leading_fixits): Likewise.
24869         (layout::annotation_line_showed_range_p): Likewise.
24870         (struct line_corrections): Likewise for field m_row.
24871         (line_corrections::line_corrections): Likewise for param "row".
24872         (layout::print_trailing_fixits): Likewise.
24873         (layout::get_state_at_point): Likewise.
24874         (layout::get_x_bound_for_row): Likewise.
24875         (layout::print_line): Likewise.
24876         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
24877         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
24878         * input.c (selftest::test_linenum_comparisons): New function.
24879         (selftest::input_c_tests): Call it.
24880         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
24881         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
24882         * selftest.h (ASSERT_GT): New macro.
24883         (ASSERT_GT_AT): New macro.
24884         (ASSERT_LT): New macro.
24885         (ASSERT_LT_AT): New macro.
24887 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
24889         PR rtl-optimization/84780
24890         * combine.c (distribute_links): Don't make a link based on pc_rtx.
24892 2018-03-14  Martin Liska  <mliska@suse.cz>
24894         * tree.c (record_node_allocation_statistics): Use
24895         get_stats_node_kind.
24896         (get_stats_node_kind): New function extracted from
24897         record_node_allocation_statistics.
24898         (free_node): Use get_stats_node_kind.
24900 2018-03-14  Richard Biener  <rguenther@suse.de>
24902         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
24903         that the value-set of ANTIC_IN doesn't grow.
24905         Revert
24906         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
24907         member.
24908         (BB_VISITED_WITH_VISITED_SUCCS): New define.
24909         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
24911 2018-03-14  Julia Koval  <julia.koval@intel.com>
24913         * config.gcc (icelake-client, icelake-server): New.
24914         (icelake): Remove.
24915         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
24916         (initial_ix86_arch_features): Ditto.
24917         (PTA_SKYLAKE): Add SGX.
24918         (PTA_ICELAKE): Remove.
24919         (PTA_ICELAKE_CLIENT): New.
24920         (PTA_ICELAKE_SERVER): New.
24921         (ix86_option_override_internal): Split up icelake on icelake client and
24922         icelake server.
24923         (get_builtin_code_for_version): Ditto.
24924         (fold_builtin_cpu): Ditto.
24925         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
24926         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
24927         * config/i386/i386.h (processor_type): Ditto.
24928         * doc/invoke.texi: Ditto.
24930 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
24932         PR sanitizer/83392
24933         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
24934         INTEGER_CST offset, add it together with bitpos / 8 and
24935         sign extend based on POINTER_SIZE.
24937         PR target/84844
24938         Revert
24939         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
24941         PR target/78090
24942         * config/i386/constraints.md (Yc): New register constraint.
24943         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
24944         Use Yc constraint for alternative 2 of operand 0.  Remove
24945         preferred_for_speed attribute.
24947 2018-03-14  Richard Biener  <rguenther@suse.de>
24949         PR tree-optimization/84830
24950         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
24951         with the old one to avoid oscillations.
24953 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
24955         PR target/83712
24956         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
24957         pseudos.
24958         (assign_by_spills): Return a flag of reload assignment failure.
24959         Do not process the reload assignment failures.  Do not spill other
24960         reload pseudos if they has the same reg class.  Update n if
24961         necessary.
24962         (lra_assign): Add a return arg.  Set up from the result of
24963         assign_by_spills call.
24964         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24965         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
24966         usage_insns if it is not NULL.
24967         (spill_hard_reg_in_range): New function.
24968         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24969         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24970         function prototypes.
24971         (lra_assign): Change prototype.
24972         * lra.c (lra): Add code to deal with fails by splitting hard reg
24973         live ranges.
24975 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
24977         * config/riscv/riscv.opt (mrelax): New option.
24978         * config/riscv/riscv.c (riscv_file_start): Emit ".option
24979         "norelax" when riscv_mrelax is disabled.
24980         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
24982 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
24984         PR target/84743
24985         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
24986         reassociation for int modes.
24988 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
24990         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
24991         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
24992         for big-endian.
24993         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
24994         * config/aarch64/aarch64-sve.md
24995         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
24996         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
24997         (*extend<mode><Vwide>2): Rename to...
24998         (aarch64_sve_extend<mode><Vwide>2): ...this.
24999         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
25000         renaming the old pattern to...
25001         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
25002         unsigned packs.
25003         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
25004         define_expand, renaming the old pattern to...
25005         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
25006         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
25007         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
25008         account when deciding which SVE instruction the optab should use.
25009         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
25011 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
25013         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
25014         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
25015         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
25016         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
25017         (tlsdesc_small_<mode>): Turn a define_expand and use
25018         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
25019         (tlsdesc_small_advsimd_<mode>): ...this.
25020         (tlsdesc_small_sve_<mode>): New pattern.
25022 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
25024         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
25025         (UNSPEC_UMUL_HIGHPART): New constants.
25026         (MUL_HIGHPART): New int iteraor.
25027         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
25028         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
25029         define_expand.
25030         (*<su>mul<mode>3_highpart): New define_insn.
25032 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
25034         PR lto/84805
25035         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
25036         incomplete types.
25038 2018-03-13  Martin Liska  <mliska@suse.cz>
25040         PR ipa/84658.
25041         * (sem_item_optimizer::sem_item_optimizer): Initialize new
25042         vector.
25043         (sem_item_optimizer::~sem_item_optimizer): Release it.
25044         (sem_item_optimizer::merge_classes): Register variable aliases.
25045         (sem_item_optimizer::fixup_pt_set): New function.
25046         (sem_item_optimizer::fixup_points_to_sets): Likewise.
25047         * ipa-icf.h: Declare new variables and functions.
25049 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
25051         PR middle-end/84834
25052         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
25053         integer_pow2p@2 and test integer_pow2p in condition.
25054         (A < 0 ? C : 0): Similarly for @1.
25056         PR middle-end/84831
25057         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
25058         characters starting at p contain '\0' character, don't look beyond
25059         that.
25061         PR target/84827
25062         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
25063         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
25065         PR target/84828
25066         * reg-stack.c (change_stack): Change update_end var from int to
25067         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
25068         also call set_block_for_insn on the newly added insns and rescan.
25070         PR target/84786
25071         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
25072         on the last operand.
25074         PR c++/84704
25075         * tree.c (stabilize_reference_1): Return save_expr (e) for
25076         STATEMENT_LIST even if it doesn't have side-effects.
25078 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
25080         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
25082 2018-03-12  Renlin Li  <renlin.li@arm.com>
25084         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
25085         aarch64_output_scalar_simd_mov_immediate.
25087 2018-03-12  Martin Sebor  <msebor@redhat.com>
25089         PR tree-optimization/83456
25090         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
25091         for perfectly overlapping calls to memcpy.
25092         (gimple_fold_builtin_memory_chk): Same.
25093         (gimple_fold_builtin_strcpy): Handle no-warning.
25094         (gimple_fold_builtin_stxcpy_chk): Same.
25095         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
25097 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
25099         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
25100         parameter.  Use it for SFmode.
25101         (rs6000_function_arg_advance_1): Adjust.
25102         (rs6000_function_arg): Adjust.
25103         (rs6000_gimplify_va_arg): Pass false for that new parameter.
25105 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
25107         PR rtl-optimization/84169
25108         PR rtl-optimization/84780
25109         * combine.c (can_combine_p): Check for a 2-insn combination whether
25110         the destination register is used between the two insns, too.
25112 2018-03-12  Richard Biener  <rguenther@suse.de>
25114         PR tree-optimization/84803
25115         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
25116         for refs DR analysis didn't process.
25118 2018-03-12  Richard Biener  <rguenther@suse.de>
25120         PR tree-optimization/84777
25121         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
25122         force-vectorize loops ignore whether we are optimizing for size.
25124 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
25126         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
25127         (TARGET_MD_ASM_ADJUST): Define.
25129 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
25130             Kito Cheng  <kito.cheng@gmail.com>
25131             Chung-Ju Wu  <jasonwucj@gmail.com>
25133         * config/nds32/nds32.c (nds32_compute_stack_frame,
25134         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
25135         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
25136         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
25137         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
25138         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
25139         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
25140         * config/nds32/nds32.md (prologue, epilogue): Use macro
25141         NDS32_V3PUSH_AVAILABLE_P to do checking.
25143 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
25145         PR debug/58150
25146         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
25147         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
25148         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
25149         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
25150         addition of most attributes on !orig_type_die or the attribute not
25151         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
25153 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
25154             Chung-Ju Wu  <jasonwucj@gmail.com>
25156         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
25157         __NDS32_VH__ macro.
25158         * config/nds32/nds32.opt (mvh): New option.
25160 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
25161             Chung-Ju Wu  <jasonwucj@gmail.com>
25163         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
25164         function.
25165         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
25166         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
25167         definition.
25169 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
25170             Chung-Ju Wu  <jasonwucj@gmail.com>
25172         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
25173         function.
25174         * config/nds32/nds32-multiple.md (strlensi): New pattern.
25175         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
25177 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
25178             Kito Cheng  <kito.cheng@gmail.com>
25179             Chung-Ju Wu  <jasonwucj@gmail.com>
25181         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
25182         UNSPEC_FFMISM and UNSPEC_FLMISM.
25183         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
25184         for ffb, ffmism and flmism.
25185         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
25186         (unspec_ffmism): Ditto.
25187         (unspec_flmism): Ditto.
25188         (nds32_expand_builtin_impl): Check if string extension is available.
25189         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
25190         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
25192 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
25194         Reverting patch:
25195         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
25197         PR target/83712
25198         * lra-assigns.c (assign_by_spills): Return a flag of reload
25199         assignment failure.  Do not process the reload assignment
25200         failures.  Do not spill other reload pseudos if they has the same
25201         reg class.
25202         (lra_assign): Add a return arg.  Set up from the result of
25203         assign_by_spills call.
25204         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
25205         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
25206         usage_insns if it is not NULL.
25207         (spill_hard_reg_in_range): New function.
25208         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
25209         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
25210         function prototypes.
25211         (lra_assign): Change prototype.
25212         * lra.c (lra): Add code to deal with fails by splitting hard reg
25213         live ranges.
25215 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
25217         PR target/84807
25218         * config/i386/i386.opt: Replace Enforcment with Enforcement.
25220 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
25222         PR debug/84620
25223         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
25224         (dw_val_node): Add val_symbolic_view.
25225         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
25226         (symview_upper_bound): New.
25227         (new_line_info_table): Initialize symviews_since_reset.
25228         (dwarf2out_source_line): Count symviews_since_reset and set
25229         symview_upper_bound.
25230         (dw_val_equal_p): Handle symview.
25231         (add_AT_symview): New.
25232         (print_dw_val): Handle symview.
25233         (attr_checksum, attr_checksum_ordered): Likewise.
25234         (same_dw_val_p, size_of_die): Likewise.
25235         (value_format, output_die): Likewise.
25236         (add_high_low_attributes): Use add_AT_symview for entry_view.
25237         (dwarf2out_finish): Reset symview_upper_bound, clear
25238         zero_view_p.
25240 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
25242         PR target/83969
25243         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
25244         Add strict argument and use it.
25245         (rs6000_split_multireg_move): Update for new strict argument.
25246         (mem_operand_gpr): Disallow all non-offsettable addresses.
25247         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
25249 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
25251         PR target/84772
25252         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
25253         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
25254         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
25256         PR c++/84767
25257         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
25258         decl, use remap_type if we want to use the type.
25260 2018-03-09  Martin Sebor  <msebor@redhat.com>
25262         PR tree-optimization/84526
25263         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
25264         Remove dead code.
25265         (builtin_access::generic_overlap): Be prepared to handle non-array
25266         base objects.
25268 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
25270         PR rtl-optimization/84682
25271         * lra-constraints.c (process_address_1): Check is_address flag
25272         for address constraints.
25273         (process_alt_operands): Likewise.
25274         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
25275         preprocess_constraints.
25276         * recog.h (preprocess_constraints): Add oploc parameter.
25277         Adjust callers.
25278         * recog.c (preprocess_constraints): Test address_operand for
25279         CT_ADDRESS constraints.
25281 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
25283         PR target/83712
25284         * lra-assigns.c (assign_by_spills): Return a flag of reload
25285         assignment failure.  Do not process the reload assignment
25286         failures.  Do not spill other reload pseudos if they has the same
25287         reg class.
25288         (lra_assign): Add a return arg.  Set up from the result of
25289         assign_by_spills call.
25290         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
25291         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
25292         usage_insns if it is not NULL.
25293         (spill_hard_reg_in_range): New function.
25294         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
25295         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
25296         function prototypes.
25297         (lra_assign): Change prototype.
25298         * lra.c (lra): Add code to deal with fails by splitting hard reg
25299         live ranges.
25301 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25303         PR target/83193
25304         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
25305         Accept complain bool parameter.  Only emit errors if it is true.
25306         (arm_parse_cpu_option_name): Likewise.
25307         (arm_target_thumb_only): Adjust callers of the above.
25308         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
25309         prototype to take a default true bool parameter.
25310         (arm_parse_arch_option_name): Likewise.
25312 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
25313             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
25315         PR jit/64089
25316         PR jit/84288
25317         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
25318         * configure: Regenerate.
25319         * configure.ac ("linker --version-script option"): New.
25320         ("linker soname option"): New.
25322 2018-03-09  Richard Biener  <rguenther@suse.de>
25324         PR tree-optimization/84775
25325         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
25326         immediate uses of predicate stmts and mark them modified.
25328         Revert
25329         PR tree-optimization/84178
25330         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
25331         to caller.
25332         (version_loop_for_if_conversion): Delay update_ssa call.
25333         (tree_if_conversion): Delay update_ssa until after predicate
25334         insertion.
25336 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
25338         PR target/84763
25339         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
25340         when the function accesses prior frames.
25342 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
25344         PR debug/84456
25345         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
25346         gen_llsym, otherwise call maybe_gen_llsym.
25348         PR inline-asm/84742
25349         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
25350         has ',' character inside of it.
25352 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25354         PR target/84748
25355         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
25356         as clobbering CC_REGNUM.
25358 2018-03-08  Richard Biener  <rguenther@suse.de>
25360         PR middle-end/84552
25361         * tree-scalar-evolution.c: Include tree-into-ssa.h.
25362         (follow_copies_to_constant): Do not follow SSA names registered
25363         for update.
25365 2018-03-08  Richard Biener  <rguenther@suse.de>
25367         PR tree-optimization/84178
25368         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
25369         to caller.
25370         (version_loop_for_if_conversion): Delay update_ssa call.
25371         (tree_if_conversion): Delay update_ssa until after predicate
25372         insertion.
25374 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
25376         PR tree-optimization/84178
25377         * tree-if-conv.c (release_bb_predicate): Remove the
25378         the assertion that the stmts have NULL use_ops.
25379         Discard the statements, asserting that they haven't
25380         yet been added to a BB.
25382 2018-03-08  Richard Biener  <rguenther@suse.de>
25384         PR tree-optimization/84746
25385         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
25386         (phi_translate): Pass in destination ANTIC_OUT set.
25387         (phi_translate_1): Likewise.  For a simplified result lookup
25388         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
25389         (phi_translate_set): Adjust.
25390         (do_pre_regular_insertion): Likewise.
25391         (do_pre_partial_partial_insertion): Likewise.
25393 2018-03-08  Martin Liska  <mliska@suse.cz>
25395         PR gcov-profile/84735
25396         * doc/gcov.texi: Document usage of profile files.
25397         * gcov-io.h: Document changes in the format.
25399 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
25401         PR debug/84404
25402         PR debug/84408
25403         * dwarf2out.c (struct dw_line_info_table): Update comments for
25404         view == -1.
25405         (FORCE_RESET_NEXT_VIEW): New.
25406         (FORCE_RESETTING_VIEW_P): New.
25407         (RESETTING_VIEW_P): Check for -1 too.
25408         (ZERO_VIEW_P): Likewise.
25409         (new_line_info_table): Force-reset next view.
25410         (dwarf2out_begin_function): Likewise.
25411         (dwarf2out_source_line): Simplify zero_view_p initialization.
25412         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
25413         view directly.  Omit view when omitting .loc at line 0.
25415 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
25417         PR tree-optimization/84740
25418         * tree-switch-conversion.c (process_switch): Call build_constructors
25419         only if info.phi_count is non-zero.
25421         PR tree-optimization/84739
25422         * tree-tailcall.c (find_tail_calls): Check call arguments against
25423         DECL_ARGUMENTS (current_function_decl) rather than
25424         DECL_ARGUMENTS (func) when checking for tail recursion.
25426 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
25428         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
25429         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
25430         Volker Reichelt's entry and add entries for people that perform
25431         GCC fuzzy testing and report numerous bugs.
25433 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
25435         PR target/82411
25436         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
25437         readonly data in sdata, if that is disabled.
25438         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
25439         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
25440         -mreadonly-in-sdata option.
25442 2018-03-07  Martin Sebor  <msebor@redhat.com>
25444         PR tree-optimization/84468
25445         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
25446         basic block when looking for nul assignment.
25448 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
25450         PR target/84277
25451         * except.h (output_function_exception_table): Adjust prototype.
25452         * except.c (output_function_exception_table): Remove FNNAME parameter
25453         and add SECTION parameter.  Ouput one part of the table at a time.
25454         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
25455         the first part of the exception table and emit unwind directives.
25456         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
25457         (i386_pe_seh_cold_init): Likewise.
25458         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
25459         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
25460         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
25461         (ix86_output_call_insn): Emit a nop in one more case for SEH.
25462         * config/i386/winnt.c: Include except.h.
25463         (struct seh_frame_state): Add reg_offset, after_prologue and
25464         in_cold_section fields.
25465         (i386_pe_seh_end_prologue): Set seh->after_prologue.
25466         (i386_pe_seh_cold_init): New function.
25467         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
25468         to seh->in_cold_section.
25469         (seh_emit_push): Record the offset of the push.
25470         (seh_emit_save): Record the offet of the save.
25471         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
25472         Test seh->after_prologue to disregard the epilogue.
25473         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
25474         (i386_pe_end_cold_function): New function.
25476 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
25478         PR fortran/84565
25479         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
25480         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
25482         PR c++/84704
25483         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
25484         on tmp_var.
25485         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
25486         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
25488         PR middle-end/84723
25489         * multiple_target.c: Include tree-inline.h and intl.h.
25490         (expand_target_clones): Diagnose and fail if node->definition and
25491         !tree_versionable_function_p (node->decl).
25493 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
25495         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
25496         sprint_ul.
25497         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
25498         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
25499         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
25501 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
25503         PR target/84710
25504         * combine.c (try_combine): Use reg_or_subregno instead of handling
25505         just paradoxical SUBREGs and REGs.
25507 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
25509         * config/arc/arc.c (arc_finalize_pic): Remove function.
25510         (arc_must_save_register): We use single base PIC register, remove
25511         checks to save/restore the PIC register.
25512         (arc_expand_prologue): Likewise.
25513         * config/arc/arc-protos.h (arc_set_default_type_attributes):
25514         Remove.
25515         (arc_verify_short): Likewise.
25516         (arc_attr_type): Likewise.
25517         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
25518         (walk_stores): Likewise.
25519         (arc_address_cost): Make it static.
25520         (arc_verify_short): Likewise.
25521         (branch_dest): Likewise.
25522         (arc_attr_type): Likewise.
25523         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
25524         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
25525         (arc_final_prescan_insn): Remove inserting the nops due to
25526         hardware hazards.  It is done in reorg step.
25527         (insn_length_variant_t): Remove.
25528         (insn_length_parameters_t): Likewise.
25529         (arc_insn_length_parameters): Likewise.
25530         (arc_get_insn_variants): Likewise.
25531         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
25533 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
25535         PR inline-asm/84683
25536         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
25537         assertion failure.
25539         PR tree-optimization/84687
25540         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
25541         on new_node->decl.
25542         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
25544 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25546         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
25547         Rename to ppc_speculation_barrier.
25548         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
25549         __builtin_ppc_speculation_barrier.
25551 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
25553         PR target/84700
25554         * combine.c (combine_simplify_rtx): Don't try to simplify if
25555         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
25556         are equal to x.
25558 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
25560         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
25561         to 32 bytes when compiling for POWER9.
25563 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
25565         PR target/84564
25566         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
25567         regparm >= 3 with no arg reg available also for calls with
25568         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
25570         PR target/84524
25571         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
25572         orig,vex.
25573         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
25575 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
25577         PR target/84264
25578         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
25580 2018-03-05  Richard Biener  <rguenther@suse.de>
25582         PR tree-optimization/84486
25583         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
25584         When inserting a __builtin_assume_aligned call set the LHS
25585         SSA name alignment info accordingly.
25587 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
25589         PR tree-optimization/84114
25590         * config/aarch64/aarch64.c (aarch64_reassociation_width)
25591         Avoid reassociation of FLOAT_MODE addition.
25593 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
25595         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
25596         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
25597         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
25598         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
25599         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
25600         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
25601         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
25602         and -mwbnoinvd.
25603         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
25604         __builtin_ia32_wbinvd): New builtins.
25605         (SPECIAL_ARGS2): New.
25606         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
25607         (SPECIAL_ARGS2): New.
25608         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
25609         (ix86_valid_target_attribute_inner_p): Ditto.
25610         (ix86_init_mmx_sse_builtins): Add special_args2.
25611         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
25612         TARGET_WBNOINVD_P): New.
25613         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
25614         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
25615         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
25616         * config/i386/immintrin.h (_wbinvd): New intrinsic.
25617         * config/i386/pconfigintrin.h: New file.
25618         * config/i386/wbnoinvdintrin.h: Ditto.
25619         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
25620         wbnoinvdintrin.h.
25621         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
25623 2018-03-05  Richard Biener  <rguenther@suse.de>
25625         PR tree-optimization/84670
25626         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
25627         member.
25628         (BB_VISITED_WITH_VISITED_SUCCS): New define.
25629         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
25630         (compute_antic_aux): Only assert the number of values in ANTIC_IN
25631         doesn't grow if all successors (recursively) were visited at least
25632         once.
25634 2018-03-05  Richard Biener  <rguenther@suse.de>
25636         PR tree-optimization/84650
25637         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
25638         if executed in the loop pipeline.
25640 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
25642         * doc/configfiles.texi (Configuration Files): Move info about
25643         conditionalizing $target-protos.h to...
25644         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
25645         differs from $target-protos.h.
25647 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
25648             Chung-Ju Wu  <jasonwucj@gmail.com>
25650         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
25651         * config/nds32/nds32-multiple.md (setmemsi): Define.
25652         * config/nds32/nds32-memory-manipulation.c
25653         (nds32_gen_dup_4_byte_to_word_value): New.
25654         (emit_setmem_word_loop): New.
25655         (emit_setmem_byte_loop): New.
25656         (nds32_expand_setmem_loop): New.
25657         (nds32_expand_setmem_loop_v3m): New.
25658         (nds32_expand_setmem_unroll): New.
25659         (nds32_expand_setmem): New.
25661 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
25662             Chung-Ju Wu  <jasonwucj@gmail.com>
25664         * config/nds32/nds32-memory-manipulation.c
25665         (nds32_emit_load_store): New.
25666         (nds32_emit_post_inc_load_store): New.
25667         (nds32_emit_mem_move): New.
25668         (nds32_emit_mem_move_block): New.
25669         (nds32_expand_movmemsi_loop_unknown_size): New.
25670         (nds32_expand_movmemsi_loop_known_size): New.
25671         (nds32_expand_movmemsi_loop): New.
25672         (nds32_expand_movmemsi_unroll): New.
25673         (nds32_expand_movmemqi): Rename ...
25674         (nds32_expand_movmemsi): ... to this.
25675         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
25676         (movmemsi): ... to this.
25677         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
25678         (nds32_expand_movmemsi): ... to this.
25680 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
25681             Monk Chiang  <sh.chiang04@gmail.com>
25682             Chung-Ju Wu  <jasonwucj@gmail.com>
25684         * config/nds32/nds32-protos.h
25685         (nds32_expand_load_multiple): New arguments.
25686         (nds32_expand_store_multiple): Ditto.
25687         (nds32_valid_multiple_load_store): Rename ...
25688         (nds32_valid_multiple_load_store_p): ... to this.
25689         * config/nds32/nds32-memory-manipulation.c
25690         (nds32_expand_load_multiple): Refine implementation.
25691         (nds32_expand_store_multiple): Ditto.
25692         * config/nds32/nds32-multiple.md
25693         (load_multiple): Update nds32_expand_load_multiple interface.
25694         (store_multiple): Update nds32_expand_store_multiple interface.
25695         * config/nds32/nds32-predicates.c
25696         (nds32_valid_multiple_load_store): Rename ...
25697         (nds32_valid_multiple_load_store_p): ... to this and refine
25698         implementation.
25699         * config/nds32/predicates.md
25700         (nds32_load_multiple_and_update_address_operation): New predicate.
25701         (nds32_store_multiple_and_update_address_operation): New predicate.
25703 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
25704             Chung-Ju Wu  <jasonwucj@gmail.com>
25706         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
25707         (combo): New attribute.
25708         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
25710 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
25712         * config/nds32/nds32.opt: Change -mcmodel= default value.
25714 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
25715             Monk Chiang  <sh.chiang04@gmail.com>
25716             Chung-Ju Wu  <jasonwucj@gmail.com>
25718         * config/nds32/constants.md (unspec_element): New enum.
25719         * config/nds32/constraints.md (Umw): New constraint.
25720         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
25721         * config/nds32/nds32-intrinsic.md: Likewise.
25722         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
25723         (nds32_valid_smw_lwm_base_p): New.
25724         (nds32_output_smw_single_word): New.
25725         (nds32_output_lmw_single_word): New.
25726         (nds32_expand_unaligned_load): New.
25727         (nds32_expand_unaligned_store): New.
25728         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
25729         (nds32_output_smw_single_word): Declare.
25730         (nds32_output_lmw_single_word): Declare.
25731         (nds32_expand_unaligned_load): Declare.
25732         (nds32_expand_unaligned_store): Declare.
25733         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
25734         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
25735         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
25736         NDS32_BUILTIN_UASTORE_DW.
25737         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
25738         predicate.
25740 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
25741             Kito Cheng  <kito.cheng@gmail.com>
25742             Chung-Ju Wu  <jasonwucj@gmail.com>
25744         * config/nds32/nds32-intrinsic.c
25745         (nds32_expand_builtin_null_ftype_reg): Delete.
25746         (nds32_expand_builtin_reg_ftype_imm): Ditto.
25747         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
25748         (nds32_read_argument): New.
25749         (nds32_legitimize_target): Ditto.
25750         (nds32_legitimize_argument): Ditto.
25751         (nds32_check_constant_argument): Ditto.
25752         (nds32_expand_unop_builtin): Ditto.
25753         (nds32_expand_unopimm_builtin): Ditto.
25754         (nds32_expand_binop_builtin): Ditto.
25755         (nds32_builtin_decl_impl): Ditto.
25756         (builtin_description): Ditto.
25757         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
25758         (nds32_init_builtins_impl): Ditto.
25759         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
25760         (nds32_builtin_decl): New.
25761         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
25762         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
25764 2018-03-02  Jeff Law  <law@redhat.com>
25766         * reorg.c (stop_search_p): Handle DEBUG_INSN.
25767         (redundant_insn, fill_simple_delay_slots): Likewise.
25768         (fill_slots_from_thread): Likewise.
25769         * resource.c (mark_referenced_resources): Likewise.
25770         (mark_set_resources, find_dead_or_set_registers): Likewise.
25772 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
25774         * substring-locations.h (format_warning_va): Formatting fix for
25775         ATTRIBUTE_GCC_DIAG.
25776         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
25777         argument.
25778         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
25779         * substring-locations.c: Include intl.h.
25780         (format_warning_va): Turned into small wrapper around
25781         format_warning_n_va, renamed to ...
25782         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
25783         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
25784         use ngettext.
25785         (format_warning_at_substring_n): New function.
25786         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
25787         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
25788         format_warning_at_substring with just a shorter name instead of
25789         const function pointer.
25790         (fmtwarn_n): New function.
25791         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
25792         appropriate, get rid of all the fmtstr temporaries, move conditionals
25793         with G_() wrapped string literals directly into fmtwarn arguments,
25794         cast dir.len to (int), formatting fixes.
25796 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
25798         * doc/invoke.texi: Remove "Cilk Plus" references.
25800 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
25801             Richard Biener  <rguenther@suse.de>
25803         PR ipa/84628
25804         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
25805         for error or warning attributes if CALL_FROM_THUNK_P is set.
25806         Formatting fixes.
25808 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
25810         PR target/56540
25811         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
25812         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
25814         PR target/56540
25815         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
25816         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
25818         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
25819         instead of -1U in last predictors element's probability member.
25821 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
25823         PR ipa/83983
25824         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
25825         arguments if they are comparable.
25827 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
25829         PR tree-optimization/84634
25830         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
25831         masks and masked_loop_p with a single loop_masks, making sure it's
25832         null for bb vectorization.
25834 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
25836         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
25837         (vect_analyze_data_ref_access): Use loop->safe_len rather than
25838         loop->force_vectorize to check whether there is no alias.
25840 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
25842         PR target/84614
25843         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
25844         prototypes.
25845         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
25846         comments.
25847         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
25848         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
25849         instead of a loop around prev_real_insn.
25850         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
25851         prev_real_insn.
25853         PR inline-asm/84625
25854         * config/i386/i386.c (ix86_print_operand): Use conditional
25855         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
25856         zero vector.
25858 2018-03-02  Richard Biener  <rguenther@suse.de>
25860         PR tree-optimization/84427
25861         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
25862         (bitmap_set_subtract_values): Rewrite to handle multiple
25863         exprs per value.
25864         (clean): Likewise.
25865         (prune_clobbered_mems): Likewise.
25866         (phi_translate): Take edge instead of pred/phiblock.
25867         (phi_translate_1): Likewise.
25868         (phi_translate_set): Likewise.  Insert all translated
25869         exprs for a value into the set, keeping possibly multiple
25870         expressions per value.
25871         (compute_antic_aux): Adjust for phi_translate changes.
25872         When intersecting union the expressions and prune those
25873         not in the final value set, keeping possibly multiple
25874         expressions per value.  Do not use value-insertion
25875         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
25876         all expressions.  Add verification that the value-sets
25877         only shrink during iteration.
25878         (compute_partial_antic_aux): Adjust for the phi_translate changes.
25879         (do_pre_regular_insertion): Likewise.
25880         (do_pre_partial_partial_insertion): Likewise.
25882 2018-03-02  Richard Biener  <rguenther@suse.de>
25884         PR target/82005
25885         * config/darwin.c (saved_debug_info_level): New static global.
25886         (darwin_asm_lto_start): Disable debug info generation for LTO out.
25887         (darwin_asm_lto_end): Restore debug info generation settings.
25889 2018-03-01  Martin Liska  <mliska@suse.cz>
25891         PR sanitizer/82484
25892         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
25893         volatile arguments.
25895 2018-03-01  Richard Biener  <rguenther@suse.de>
25897         PR debug/84645
25898         * dwarf2out.c (gen_variable_die): Properly handle late VLA
25899         type annotation with LTO when debug was disabled at compile-time.
25901 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
25903         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
25904         XINT with INTVAL.
25905         (mips_final_postscan_insn): Likewise.
25907 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
25909         PR rtl-optimization/84528
25910         * alias.c (init_alias_target): Add commentary.
25911         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
25912         a unique base value if the frame pointer is not eliminated
25913         to the stack pointer.
25915 2018-03-01  Tom de Vries  <tom@codesourcery.com>
25917         PR rtl-optimization/83327
25918         * lra-int.h (hard_regs_spilled_into): Declare.
25919         * lra.c (hard_regs_spilled_into): Define.
25920         (init_reg_info): Init hard_regs_spilled_into.
25921         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
25922         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
25923         (process_bb_lives): Handle hard_regs_spilled_into.
25924         (lra_create_live_ranges_1): Before doing liveness propagation, clear
25925         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
25927 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
25929         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
25930         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
25931         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
25932         * config/rs6000/aix72.h: New file.
25934 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
25936         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
25937         instead of warning_at with conditional singular and plural messages
25938         where possible.
25940         PR target/52991
25941         * stor-layout.c (update_alignment_for_field): For
25942         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
25943         && !DECL_PACKED (field), do the alignment update, just use
25944         only desired_align instead of MAX (type_align, desired_align)
25945         as the alignment.
25946         (place_field): Don't do known_align < desired_align handling
25947         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
25948         is non-NULL, instead do it after rli->prev_field handling and
25949         only if not within a bitfield word.  For DECL_PACKED (field)
25950         use type_align of BITS_PER_UNIT.
25952 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
25954         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
25955         superfluous parentheses and trailing spaces.
25957 2018-02-28  Richard Biener  <rguenther@suse.de>
25959         PR tree-optimization/84584
25960         * graphite-scop-detection.c (scop_detection::add_scop): Discard
25961         SCoPs with fake exit edge.
25963 2018-02-28  Martin Liska  <mliska@suse.cz>
25965         PR testsuite/84597
25966         * timevar.c (timer::print): Fix format to properly print 100%
25967         values.
25969 2018-02-28  Richard Biener  <rguenther@suse.de>
25971         PR middle-end/84607
25972         * genmatch.c (capture_info::walk_match): Do not mark
25973         captured expressions without operands as expr_p given
25974         they act more like predicates and should be subject to
25975         "lost tail" side-effect preserving.
25977 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
25979         PR rtl-optimization/81611
25980         * auto-inc-dec.c (attempt_change): Move dead note from
25981         mem_insn if it's the next use of regno
25982         (find_address): Take address use of reg holding
25983         non-incremented value.  Add parm to limit search to the named
25984         reg only.
25985         (merge_in_block): Attempt to use a mem insn that is the next
25986         use of the original regno.
25988 2018-02-27  Martin Sebor  <msebor@redhat.com>
25990         PR c++/83871
25991         * doc/invoke.texi (-Wmissing-attributes): New option.
25992         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
25994 2018-02-27  Martin Sebor  <msebor@redhat.com>
25996         PR translation/84207
25997         * diagnostic-core.h (warning_n, error_n, inform_n): Change
25998         n argument to unsigned HOST_WIDE_INT.
25999         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
26000         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
26001         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
26002         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
26004 2018-02-27  Richard Biener  <rguenther@suse.de>
26006         PR tree-optimization/84512
26007         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
26008         Do not use the estimate returned from record_stmt_cost for
26009         the scalar iteration cost but sum properly using add_stmt_cost.
26011 2018-02-27  Richard Biener  <rguenther@suse.de>
26013         PR tree-optimization/84466
26014         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
26015         Adjust last change to less strictly validate use operands.
26017 2018-02-27  Martin Liska  <mliska@suse.cz>
26019         PR gcov-profile/84548
26020         * gcov.c (process_file): Allow partial overlap and consider it
26021         also as group functions.
26022         (output_lines): Properly calculate range of lines for a group.
26024 2018-02-27  Martin Liska  <mliska@suse.cz>
26026         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
26027         'ggc' suffixes.  Change first column width.
26028         (timer::print): Fix formatting of the column.
26030 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
26032         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
26033         preserve inline entry blocks for the sake of debug inline
26034         entry point markers alone.
26035         (remove_unused_locals): Suggest in comments a better place to
26036         force the preservation of inline entry blocks that are
26037         otherwise unused, but do not preserve them.
26039 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
26041         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
26043 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
26045         PR target/84039
26046         * config/i386/constraints.md (Bs): Replace
26047         ix86_indirect_branch_register with
26048         TARGET_INDIRECT_BRANCH_REGISTER.
26049         (Bw): Likewise.
26050         * config/i386/i386.md (indirect_jump): Likewise.
26051         (tablejump): Likewise.
26052         (*sibcall_memory): Likewise.
26053         (*sibcall_value_memory): Likewise.
26054         Peepholes of indirect call and jump via memory: Likewise.
26055         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
26056         (*sibcall_value_GOT_32): Likewise.
26057         * config/i386/predicates.md (indirect_branch_operand): Likewise.
26058         (GOT_memory_operand): Likewise.
26059         (call_insn_operand): Likewise.
26060         (sibcall_insn_operand): Likewise.
26061         (GOT32_symbol_operand): Likewise.
26062         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
26064 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
26066         PR rtl-optimization/83496
26067         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
26068         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
26069         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
26070         redundant insn, if any.
26071         (relax_delay_slots): Likewise.
26072         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
26074 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
26076         PR tree-optimization/83965
26077         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
26078         that grouped statements are part of a reduction chain.  Return
26079         true if the statement is not marked as a reduction itself but
26080         is part of a group.
26081         (vect_recog_dot_prod_pattern): Don't check whether the statement
26082         is part of a group here.
26083         (vect_recog_sad_pattern): Likewise.
26084         (vect_recog_widen_sum_pattern): Likewise.
26086 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
26088         PR debug/84545
26089         * final.c (rest_of_clean_state): Also look for calls inside sequences.
26091 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
26093         PR target/84530
26094         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
26095         the bool argument.
26096         (ix86_output_indirect_function_return): New prototype.
26097         (ix86_split_simple_return_pop_internal): Likewise.
26098         * config/i386/i386.c (indirect_return_via_cx): New.
26099         (indirect_return_via_cx_bnd): Likewise.
26100         (indirect_thunk_name): Handle return va CX_REG.
26101         (output_indirect_thunk_function): Create alias for
26102         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
26103         (ix86_output_indirect_jmp): Remove the bool argument.
26104         (ix86_output_indirect_function_return): New function.
26105         (ix86_split_simple_return_pop_internal): Likewise.
26106         * config/i386/i386.md (*indirect_jump): Don't pass false
26107         to ix86_output_indirect_jmp.
26108         (*tablejump_1): Likewise.
26109         (simple_return_pop_internal): Change it to define_insn_and_split.
26110         Call ix86_split_simple_return_pop_internal to split it for
26111         -mfunction-return=.
26112         (simple_return_indirect_internal): Call
26113         ix86_output_indirect_function_return instead of
26114         ix86_output_indirect_jmp.
26116 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
26118         PR bootstrap/84405
26119         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
26120         memset and value initialization afterwards.
26122 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
26124         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
26126 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26128         PR target/84521
26129         * common/config/aarch64/aarch64-common.c
26130         (aarch_option_optimization_table[]): Switch
26131         off fomit-frame-pointer
26133 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
26134             Chung-Ju Wu  <jasonwucj@gmail.com>
26136         * config/nds32/nds32-multiple.md (load_multiple): Disallow
26137         volatile memory.
26138         (store_multiple): Ditto.
26140 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
26142         * config.gcc: Add --with-cpu support for nds32 target.
26143         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
26144         * config/nds32/nds32.opt: Add -mcpu= option.
26146 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
26148         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
26149         isel=yes): Warn for these deprecated options.
26151 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
26153         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
26154         ISA_2_5_MASKS_EMBEDDED.
26156 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
26158         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
26159         p->max as pointers rather than using iterative_hash_expr.
26161 2018-02-23  Carl Love  <cel@us.ibm.com>
26163         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
26164         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
26165         BU_P8V_OVERLOAD_2.
26166         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
26167         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
26168         P8V_BUILTIN_VEC_VUNSIGNED2.
26170 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
26172         PR target/81572
26173         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
26174         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
26175         LRA_UNKNOWN_ALT.
26176         * lra-constraints.c (curr_insn_transform): Set up
26177         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
26178         LRA_UNKNOWN_ALT.
26179         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
26180         * lra-eliminations.c (spill_pseudos): Ditto.
26181         (process_insn_for_elimination): Ditto.
26182         * lra-lives.c (reg_early_clobber_p): Use the new macros.
26183         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
26184         LRA_NON_CLOBBERED_ALT.
26186 2018-02-22  Martin Sebor  <msebor@redhat.com>
26188         PR tree-optimization/84480
26189         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
26190         to maybe_diag_stxncpy_trunc.  Call it.
26191         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
26192         from gimple_fold_builtin_strcpy.  Print inlining stack.
26193         (handle_builtin_stxncpy): Print inlining stack.
26194         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
26196 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
26198         PR target/84176
26199         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
26200         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
26201         and -fcheck-pointer-bounds are used together.
26202         (indirect_thunk_prefix): New enum.
26203         (indirect_thunk_need_prefix): New function.
26204         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
26205         "_nt" instead of "_bnd" for NOTRACK prefix.
26206         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
26207         (output_indirect_thunk_function): Likewise.
26208         (): Likewise.
26209         (ix86_code_end): Update output_indirect_thunk_function calls.
26210         (ix86_output_indirect_branch_via_reg): Replace
26211         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
26212         (ix86_output_indirect_branch_via_push): Likewise.
26213         (ix86_output_function_return): Likewise.
26214         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
26215         incompatible with -fcf-protection=branch and
26216         -fcheck-pointer-bounds.
26218 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
26220         PR target/83335
26221         * config/aarch64/aarch64.c (aarch64_print_address_internal):
26222         Change gcc_assert call to output_operand_lossage.
26224 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
26226         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
26228 2018-02-22  DJ Delorie  <dj@redhat.com>
26229             Sebastian Perta  <sebastian.perta@renesas.com>
26230             Oleg Endo  <olegendo@gcc.gnu.org>
26232         * config/rx/rx.c (rx_rtx_costs): New function.
26233         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
26235 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
26237         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
26239 2018-02-22  Martin Liska  <mliska@suse.cz>
26241         PR driver/83193
26242         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
26243         Add "native" as a possible value.
26245 2018-02-22  Martin Liska  <mliska@suse.cz>
26247         PR driver/83193
26248         * config/i386/i386.c (ix86_option_override_internal):
26249         Add "native" as a possible value for -march and -mtune.
26251 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
26253         PR target/84502
26254         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
26255         to all type variants.
26257         PR tree-optimization/84503
26258         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
26259         width as info->bitpos + info->bitsize - start.
26260         (merged_store_group::merge_overlapping): Simplify width computation.
26261         (check_no_overlap): New function.
26262         (imm_store_chain_info::try_coalesce_bswap): Compute expected
26263         start + width and last_order of the group, fail if check_no_overlap
26264         fails.
26265         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
26266         to group if check_no_overlap fails.
26268 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
26270         * config/rs6000/altivec.md: Delete contraint arguments to
26271         define_expand, define_split, and define_peephole2, and in
26272         define_insn_and_split if always unused.
26273         * config/rs6000/darwin.md: Ditto.
26274         * config/rs6000/dfp.md: Ditto.
26275         * config/rs6000/rs6000.md: Ditto.
26276         * config/rs6000/sync.md: Ditto.
26277         * config/rs6000/vector.md: Ditto.
26278         * config/rs6000/vsx.md: Ditto.
26280 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
26282         * config/rs6000/altivec.md: Write output control strings as braced
26283         blocks instead of double-quoted strings.
26284         * config/rs6000/darwin.md: Ditto.
26285         * config/rs6000/rs6000.md: Ditto.
26286         * config/rs6000/vector.md: Ditto.
26287         * config/rs6000/vsx.md: Ditto.
26289 2018-02-21  Jason Merrill  <jason@redhat.com>
26291         PR c++/84314 - ICE with templates and fastcall attribute.
26292         * attribs.c (build_type_attribute_qual_variant): Remove assert.
26294 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
26296         * ipa-cp.c (determine_versionability): Fix comment typos.
26298 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
26300         PR c/84229
26301         * ipa-cp.c (determine_versionability): Do not version functions caling
26302         va_arg_pack.
26304 2018-02-21  Martin Liska  <mliska@suse.cz>
26306         PR driver/83193
26307         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
26308         Add "native" as a possible value.
26309         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
26310         the macro when native cpu detection is available.
26312 2018-02-21  Martin Liska  <mliska@suse.cz>
26314         PR driver/83193
26315         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
26316         Add "native" as a possible value.
26317         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
26318         when native cpu detection is available.
26320 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
26321             Martin Sebor  <msebor@redhat.com>
26323         PR tree-optimization/84478
26324         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
26325         false.
26326         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
26327         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
26328         support which is conservatively correct, for 2 only stay conservative
26329         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
26330         argument to the 2 argument get_range_strlen, adjust 6 arg
26331         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
26332         false.
26333         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
26334         (gimple_fold_builtin_strlen): Pass true as last argument to
26335         get_range_strlen.
26337 2018-02-20  Martin Sebor  <msebor@redhat.com>
26339         PR middle-end/84095
26340         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
26341         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
26342         (builtin_memref::builtin_memref): Factor out parts into
26343         set_base_and_offset and call it.
26345 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
26347         PR middle-end/84406
26348         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
26349         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
26350         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
26351         search at the associated MODE_INT.
26353 2018-02-20  Jeff Law  <law@redhat.com>
26355         PR middle-end/82123
26356         PR tree-optimization/81592
26357         PR middle-end/79257
26358         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
26359         for range data rather than using global data.
26360         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
26361         range data rather than using global data.
26362         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
26363         pass it to children as needed.
26364         (struct directive::fmtresult): Similarly.
26365         (struct directive::set_width): Similarly.
26366         (struct directive::set_precision): Similarly.
26367         (format_integer, format_directive, parse_directive): Similarly.
26368         (format_none): Accept unnamed vr_values parameter.
26369         (format_percent, format_floating, format_character): Similarly.
26370         (format_string, format_plain): Similarly.
26371         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
26372         the EVRP range analyzer for range data rather than using global data.
26373         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
26374         gimple-ssa-evrp-analyze.h
26375         (class sprintf_dom_walker): Add after_dom_children member function.
26376         Add evrp_range_analyzer member.
26377         (sprintf_dom_walker::before_dom_children): Call into the EVRP
26378         range analyzer as needed.
26379         (sprintf_dom_walker::after_dom_children): New member function.
26380         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
26381         if not optimizing.
26382         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
26383         (evrp_range_analyzer::pop_to_marker): Likewise.
26385 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
26387         PR tree-optimization/84419
26388         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
26389         with the required type if its current type is compatible but
26390         different.
26392 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
26394         PR middle-end/82004
26395         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
26396         after vectorization.
26398 2018-02-20  Martin Liska  <mliska@suse.cz>
26400         PR driver/83193
26401         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
26402         possible values if we don't have a hint.
26404 2018-02-20  Martin Liska  <mliska@suse.cz>
26406         PR c/84310
26407         PR target/79747
26408         * final.c (shorten_branches): Build align_tab array with one
26409         more element.
26410         * opts.c (finish_options): Add alignment option limit check.
26411         (MAX_CODE_ALIGN): Likewise.
26412         (MAX_CODE_ALIGN_VALUE): Likewise.
26413         * doc/invoke.texi: Document maximum allowed option value for
26414         all -falign-* options.
26416 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
26418         PR target/84146
26419         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
26420         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
26421         * var-tracking.c (emit_note_insn_var_location): Remove all references
26422         to NOTE_INSN_CALL_ARG_LOCATION.
26423         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
26424         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
26425         Use copy_rtx_if_shared.
26426         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
26427         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
26428         (dwarf2out_var_location): Remove handling of
26429         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
26430         on call_insn.
26431         * final.c (final_scan_insn): Remove all references to
26432         NOTE_INSN_CALL_ARG_LOCATION.
26433         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
26434         before dumping final insns.
26435         * except.c (emit_note_eh_region_end): Remove all references to
26436         NOTE_INSN_CALL_ARG_LOCATION.
26437         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
26438         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
26439         * config/arc/arc.c (hwloop_optimize): Likewise.
26440         * config/arm/arm.c (create_fix_barrier): Likewise.
26441         * config/s390/s390.c (s390_chunkify_start): Likewise.
26442         * config/sh/sh.c (find_barrier): Likewise.
26443         * config/i386/i386.c (rest_of_insert_endbranch,
26444         ix86_seh_fixup_eh_fallthru): Likewise.
26445         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
26446         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
26447         * config/frv/frv.c (frv_function_prologue): Likewise.
26448         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
26449         reg note.
26450         (note_outside_basic_block_p): Remove all references to
26451         NOTE_INSN_CALL_ARG_LOCATION.
26452         * gengtype.c (adjust_field_rtx_def): Likewise.
26453         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
26454         Likewise.
26455         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
26456         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
26458         PR c++/84444
26459         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
26460         is ADDR_EXPR.
26462         PR tree-optimization/84452
26463         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
26464         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
26465         is NULL.
26467 2018-02-19  Martin Liska  <mliska@suse.cz>
26469         PR sanitizer/82183
26470         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
26472 2018-02-19  Martin Liska  <mliska@suse.cz>
26473             Richard Sandiford  <richard.sandiford@linaro.org>
26475         PR tree-optimization/82491
26476         * gimple-fold.c (get_base_constructor): Make earlier bail out
26477         to prevent ubsan.
26479 2018-02-19  Carl Love  <cel@us.ibm.com>
26481         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
26482         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
26483         BU_P8V_OVERLOAD_1.
26484         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
26485         P8V_BUILTIN_VEC_NEG.
26487 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
26489         * config/rl78/rl78.md (movdf): New define expand.
26491 2018-02-19  Martin Liska  <mliska@suse.cz>
26493         PR other/80589
26494         * doc/invoke.texi: Fix typo.
26495         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
26497 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
26499         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
26500         handle rs6000_single_float and rs6000_double_float specially for
26501         e500 family CPUs.
26503 2018-02-16  Jeff Law  <law@redhat.com>
26505         * config/rx/rx.c (add_pop_cfi_notes): New function.;
26506         (pop_regs): Use it.
26508 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
26510         PR ipa/84425
26511         * ipa-inline.c (inline_small_functions): Fix a typo.
26513 2018-02-16  Nathan Sidwell  <nathan@acm.org>
26515         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
26517 2018-02-16  Carl Love  <cel@us.ibm.com>
26519         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
26520         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
26521         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
26522         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
26523         expansion to P8V_BUILTIN_VEC_FLOAT2.
26525 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
26527         PR rtl-optimization/70023
26528         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
26529         src_regno into account.
26531 2018-02-16  Carl Love  <cel@us.ibm.com>
26533         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
26534         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
26535         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
26536         * config/rs6000/rs6000.c: Remove case statements for
26537         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
26538         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
26539         and P9V_BUILTIN_VEC_VINSERT4B.
26540         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
26541         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
26542         * config/rs6000/vsx.md:
26543         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
26544         vec_insert4b.
26546 2018-02-16  Carl Love  <cel@us.ibm.com>
26548         * config/rs6000/altivec.h: Add builtin names vec_extract4b
26549         vec_insert4b.
26550         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
26551         definitions.
26552         * config/rs6000/rs6000-c.c: Add the definitions for
26553         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
26554         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
26555         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
26556         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
26557         definition for insert4b and define insn *insert3b_internal.
26558         * doc/extend.texi: Add documentation for vec_extract4b.
26560 2018-02-16  Nathan Sidwell  <nathan@acm.org>
26562         * doc/extend.texi (Backwards Compatibility): Mention friend
26563         injection.  Note for-scope is deprecated.
26564         * doc/invoke.texi (-ffriend-injection): Deprecate.
26566 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
26568         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
26569         that moved to I2, also allow destinations that are a paradoxical
26570         subreg (instead of a normal reg).
26572 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
26574         PR target/83831
26575         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
26576         to QImode.
26578 2018-02-16  Richard Biener  <rguenther@suse.de>
26580         PR tree-optimization/84037
26581         PR tree-optimization/84016
26582         PR target/82862
26583         * config/i386/i386.c (ix86_builtin_vectorization_cost):
26584         Adjust vec_construct for the fact we need additional higher latency
26585         128bit inserts for AVX256 and AVX512 vector builds.
26586         (ix86_add_stmt_cost): Scale vector construction cost for
26587         elementwise loads.
26589 2018-02-16  Richard Biener  <rguenther@suse.de>
26591         PR tree-optimization/84417
26592         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
26593         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
26594         (non_rewritable_lvalue_p): Likewise, use poly-ints.
26596 2018-02-16  Martin Liska  <mliska@suse.cz>
26598         PR sanitizer/84307
26599         * internal-fn.def (ASAN_CHECK): Set proper flags.
26600         (ASAN_MARK): Likewise.
26602 2018-02-16  Julia Koval  <julia.koval@intel.com>
26604         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
26605         from PTA_CANNONLAKE.
26607 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
26609         PR target/84272
26610         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
26611         Use ++iter rather than iter++ for std::list iterators.
26612         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
26613         defer deleting them until all nodes in the forest are processed.  Do
26614         free even leaf nodes.  Change to_process into auto_vec.
26616         PR bootstrap/84405
26617         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
26618         * vec.h (vec_default_construct): Use memset instead of placement new
26619         if BROKEN_VALUE_INITIALIZATION is defined.
26620         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
26621         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
26622         is defined.
26624         PR rtl-optimization/83723
26625         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
26626         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
26627         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
26628         recursive calls.
26629         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
26630         callers.
26631         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
26633 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
26635         PR rtl-optimization/81443
26636         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
26637         from inner REGs to paradoxical SUBREGs.
26639 2018-02-16  Richard Biener  <rguenther@suse.de>
26641         PR tree-optimization/84399
26642         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
26643         For operands we can analyze at their definition make sure we can
26644         analyze them at each use as well.
26646 2018-02-16  Richard Biener  <rguenther@suse.de>
26648         PR tree-optimization/84190
26649         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
26650         volatile accesses if the decl isn't volatile.
26652 2018-02-15  Jason Merrill  <jason@redhat.com>
26654         PR c++/84314 - ICE with templates and fastcall attribute.
26655         * attribs.c (build_type_attribute_qual_variant): Don't clobber
26656         TYPE_CANONICAL on an existing type.
26658 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
26660         PR tree-optimization/84383
26661         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
26662         dstoff nor call operand_equal_p if dstbase is NULL.
26664         PR tree-optimization/84334
26665         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
26666         also a CONSTANT_CLASS_P, punt.
26668 2018-02-14  Jim Wilson  <jimw@sifive.com>
26670         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
26671         first SMALL_OPERAND check.  New local min_second_step.  Move assert
26672         to where locals are set.  Add TARGET_RVC support.
26673         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
26675 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
26677         * doc/invoke.texi: Correct -Wformat-overflow code sample.
26679 2018-02-14  Martin Sebor  <msebor@redhat.com>
26681         PR tree-optimization/83698
26682         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
26683         arrays constrain the offset range to their bounds.
26684         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
26685         (builtin_access::overlap): Avoid setting the size of overlap if it's
26686         already been set.
26687         (maybe_diag_overlap): Also consider arrays when deciding what values
26688         of offsets to include in diagnostics.
26690 2018-02-14  Martin Sebor  <msebor@redhat.com>
26692         PR c/84108
26693         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
26694         that correspond to the kind of a declaration.
26696 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
26698         PR target/83984
26699         * config/pa/pa.md: Load address of PIC label using the linkage table
26700         if the label is nonlocal.
26702 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
26704         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
26705         warning message if user requests -maltivec=be.
26706         * doc/invoke.texi: Document deprecation of -maltivec=be.
26708 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
26710         PR target/84220
26711         * config/rs6000/rs6000-c.c: Update definitions for
26712         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
26713         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
26715 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26717         PR target/84239
26718         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
26719         add _get_ssp intrinsics. Remove argument from
26720         __builtin_ia32_rdssp[d|q].
26721         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
26722         * config/i386/i386-builtin.def: Remove argument from
26723         __builtin_ia32_rdssp[d|q].
26724         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
26725         ix86_expand_special_args_builtin for _rdssp[d|q].
26726         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
26727         Clear register before usage.
26728         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
26729         Add documentation for new _get_ssp and _inc_ssp intrinsics.
26731 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
26733         PR tree-optimization/84357
26734         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
26735         operand 1 of an ARRAY_REF too.
26737 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
26739         PR target/83831
26740         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
26741         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
26742         declarations.
26743         (set_of_reg): New struct.
26744         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
26745         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
26746         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
26747         functions.
26748         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
26749         Split into bitclr, bitset, bitinvert patterns if appropriate.
26750         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
26751         use rx_fuse_in_memory_bitop.
26752         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
26753         to named insn, correct maximum insn length.
26755 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
26757         PR target/79242
26758         * machmode.def: Define a complex mode for PARTIAL_INT.
26759         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
26760         MODE_PARTIAL_INT.
26761         * doc/rtl.texi: Document CSPImode.
26762         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
26763         handling.
26764         (msp430_hard_regno_nregs_with_padding): Likewise.
26766 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
26768         PR target/84279
26769         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
26771 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
26773         PR rtl-optimization/84169
26774         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
26775         we generated a parallel as new i3 and we split that to new i2 and i3
26776         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
26777         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
26778         those to i2, not i1.  Partially rewrite this scan code.
26780 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
26782         PR c/82210
26783         * stor-layout.c (place_field): For variable length fields, adjust
26784         offset_align afterwards not just based on the field's alignment,
26785         but also on the size.
26787         PR middle-end/84309
26788         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
26789         of exps and logs in the use_exp2 case.
26791 2018-02-13  Jeff Law  <law@redhat.com>
26793         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
26794         entry for "vector".
26796         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
26797         ARGS as unused.
26799 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
26801         PR debug/84342
26802         PR debug/84319
26803         * common.opt (gas-loc-support, gas-locview-support): New.
26804         (ginline-points, ginternal-reset-location-views): New.
26805         * doc/invoke.texi: Document them.  Use @itemx where intended.
26806         (gvariable-location-views): Adjust.
26807         * target.def (reset_location_view): New.
26808         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
26809         (TARGET_RESET_LOCATION_VIEW): New.
26810         * doc/tm.texi: Rebuilt.
26811         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
26812         (dwarf2out_default_as_locview_support): New.
26813         (output_asm_line_debug_info): Use option variables.
26814         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
26815         (output_loc_list): Likewise.
26816         (add_high_low_attributes): Check option variables.
26817         Don't output entry view attribute in strict mode.
26818         (gen_inlined_subroutine_die): Check option variables.
26819         (dwarf2out_inline_entry): Likewise.
26820         (init_sections_and_labels): Likewise.
26821         (dwarf2out_early_finish): Likewise.
26822         (maybe_reset_location_view): New, from...
26823         (dwarf2out_var_location): ... here.  Call it.
26824         * debug.h (dwarf2out_default_as_loc_support): Declare.
26825         (dwarf2out_default_as_locview_support): Declare.
26826         * hooks.c (hook_int_rtx_insn_0): New.
26827         * hooks.h (hook_int_rtx_insn_0): Declare.
26828         * toplev.c (process_options): Take -gas-loc-support and
26829         -gas-locview-support from dwarf2out.  Enable
26830         -gvariable-location-views by default only with locview
26831         assembler support.  Enable -ginternal-reset-location-views by
26832         default only if the target defines the corresponding hook.
26833         Enable -ginline-points by default if location views are
26834         enabled; force it disabled if statement frontiers are
26835         disabled.
26836         * tree-inline.c (expand_call_inline): Check option variables.
26837         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
26839 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
26841         PR tree-optimization/84321
26842         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
26843         handling.  Also check whether the anti-range contains any values
26844         that satisfy the mask; switch to a VR_RANGE if not.
26846 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
26848         PR sanitizer/84340
26849         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
26851 2018-02-13  Martin Jambor  <mjambor@suse.cz>
26853         PR c++/83990
26854         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
26855         of call statements, also set location of a load to a temporary.
26857 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
26859         * config/rl78/rl78.c (add_vector_labels): New function.
26860         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
26861         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
26862         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
26863         which checks that no arguments are passed.
26864         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
26865         * doc/extend.texi: Documentation for the new attribute.
26867 2018-02-13  Andreas Schwab  <schwab@suse.de>
26869         * config/riscv/linux.h (CPP_SPEC): Define.
26871 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
26873         PR target/84335
26874         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
26875         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
26876         OPTION_MASK_ISA_AES as first argument to def_builtin_const
26877         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
26878         instead of OPTION_MASK_ISA_PCLMUL as first argument to
26879         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
26880         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
26881         temporarily for AES and PCLMUL builtins.
26883         PR tree-optimization/84339
26884         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
26885         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
26886         Formatting fixes.
26888         PR middle-end/84309
26889         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
26890         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
26891         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
26892         inline function.
26893         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
26894         inline function.
26895         * omp-simd-clone.h: New file.
26896         * omp-simd-clone.c: Include omp-simd-clone.h.
26897         (expand_simd_clones): No longer static.
26898         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
26899         cgraph.h and omp-simd-clone.h.
26900         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
26901         (vect_recog_widen_shift_pattern): Formatting fix.
26902         (vect_pattern_recog_1): Don't check optab for calls.
26904         PR target/84336
26905         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
26906         operands[2] into a REG before using gen_lowpart on it.
26908 2018-02-12  Jeff Law  <law@redhat.com>
26910         PR target/83760
26911         * config/sh/sh.c (find_barrier): Consider a sibling call
26912         a barrier as well.
26914         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
26915         successfully back substituting a reg.
26917 2018-02-12  Richard Biener  <rguenther@suse.de>
26919         PR tree-optimization/84037
26920         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
26921         parameter, move visited init to caller.
26922         (vect_slp_analyze_operations): Separate cost from validity
26923         check, initialize visited once for all instances.
26924         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
26925         for all instances.
26926         * tree-vect-stmts.c (vect_model_simple_cost): Make early
26927         out an assert.
26928         (vect_model_promotion_demotion_cost): Likewise.
26929         (vectorizable_bswap): Guard cost modeling with !slp_node
26930         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
26931         SLP stmts.
26932         (vectorizable_call): Likewise.
26933         (vectorizable_conversion): Likewise.
26934         (vectorizable_assignment): Likewise.
26935         (vectorizable_shift): Likewise.
26936         (vectorizable_operation): Likewise.
26937         (vectorizable_store): Likewise.
26938         (vectorizable_load): Likewise.
26939         (vectorizable_condition): Likewise.
26940         (vectorizable_comparison): Likewise.
26942 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
26944         PR sanitizer/84307
26945         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
26946         (ASAN_MARK): Fix fnspec to account for return value, change pointer
26947         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
26949 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
26951         PR middle-end/83665
26952         * params.def (inline-min-speedup): Increase from 8 to 15.
26953         (max-inline-insns-auto): Decrease from 40 to 30.
26954         * ipa-split.c (consider_split): Add some buffer for function to
26955         be considered inlining candidate.
26956         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
26957         default values.
26959 2018-02-12  Richard Biener  <rguenther@suse.de>
26961         PR tree-optimization/84037
26962         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
26963         matched stmts if we cannot swap the non-matched ones.
26965 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
26967         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
26968         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
26969         _mm_maskz_scalef_round_ss): New intrinsics.
26970         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
26971         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
26972         __builtin_ia32_scalefss_round): Remove.
26973         (__builtin_ia32_scalefsd_mask_round,
26974         __builtin_ia32_scalefss_mask_round): New intrinsics.
26975         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
26976         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
26977         ((match_operand:VF_128 2 "<round_nimm_predicate>"
26978         "<round_constraint>")): Changed to ...
26979         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
26980         "<round_scalar_constraint>")): ... this.
26981         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
26982         %0, %1, %2<round_op3>}"): Changed to ...
26983         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
26984         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
26985         %2<round_scalar_mask_op3>}"): ... this.
26986         * config/i386/subst.md (round_scalar_nimm_predicate): New.
26988 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
26990         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
26991         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
26992         (_mm_maskz_sqrt_round_ss): New intrinsics.
26993         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
26994         (__builtin_ia32_sqrtsd_mask_round)
26995         (__builtin_ia32_sqrtss_mask_round): New builtins.
26996         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
26997         (__builtin_ia32_sqrtss_round): Remove.
26998         (__builtin_ia32_sqrtsd_mask_round)
26999         (__builtin_ia32_sqrtss_mask_round): New builtins.
27000         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
27001         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
27002         ((match_operand:VF_128 1 "vector_operand"
27003         "xBm,<round_constraint>")): Changed to ...
27004         ((match_operand:VF_128 1 "vector_operand"
27005         "xBm,<round_scalar_constraint>")): ... this.
27006         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
27007         %0, %2, %<iptr>1<round_op3>}): Changed to ...
27008         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
27009         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
27010         %<iptr>1<round_scalar_mask_op3>}): ... this.
27011         ((set_attr "prefix" "<round_prefix>")): Changed to ...
27012         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
27014 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
27016         PR target/84266
27017         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
27018         Cast vec_cmpeq result to correct type.
27019         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
27020         Cast vec_cmpgt result to correct type.
27022 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
27024         * final.c (final_scan_insn_1): Renamed from...
27025         (final_scan_insn): ... this.  New wrapper, to recover
27026         seen from the outermost call in recursive ones.
27027         * config/sparc/sparc.c (output_return): Drop seen from call.
27028         (output_sibcall): Likewise.
27029         * config/visium/visium.c (output_branch): Likewise.
27031 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
27033         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
27034         function label.
27036 2018-02-10  Alan Modra  <amodra@gmail.com>
27038         PR target/84300
27039         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
27040         Specify LR as an input.
27042 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
27044         PR sanitizer/83987
27045         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
27046         remove_member_access_dummy_vars): New functions.
27047         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
27048         lower_omp_1, execute_lower_omp): Use them.
27050         PR rtl-optimization/84308
27051         * shrink-wrap.c (spread_components): Release todo vector.
27053 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
27055         PR rtl-optimization/57193
27056         * ira-color.c (struct allocno_color_data): Add member
27057         conflict_allocno_hard_prefs.
27058         (update_conflict_allocno_hard_prefs): New.
27059         (bucket_allocno_compare_func): Add a preference based on
27060         conflict_allocno_hard_prefs.
27061         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
27062         (color_allocnos): Remove a dead code.  Initiate
27063         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
27065 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
27067         PR target/84226
27068         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
27069         constraint from =wa to wa.  Avoid a subreg on the output operand,
27070         instead use a pseudo and subreg it in a move.
27071         (p9_xxbrd_<mode>): Changed to ...
27072         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
27073         (p9_xxbrd_v2df): New expander.
27074         (p9_xxbrw_<mode>): Changed to ...
27075         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
27076         (p9_xxbrw_v4sf): New expander.
27078 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
27080         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
27082 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
27084         PR target/83926
27085         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
27086         multiply in 32-bit mode.
27087         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
27088         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
27089         mode.
27091 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
27093         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
27094         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
27095         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
27096         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
27098 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
27100         PR lto/84213
27101         * dwarf2out.c (is_trivial_indirect_ref): New function.
27102         (dwarf2out_late_global_decl): Do not generate a location
27103         attribute for variables that have a non-trivial DECL_VALUE_EXPR
27104         and that are not defined in the current unit.
27106 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
27108         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
27109         instead of a libcall for UNORDERED.
27111 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
27113         PR target/82641
27114         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
27115         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
27117 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27119         PR target/PR84295
27120         * config/s390/s390.c (s390_set_current_function): Invoke
27121         s390_indirect_branch_settings also if fndecl didn't change.
27123 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
27125         * config/rs6000/rs6000.md (blockage): Set length to zero.
27127 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
27129         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
27131 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
27133         PR sanitizer/84285
27134         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
27135         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
27136         -static-lib*san.
27138         PR debug/84252
27139         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
27140         PARALLEL incoming that failed vt_get_decl_and_offset check.
27142         PR middle-end/84237
27143         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
27144         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
27145         TREE_READONLY bit.
27146         (get_variable_section): For decls in named .bss* sections pass true as
27147         second argument to bss_initializer_p.
27149 2018-02-09  Marek Polacek  <polacek@redhat.com>
27150             Jakub Jelinek  <jakub@redhat.com>
27152         PR c++/83659
27153         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
27154         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
27155         Sync some changes from cxx_fold_indirect_ref.
27157 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
27159         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
27160         markers.
27161         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
27162         (BLOCK_INLINE_ENTRY_LABEL): New.
27163         (dwarf2out_var_location): Disregard inline entry markers.
27164         (inline_entry_data): New struct.
27165         (inline_entry_data_hasher): New hashtable type.
27166         (inline_entry_data_hasher::hash): New.
27167         (inline_entry_data_hasher::equal): New.
27168         (inline_entry_data_table): New variable.
27169         (add_high_low_attributes): Add DW_AT_entry_pc and
27170         DW_AT_GNU_entry_view attributes if a pending entry is found
27171         in inline_entry_data_table.  Add old entry_pc attribute only
27172         if debug nonbinding markers are disabled.
27173         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
27174         markers are enabled.
27175         (block_within_block_p, dwarf2out_inline_entry): New.
27176         (dwarf2out_finish): Check that no entries remained in
27177         inline_entry_data_table.
27178         * final.c (reemit_insn_block_notes): Handle inline entry notes.
27179         (final_scan_insn, notice_source_line): Likewise.
27180         (rest_of_clean_state): Skip inline entry markers.
27181         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
27182         markers.
27183         * gimple.c (gimple_build_debug_inline_entry): New.
27184         * gimple.h (enum gimple_debug_subcode): Add
27185         GIMPLE_DEBUG_INLINE_ENTRY.
27186         (gimple_build_debug_inline_entry): Declare.
27187         (gimple_debug_inline_entry_p): New.
27188         (gimple_debug_nonbind_marker_p): Adjust.
27189         * insn-notes.def (INLINE_ENTRY): New.
27190         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
27191         inline entry marker notes.
27192         (print_insn): Likewise.
27193         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
27194         (INSN_DEBUG_MARKER_KIND): Likewise.
27195         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
27196         * tree-inline.c (expand_call_inline): Build and insert
27197         debug_inline_entry stmt.
27198         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
27199         inline entry blocks early, if nonbind markers are enabled.
27200         (dump_scope_block): Dump fragment info.
27201         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
27202         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
27203         (gimple_build_debug_inline_entry): New.
27204         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
27205         Enable/disable inline entry points too.
27206         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
27207         (DEBUG_INSN): Describe inline entry markers.
27209         * common.opt (gvariable-location-views): New.
27210         (gvariable-location-views=incompat5): New.
27211         * config.in: Rebuilt.
27212         * configure: Rebuilt.
27213         * configure.ac: Test assembler for view support.
27214         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
27215         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
27216         * dwarf2out.c (var_loc_view): New typedef.
27217         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
27218         (dwarf2out_locviews_in_attribute): New.
27219         (dwarf2out_locviews_in_loclist): New.
27220         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
27221         (enum dw_line_info_opcode): Add LI_adv_address.
27222         (struct dw_line_info_table): Add view.
27223         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
27224         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
27225         (zero_view_p): New variable.
27226         (ZERO_VIEW_P): New macro.
27227         (output_asm_line_debug_info): New.
27228         (struct var_loc_node): Add view.
27229         (add_AT_view_list, AT_loc_list): New.
27230         (add_var_loc_to_decl): Add view param.  Test it against last.
27231         (new_loc_list): Add view params.  Record them.
27232         (AT_loc_list_ptr): Handle loc and view lists.
27233         (view_list_to_loc_list_val_node): New.
27234         (print_dw_val): Handle dw_val_class_view_list.
27235         (size_of_die): Likewise.
27236         (value_format): Likewise.
27237         (loc_list_has_views): New.
27238         (gen_llsym): Set vl_symbol too.
27239         (maybe_gen_llsym, skip_loc_list_entry): New.
27240         (dwarf2out_maybe_output_loclist_view_pair): New.
27241         (output_loc_list): Output view list or entries too.
27242         (output_view_list_offset): New.
27243         (output_die): Handle dw_val_class_view_list.
27244         (output_dwarf_version): New.
27245         (output_compilation_unit_header): Use it.
27246         (output_skeleton_debug_sections): Likewise.
27247         (output_rnglists, output_line_info): Likewise.
27248         (output_pubnames, output_aranges): Update version comments.
27249         (output_one_line_info_table): Output view numbers in asm comments.
27250         (dw_loc_list): Determine current endview, pass it to new_loc_list.
27251         Call maybe_gen_llsym.
27252         (loc_list_from_tree_1): Adjust.
27253         (add_AT_location_description): Create view list attribute if
27254         needed, check it's absent otherwise.
27255         (convert_cfa_to_fb_loc_list): Adjust.
27256         (maybe_emit_file): Call output_asm_line_debug_info for test.
27257         (dwarf2out_var_location): Reset views as needed.  Precompute
27258         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
27259         attribute.  Set view.
27260         (new_line_info_table): Reset next view.
27261         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
27262         (dwarf2out_source_line): Likewise.  Output view resets and labels to
27263         the assembler, or select appropriate line info opcodes.
27264         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
27265         (optimize_string_length): Catch it.  Adjust.
27266         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
27267         dw_val_class_view_list, and remove it if no longer needed.
27268         (hash_loc_list): Hash view numbers.
27269         (loc_list_hasher::equal): Compare them.
27270         (optimize_location_lists): Check whether a view list symbol is
27271         needed, and whether the locview attribute is present, and
27272         whether they match.  Remove the locview attribute if no longer
27273         needed.
27274         (index_location_lists): Call skip_loc_list_entry for test.
27275         (dwarf2out_finish): Call output_asm_line_debug_info for test.
27276         Use output_dwarf_version.
27277         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
27278         (struct dw_val_node): Add val_view_list.
27279         * final.c (SEEN_NEXT_VIEW): New.
27280         (set_next_view_needed): New.
27281         (clear_next_view_needed): New.
27282         (maybe_output_next_view): New.
27283         (final_start_function): Rename to...
27284         (final_start_function_1): ... this.  Take pointer to FIRST,
27285         add SEEN parameter.  Emit param bindings in the initial view.
27286         (final_start_function): Reintroduce SEEN-less interface.
27287         (final): Rename to...
27288         (final_1): ... this.  Take SEEN parameter.  Output final pending
27289         next view at the end.
27290         (final): Reintroduce seen-less interface.
27291         (final_scan_insn): Output pending next view before switching
27292         sections or ending a block.  Mark the next view as needed when
27293         outputting variable locations.  Notify debug backend of section
27294         changes, and of location view changes.
27295         (rest_of_handle_final): Adjust.
27296         * toplev.c (process_options): Autodetect value for debug variable
27297         location views option.  Warn on incompat5 without -gdwarf-5.
27298         * doc/invoke.texi (gvariable-location-views): New.
27299         (gvariable-location-views=incompat5): New.
27300         (gno-variable-location-views): New.
27302 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
27304         PR tree-optimization/84136
27305         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
27306         that the result of find_edge is non-NULL.
27308 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
27310         PR target/83008
27311         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
27312         storing integer register in SImode.  Fix cost of 256 and 512
27313         byte aligned SSE register store.
27315 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
27317         * config/i386/i386.c (ix86_multiplication_cost): Fix
27318         multiplication cost for TARGET_AVX512DQ.
27320 2018-02-08  Marek Polacek  <polacek@redhat.com>
27322         PR tree-optimization/84238
27323         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
27324         get_range_strlen.
27326 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
27328         PR tree-optimization/84265
27329         * tree-vect-stmts.c (vectorizable_store): Don't treat
27330         VMAT_CONTIGUOUS accesses as grouped.
27331         (vectorizable_load): Likewise.
27333 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
27335         PR tree-optimization/81635
27336         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
27337         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
27338         (test_round_for_mask): New functions.
27339         (wide_int_cc_tests): Call test_round_for_mask.
27340         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
27341         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
27342         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
27343         range returned by get_range_info.
27345 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
27347         PR ipa/81360
27348         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
27349         * symtab.c: Include builtins.h
27350         (symtab_node::output_to_lto_symbol_table_p): Move here
27351         from lto-streamer-out.c:output_symbol_p.
27352         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
27353         (output_symbol_p): Move all logic to symtab.c
27354         (produce_symtab): Update.
27356 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27358         * config/s390/s390-opts.h (enum indirect_branch): Define.
27359         * config/s390/s390-protos.h (s390_return_addr_from_memory)
27360         (s390_indirect_branch_via_thunk)
27361         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
27362         (enum s390_indirect_branch_type): Define.
27363         * config/s390/s390.c (struct s390_frame_layout, struct
27364         machine_function): Remove.
27365         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
27366         (indirect_branch_table_label_no, indirect_branch_table_name):
27367         Define variables.
27368         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
27369         (enum s390_indirect_branch_option): Define.
27370         (s390_return_addr_from_memory): New function.
27371         (s390_handle_string_attribute): New function.
27372         (s390_attribute_table): Add new attribute handler.
27373         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
27374         (s390_indirect_branch_via_thunk): New function.
27375         (s390_indirect_branch_via_inline_thunk): New function.
27376         (s390_function_ok_for_sibcall): When jumping via thunk disallow
27377         sibling call optimization for non z10 compiles.
27378         (s390_emit_call): Force indirect branch target to be a single
27379         register.  Add r1 clobber for non-z10 compiles.
27380         (s390_emit_epilogue): Emit return jump via return_use expander.
27381         (s390_reorg): Handle JUMP_INSNs as execute targets.
27382         (s390_option_override_internal): Perform validity checks for the
27383         new command line options.
27384         (s390_indirect_branch_attrvalue): New function.
27385         (s390_indirect_branch_settings): New function.
27386         (s390_set_current_function): Invoke s390_indirect_branch_settings.
27387         (s390_output_indirect_thunk_function):  New function.
27388         (s390_code_end): Implement target hook.
27389         (s390_case_values_threshold): Implement target hook.
27390         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
27391         macros.
27392         * config/s390/s390.h (struct s390_frame_layout)
27393         (struct machine_function): Move here from s390.c.
27394         (TARGET_INDIRECT_BRANCH_NOBP_RET)
27395         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
27396         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
27397         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
27398         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
27399         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
27400         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
27401         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
27402         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
27403         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
27404         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
27405         (mnemonic attribute): Add values which aren't recognized
27406         automatically.
27407         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
27408         pattern for branch conversion.  Fix mnemonic attribute.
27409         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
27410         indirect branch via thunk if requested.
27411         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
27412         ("*indirect_jump"): Disable for branch conversion using out of
27413         line thunks.
27414         ("indirect_jump_via_thunk<mode>_z10")
27415         ("indirect_jump_via_thunk<mode>")
27416         ("indirect_jump_via_inlinethunk<mode>_z10")
27417         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
27418         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
27419         ("casesi_jump_via_inlinethunk<mode>_z10")
27420         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
27421         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
27422         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
27423         ("*indirect2_jump"): Disable for branch conversion.
27424         ("casesi_jump"): Turn into expander and expand patterns for branch
27425         conversion.
27426         ("return_use"): New expander.
27427         ("*return"): Emit return via thunk and rename it to ...
27428         ("*return<mode>"): ... this one.
27429         * config/s390/s390.opt: Add new options and and enum for the
27430         option values.
27432 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
27434         * lra-constraints.c (match_reload): Unconditionally use
27435         gen_lowpart_SUBREG, rather than selecting between that
27436         and equivalent gen_rtx_SUBREG code.
27438 2018-02-08  Richard Biener  <rguenther@suse.de>
27440         PR tree-optimization/84233
27441         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
27442         changed flag instead of boguously re-using phi_inserted.
27444 2018-02-08  Martin Jambor  <mjambor@suse.cz>
27446         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
27447         static local variables.
27449 2018-02-08  Richard Biener  <rguenther@suse.de>
27451         PR tree-optimization/84278
27452         * tree-vect-stmts.c (vectorizable_store): When looking for
27453         smaller vector types to perform grouped strided loads/stores
27454         make sure the mode is supported by the target.
27455         (vectorizable_load): Likewise.
27457 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
27459         * config/aarch64/aarch64.c (aarch64_components_for_bb):
27460         Increase LDP/STP opportunities by adding adjacent callee-saves.
27462 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
27464         PR rtl-optimization/84068
27465         PR rtl-optimization/83459
27466         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
27468 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
27470         PR tree-optimization/84224
27471         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
27472         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
27473         non-zero arguments.
27475 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
27477         PR target/84113
27478         * config/rs6000/altivec.md (*restore_world): Remove LR use.
27479         * config/rs6000/predicates.md (restore_world_operation): Adjust op
27480         count, remove one USE.
27482 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
27484         * doc/install.texi (Configuration): Document the
27485         --with-long-double-format={ibm,ieee} PowerPC configuration
27486         options.
27488         PR target/84154
27489         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
27490         Convert from define_expand to be define_insn_and_split.  Rework
27491         float/double/_Float128 conversions to QI/HI/SImode to work with
27492         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
27493         conversions to QI/HImode types did a store and then a load to
27494         truncate the value.  For conversions to VSX registers, don't split
27495         the insn, instead emit the code directly.  Use the code iterator
27496         any_fix to combine signed and unsigned conversions.
27497         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
27498         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
27499         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
27500         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
27501         (fix_<mode>di2_hw): Likewise.
27502         (fixuns_<mode>di2_hw): Likewise.
27503         (fix_<mode>si2_hw): Likewise.
27504         (fixuns_<mode>si2_hw): Likewise.
27505         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
27506         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
27507         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
27508         fix<uns>_trunc<SFDF:mode>si2_p8.
27509         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
27510         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
27511         (fix<uns>_<mode>_mem): Likewise.
27512         (fctiw<u>z_<mode>_mem): Likewise.
27513         (fix<uns>_<mode>_mem): Likewise.
27514         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
27515         the register allocator from doing a direct move to the GPRs to do
27516         a store, and instead use the ISA 3.0 store byte/half-word from
27517         vector register instruction.  For IEEE 128-bit floating point,
27518         also optimize stores of 32-bit ints.
27519         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
27521 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
27523         * genextract.c (push_pathstr_operand): New function to support
27524         [a-zA-Z].
27525         (walk_rtx): Call push_pathstr_operand.
27526         (print_path): Support [a-zA-Z].
27528 2018-02-07  Richard Biener  <rguenther@suse.de>
27530         PR tree-optimization/84037
27531         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
27532         (cse_and_gimplify_to_preheader): Declare.
27533         (vect_get_place_in_interleaving_chain): Likewise.
27534         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
27535         ivexpr_map.
27536         (_loop_vec_info::~_loop_vec_info): Delete it.
27537         (cse_and_gimplify_to_preheader): New function.
27538         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
27539         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
27540         (vectorizable_load): Likewise.  For grouped stores always base
27541         the IV on the first element.
27542         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
27543         condition before gimplifying.
27545 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
27547         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
27548         *DIV_EXPR and *MOD_EXPR.
27550 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
27552         PR target/84248
27553         * config/i386/i386.c (ix86_option_override_internal): Mask out
27554         the CF_SET bit when checking -fcf-protection.
27556 2018-02-07  Tom de Vries  <tom@codesourcery.com>
27558         PR libgomp/84217
27559         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
27560         enough.
27562 2018-02-07  Richard Biener  <rguenther@suse.de>
27564         PR tree-optimization/84204
27565         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
27566         this place.
27568         PR tree-optimization/84205
27569         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
27570         special-case isl_ast_op_zdiv_r.
27572         PR tree-optimization/84223
27573         * graphite-scop-detection.c (gather_bbs::before_dom_children):
27574         Only add conditions from within the region.
27575         (gather_bbs::after_dom_children): Adjust.
27577 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
27579         PR target/84209
27580         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
27581         * config/avr/avr.md: Only post-reload split REG-REG moves if
27582         either register is GENERAL_REG_P.
27584 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
27586         PR tree-optimization/84235
27587         * tree-ssa-scopedtables.c
27588         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
27589         if the subtraction is performed in floating point type where NaNs are
27590         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
27591         build 1.  Formatting fix.
27593 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
27595         PR target/84146
27596         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
27597         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
27598         and skip it regardless of bb boundaries.  Use CALL_P macro,
27599         don't test INSN_P (insn) together with CALL_P or JUMP_P check
27600         unnecessarily, formatting fix.
27602 2018-02-06  Michael Collison  <michael.collison@arm.com>
27604         * config/arm/thumb2.md:
27605         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
27606         (*thumb_mov_notscc): Ditto.
27608 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
27610         PR target/84154
27611         * config/rs6000/rs6000.md (su code attribute): Use "u" for
27612         unsigned_fix, not "s".
27614 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27616         * configure.ac (gcc_fn_eh_frame_ro): New function.
27617         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
27618         correct .eh_frame permissions.
27619         * configure: Regenerate.
27621 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
27623         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
27624         irrelevant options.
27626 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27628         * config/rs6000/rs6000.c (rs6000_option_override_internal):
27629         Display warning message for -mno-speculate-indirect-jumps.
27631 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
27633         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
27634         Undocumented.
27635         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
27637 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
27639         PR tree-optimization/84225
27640         * tree-eh.c (find_trapping_overflow): Only call
27641         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
27643 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
27645         PR target/84145
27646         * config/i386/i386.c: Reimplement the check of possible options
27647         -mibt/-mshstk conbination. Change error messages.
27648         * doc/invoke.texi: Fix a typo: remove extra '='.
27650 2018-02-06  Marek Polacek  <polacek@redhat.com>
27652         PR tree-optimization/84228
27653         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
27655 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
27657         PR target/82641
27658         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
27659         emitted arch directives.
27660         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
27661         __ARM_FEATURE_COPROC before changing architectures.
27663 2018-02-06  Richard Biener  <rguenther@suse.de>
27665         * config/i386/i386.c (print_reg): Fix typo.
27666         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
27668 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
27670         * configure: Regenerate.
27672 2018-02-05  Martin Sebor  <msebor@redhat.com>
27674         PR tree-optimization/83369
27675         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
27676         inlining context.
27678 2018-02-05  Martin Liska  <mliska@suse.cz>
27680         * doc/invoke.texi: Cherry-pick upstream r323995.
27682 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
27684         * ira.c (ira_init_register_move_cost): Adjust comment.
27686 2018-02-05  Martin Liska  <mliska@suse.cz>
27688         PR gcov-profile/84137
27689         * doc/gcov.texi: Fix typo in documentation.
27691 2018-02-05  Martin Liska  <mliska@suse.cz>
27693         PR gcov-profile/83879
27694         * doc/gcov.texi: Document necessity of --dynamic-list-data when
27695         using dlopen functionality.
27697 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
27699         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
27700         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
27701         _mm_maskz_range_ss, _mm_mask_range_round_ss,
27702         _mm_maskz_range_round_ss): New intrinsics.
27703         (__builtin_ia32_rangesd128_round)
27704         (__builtin_ia32_rangess128_round): Remove.
27705         (__builtin_ia32_rangesd128_mask_round,
27706         __builtin_ia32_rangess128_mask_round): New builtins.
27707         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
27708         __builtin_ia32_rangess128_round): Remove.
27709         (__builtin_ia32_rangesd128_mask_round,
27710         __builtin_ia32_rangess128_mask_round): New builtins.
27711         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
27712         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
27713         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
27714         "<round_saeonly_constraint>")): Changed to ...
27715         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
27716         "<round_saeonly_scalar_constraint>")): ... this.
27717         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
27718         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
27719         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
27720         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
27721         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
27723 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
27725         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
27726         options.
27727         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
27728         Remove all values except native, 8540 and 8548.
27730 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
27732         * config/i386/i386.c (ix86_output_function_return): Pass
27733         INVALID_REGNUM, instead of -1, as invalid register number to
27734         indirect_thunk_name and output_indirect_thunk.
27736 2018-02-02  Julia Koval  <julia.koval@intel.com>
27738         * config.gcc: Add -march=icelake.
27739         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
27740         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
27741         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
27742         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
27743         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
27744         (processor_target_table): Add icelake.
27745         (ix86_option_override_internal): Handle new PTAs.
27746         (get_builtin_code_for_version): Handle icelake.
27747         (M_INTEL_COREI7_ICELAKE): New.
27748         (fold_builtin_cpu): Handle icelake.
27749         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
27750         * doc/invoke.texi: Add -march=icelake.
27752 2018-02-02  Julia Koval  <julia.koval@intel.com>
27754         * config/i386/i386.c (ix86_option_override_internal): Change flags type
27755         to wide_int_bitmask.
27756         * wide-int-bitmask.h: New.
27758 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
27760         PR target/84066
27761         * config/i386/i386.md: Replace Pmode with word_mode in
27762         builtin_setjmp_setup and builtin_longjmp to support x32.
27764 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
27766         PR target/56010
27767         PR target/83743
27768         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
27769         #include "opts.h".
27770         (rs6000_supported_cpu_names): New static variable.
27771         (linux_cpu_translation_table): Likewise.
27772         (elf_platform) <cpu>: Define new static variable and use it.
27773         Translate kernel AT_PLATFORM name to canonical name if needed.
27774         Error if platform name is unknown.
27776 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
27778         PR target/84089
27779         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
27781 2018-02-01  Jeff Law  <law@redhat.com>
27783         PR target/84128
27784         * config/i386/i386.c (release_scratch_register_on_entry): Add new
27785         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
27786         the scratch if RELEASE_VIA_POP is false.
27787         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
27788         If we have to save a temporary register, decrement SIZE appropriately.
27789         Pass new arguments to release_scratch_register_on_entry.
27790         (ix86_adjust_stack_and_probe): Likewise.
27791         (ix86_emit_probe_stack_range): Pass new arguments to
27792         release_scratch_register_on_entry.
27794 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
27796         PR rtl-optimization/84157
27797         * combine.c (change_zero_ext): Use REG_P predicate in
27798         front of HARD_REGISTER_P predicate.
27800 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
27802         * config/avr/avr.c (avr_option_override): Move disabling of
27803         -fdelete-null-pointer-checks to...
27804         * common/config/avr/avr-common.c (avr_option_optimization_table):
27805         ...here.
27807 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
27809         PR tree-optimization/81635
27810         * tree-data-ref.c (split_constant_offset_1): For types that
27811         wrap on overflow, try to use range info to prove that wrapping
27812         cannot occur.
27814 2018-02-01  Renlin Li  <renlin.li@arm.com>
27816         PR target/83370
27817         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
27818         TAILCALL_ADDR_REGS.
27819         (aarch64_register_move_cost): Likewise.
27820         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
27821         TAILCALL_ADDR_REGS.
27822         (REG_CLASS_NAMES): Likewise.
27823         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
27824         TAILCALL_ADDR_REGS. Remove IP registers.
27825         * config/aarch64/aarch64.md (Ucs): Update register constraint.
27827 2018-02-01  Richard Biener  <rguenther@suse.de>
27829         * domwalk.h (dom_walker::dom_walker): Add additional constructor
27830         for specifying RPO order and allow NULL for that.
27831         * domwalk.c (dom_walker::dom_walker): Likewise.
27832         (dom_walker::walk): Handle NULL RPO order.
27833         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
27834         in RPO order.
27835         (rewrite_update_dom_walker): Likewise.
27836         (mark_def_dom_walker): Likewise.
27838 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
27840         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
27841         (aarch64_maybe_expand_sve_subreg_move): Declare.
27842         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
27843         * config/aarch64/predicates.md (aarch64_any_register_operand): New
27844         predicate.
27845         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
27846         that are semantically a reverse operation.
27847         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
27848         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
27849         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
27850         functions.
27851         (aarch64_can_change_mode_class): For big-endian, forbid changes
27852         between two SVE modes if they have different element sizes.
27854 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
27856         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
27857         the TImode handling for big-endian targets.
27859 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
27861         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
27862         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
27863         not just bytes.
27864         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
27865         Remove BSWAP handing for big-endian targets and use the form of
27866         LD1RQ appropariate for the mode.
27868 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
27870         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
27871         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
27872         duplicated element.
27874 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
27876         PR tearget/83845
27877         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
27878         check for operands that need to go through aarch64_sve_reload_be.
27880 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
27882         PR tree-optimization/81661
27883         PR tree-optimization/84117
27884         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
27885         * tree-eh.c: Include gimplify.h.
27886         (find_trapping_overflow, replace_trapping_overflow,
27887         rewrite_to_non_trapping_overflow): New functions.
27888         * tree-vect-loop.c: Include tree-eh.h.
27889         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
27890         * tree-data-ref.c: Include tree-eh.h.
27891         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
27893 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
27895         PR rtl-optimization/84123
27896         * combine.c (change_zero_ext): Check if hard register satisfies
27897         can_change_dest_mode before calling gen_lowpart_SUBREG.
27899 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
27901         PR target/82444
27902         * ira.c (ira_init_register_move_cost): Remove assert.
27904 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
27906         PR rtl-optimization/84071
27907         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
27908         * doc/tm.texi: Regenerate.
27910 2018-01-31  Richard Biener  <rguenther@suse.de>
27912         PR tree-optimization/84132
27913         * tree-data-ref.c (analyze_miv_subscript): Properly
27914         check whether evolution_function_is_affine_multivariate_p
27915         before calling gcd_of_steps_may_divide_p.
27917 2018-01-31  Julia Koval  <julia.koval@intel.com>
27919         PR target/83618
27920         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
27921         * config/i386/i386.md (rdpid_rex64) New.
27922         (rdpid): Make 32bit only.
27924 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
27926         PR lto/84105
27927         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
27928         an IDENTIFIER_NODE for FUNCTION_TYPE's.
27930 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
27932         Revert
27933         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
27935         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
27937 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
27939         PR rtl-optimization/84071
27940         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
27941         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
27943 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
27945         * config/arc/arc.c (arc_handle_aux_attribute): New function.
27946         (arc_attribute_table): Add 'aux' attribute.
27947         (arc_in_small_data_p): Consider aux like variables.
27948         (arc_is_aux_reg_p): New function.
27949         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
27950         (arc_get_aux_arg): New function.
27951         (prepare_move_operands): Handle aux-register access.
27952         (arc_handle_aux_attribute): New function.
27953         * doc/extend.texi (ARC Variable attributes): Add subsection.
27955 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
27957         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
27958         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
27959         (arc_attribute_table): Add 'uncached' attribute.
27960         (arc_print_operand): Print '.di' flag for uncached memory
27961         accesses.
27962         (arc_in_small_data_p): Do not consider for small data the uncached
27963         types.
27964         (arc_is_uncached_mem_p): New function.
27965         * config/arc/predicates.md (compact_store_memory_operand): Check
27966         for uncached memory accesses.
27967         (nonvol_nonimm_operand): Likewise.
27968         * doc/extend.texi (ARC Type Attribute): New subsection.
27970 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
27972         PR c/84100
27973         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
27974         falign-loops=): Add Optimization flag.
27976 2018-01-30  Jeff Law  <law@redhat.com>
27978         PR target/84064
27979         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
27980         INT_REGISTERS_SAVED.  Check it prior to calling
27981         get_scratch_register_on_entry.
27982         (ix86_adjust_stack_and_probe): Similarly.
27983         (ix86_emit_probe_stack_range): Similarly.
27984         (ix86_expand_prologue): Corresponding changes.
27986 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27988         PR target/40411
27989         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
27990         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
27992 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
27994         PR target/84112
27995         * lra-constraints.c (curr_insn_transform): Process AND in the
27996         address.
27998 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
28000         PR rtl-optimization/83986
28001         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
28002         dependence against last_pending_memory_flush in addition to
28003         pending_jump_insns.
28005 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
28007         PR tree-optimization/81611
28008         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
28009         copies.
28011 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
28013         PR target/83758
28014         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
28015         a reg rtx.
28017 2018-01-30  Richard Biener  <rguenther@suse.de>
28018             Jakub Jelinek  <jakub@redhat.com>
28020         PR tree-optimization/84111
28021         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
28022         inner loops added during recursion, as they don't have up-to-date
28023         SSA form.
28025 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
28027         PR ipa/81360
28028         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
28029         (can_inline_edge_by_limits_p): ... here.
28030         (can_early_inline_edge_p, check_callers,
28031         update_caller_keys, update_callee_keys, recursive_inlining,
28032         add_new_edges_to_heap, speculation_useful_p,
28033         inline_small_functions,
28034         inline_small_functions, flatten_function,
28035         inline_to_all_callers_1): Update.
28037 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
28039         * profile-count.c (profile_count::combine_with_ipa_count): Handle
28040         zeros correctly.
28042 2018-01-30  Richard Biener  <rguenther@suse.de>
28044         PR tree-optimization/83008
28045         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
28046         invariant and constant vector uses in stmts when they need
28047         more than one stmt.
28049 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28051         PR bootstrap/84017
28052         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
28053         * configure: Regenerate.
28055 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
28057         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
28058         pattern.
28059         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
28060         Use gen_rtx_REG rather than gen_lowpart.
28062 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
28064         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
28065         rather than 0 when creating partial subregs.
28067 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
28069         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
28070         of usage.
28072 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
28074         PR target/81550
28075         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
28076         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
28077         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
28078         flags.  This restores the settings used before the 2017-07-24.
28079         Turning off pre increment/decrement/modify allows IVOPTS to
28080         optimize DF/SF loops where the index is an int.
28082 2018-01-29  Richard Biener  <rguenther@suse.de>
28083             Kelvin Nilsen  <kelvin@gcc.gnu.org>
28085         PR bootstrap/80867
28086         * tree-vect-stmts.c (vectorizable_call): Don't call
28087         targetm.vectorize_builtin_md_vectorized_function if callee is
28088         NULL.
28090 2018-01-22  Carl Love  <cel@us.ibm.com>
28092         * doc/extend.tex: Fix typo in second arg in
28093         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
28095 2018-01-29  Richard Biener  <rguenther@suse.de>
28097         PR tree-optimization/84086
28098         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
28099         (flush_ssaname_freelist): When SSA names were released reset
28100         the SCEV hash table.
28102 2018-01-29  Richard Biener  <rguenther@suse.de>
28104         PR tree-optimization/84057
28105         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
28106         removed paths when removing edges.
28108 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
28110         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
28111         -mfunction-return=@var{choice}.
28113 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
28115         PR diagnostic/84034
28116         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
28117         Handle CR like TAB.
28118         (layout::print_source_line): Likewise.
28119         (test_get_line_width_without_trailing_whitespace): Add test cases.
28121 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
28123         PR middle-end/84040
28124         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
28125         debug insns.
28127 2018-01-26  Jim Wilson  <jimw@sifive.com>
28129         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
28131         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
28132         specified.
28134 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28136         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
28137         and CMP + SUB-immediate -> SUBS.
28139 2018-01-26  Martin Sebor  <msebor@redhat.com>
28141         PR tree-optimization/83896
28142         * tree-ssa-strlen.c (get_string_len): Rename...
28143         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
28144         Avoid assuming length is constant.
28145         (handle_char_store): Use HOST_WIDE_INT for string length.
28147 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
28149         PR target/81763
28150         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
28151         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
28153 2018-01-26  Richard Biener  <rguenther@suse.de>
28155         PR rtl-optimization/84003
28156         * dse.c (record_store): Only record redundant stores when
28157         the earlier store aliases at least all accesses the later one does.
28159 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
28161         PR rtl-optimization/83985
28162         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
28163         REG_CFA_RESTORE insns.
28164         (delete_unmarked_insns): Don't ignore separate shrink wrapping
28165         REG_CFA_RESTORE insns here.
28167         PR c/83989
28168         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
28169         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
28171 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
28173         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
28174         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
28175         (arc_init): Likewise.
28176         (arc_override_options): Likewise.
28177         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
28178         value.
28179         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
28180         support.
28181         * config/arc/arc.h (TARGET_DBNZ): Define.
28182         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
28183         properly set the tune attribute.
28184         (dbnz): Use TARGET_DBNZ guard.
28185         * config/arc/arc.opt (mtune): Add core3 option.
28187 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
28189         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
28190         recognize new pic like addresses.
28191         (arc_delegitimize_address): Clean up.
28193 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
28195         * config/arc/arc-arches.def: Option mrf16 valid for all
28196         architectures.
28197         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
28198         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
28199         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
28200         * config/arc/arc-tables.opt: Regenerate.
28201         * config/arc/arc.c (arc_conditional_register_usage): Handle
28202         reduced register file case.
28203         (arc_file_start): Set must have build attributes.
28204         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
28205         mrf16 option value.
28206         * config/arc/arc.opt (mrf16): Add new option.
28207         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
28208         * config/arc/genmultilib.awk: Handle new mrf16 option.
28209         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
28210         * config/arc/t-multilib: Regenerate.
28211         * doc/invoke.texi (ARC Options): Document mrf16 option.
28213 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
28215         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
28216         * config/arc/arc.c (arc_handle_secure_attribute): New function.
28217         (arc_attribute_table): Add 'secure_call' attribute.
28218         (arc_print_operand): Print secure call operand.
28219         (arc_function_ok_for_sibcall): Don't optimize tail calls when
28220         secure.
28221         (arc_is_secure_call_p): New function.  * config/arc/arc.md
28222         (call_i): Add support for sjli instruction.
28223         (call_value_i): Likewise.
28224         * config/arc/constraints.md (Csc): New constraint.
28226 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
28227             John Eric Martin  <John.Martin@emmicro-us.com>
28229         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
28230         * config/arc/arc.c (_arc_jli_section): New struct.
28231         (arc_jli_section): New type.
28232         (rc_jli_sections): New static variable.
28233         (arc_handle_jli_attribute): New function.
28234         (arc_attribute_table): Add jli_always and jli_fixed attribute.
28235         (arc_file_end): New function.
28236         (TARGET_ASM_FILE_END): Define.
28237         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
28238         (arc_add_jli_section): New function.
28239         (jli_call_scan): Likewise.
28240         (arc_reorg): Call jli_call_scan.
28241         (arc_output_addsi): Remove 'S' from printing asm operand.
28242         (arc_is_jli_call_p): New function.
28243         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
28244         operand.
28245         (movhi_insn): Likewise.
28246         (movsi_insn): Likewise.
28247         (movsi_set_cc_insn): Likewise.
28248         (loadqi_update): Likewise.
28249         (load_zeroextendqisi_update): Likewise.
28250         (load_signextendqisi_update): Likewise.
28251         (loadhi_update): Likewise.
28252         (load_zeroextendhisi_update): Likewise.
28253         (load_signextendhisi_update): Likewise.
28254         (loadsi_update): Likewise.
28255         (loadsf_update): Likewise.
28256         (movsicc_insn): Likewise.
28257         (bset_insn): Likewise.
28258         (bxor_insn): Likewise.
28259         (bclr_insn): Likewise.
28260         (bmsk_insn): Likewise.
28261         (bicsi3_insn): Likewise.
28262         (cmpsi_cc_c_insn): Likewise.
28263         (movsi_ne): Likewise.
28264         (movsi_cond_exec): Likewise.
28265         (clrsbsi2): Likewise.
28266         (norm_f): Likewise.
28267         (normw): Likewise.
28268         (swap): Likewise.
28269         (divaw): Likewise.
28270         (flag): Likewise.
28271         (sr): Likewise.
28272         (kflag): Likewise.
28273         (ffs): Likewise.
28274         (ffs_f): Likewise.
28275         (fls): Likewise.
28276         (call_i): Remove 'S' asm letter, add jli instruction.
28277         (call_value_i): Likewise.
28278         * config/arc/arc.op (mjli-always): New option.
28279         * config/arc/constraints.md (Cji): New constraint.
28280         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
28281         operand.
28282         (subsf3_fpx): Likewise.
28283         (mulsf3_fpx): Likewise.
28284         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
28285         asm operand.
28286         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
28287         function attrbutes.
28288         * doc/invoke.texi (ARC): Document mjli-always option.
28290 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
28292         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
28293         avoid addition with 0 and use incw and decw where possible.
28295 2018-01-26  Richard Biener  <rguenther@suse.de>
28297         PR tree-optimization/81082
28298         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
28299         association if it requires casting to unsigned.
28300         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
28301         from fold_plusminus_mult_expr to catch important cases late when
28302         range info is available.
28304 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
28306         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
28307         * configure.ac (hidden_linkonce): New test.
28308         * configure: Regenerate.
28309         * config.in: Regenerate.
28311 2018-01-26  Julia Koval  <julia.koval@intel.com>
28313         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
28314         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
28315         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
28316         _mm_mask_bitshuffle_epi64_mask): Fix type.
28317         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
28318         USI_FTYPE_V4DI_V4DI_USI): Remove.
28319         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
28320         __builtin_ia32_vpshufbitqmb256_mask,
28321         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
28322         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
28323         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
28325 2018-01-26  Alan Modra  <amodra@gmail.com>
28327         PR target/84033
28328         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
28329         UNSPEC_VBPERMQ.  Sort other unspecs.
28331 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
28333         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
28335 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
28337         PR middle-end/83055
28338         * predict.c (drop_profile): Do not push/pop cfun; update also
28339         node->count.
28340         (handle_missing_profiles): Fix logic looking for zero profiles.
28342 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
28344         PR middle-end/83977
28345         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
28346         on functions with #pragma omp declare simd or functions with simd
28347         attribute.
28348         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
28349         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
28350         Remove trailing \n from warning_at calls.
28352 2018-01-25  Tom de Vries  <tom@codesourcery.com>
28354         PR target/84028
28355         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
28356         for neutered workers.
28358 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
28360         PR target/68467
28361         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
28362         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
28364 2018-01-24  Jeff Law  <law@redhat.com>
28366         PR target/83994
28367         * i386.c (get_probe_interval): Move to earlier point.
28368         (ix86_compute_frame_layout): If -fstack-clash-protection and
28369         the frame is larger than the probe interval, then use pushes
28370         to save registers rather than reg->mem moves.
28371         (ix86_expand_prologue): Remove conditional for int_registers_saved
28372         assertion.
28374 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
28376         PR target/84014
28377         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
28378         min/max for never referenced object.
28380 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
28382         PR middle-end/83977
28383         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
28384         here.
28385         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
28386         attributes from DECL_ATTRIBUTES (decl) without affecting
28387         DECL_ATTRIBUTES (current_function_decl).
28388         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
28389         functions with non-NULL DECL_ABSTRACT_ORIGIN.
28391 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
28393         PR tree-optimization/83979
28394         * fold-const.c (fold_comparison): Use constant_boolean_node
28395         instead of boolean_{true,false}_node.
28397 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
28399         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
28400         with zero counts.
28402 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28404         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
28405         Simplify the clause that sets the length attribute.
28406         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
28407         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
28408         clause that sets the length attribute.
28409         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
28411 2018-01-24  Tom de Vries  <tom@codesourcery.com>
28413         PR target/83589
28414         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
28415         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
28416         Add strict parameter.
28417         (prevent_branch_around_nothing): Insert dummy insn between branch to
28418         label and label with no ptx insn inbetween.
28419         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
28421 2018-01-24  Tom de Vries  <tom@codesourcery.com>
28423         PR target/81352
28424         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
28425         for neutered threads in warp.
28426         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
28428 2018-01-24  Richard Biener  <rguenther@suse.de>
28430         PR tree-optimization/83176
28431         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
28432         operands.
28434 2018-01-24  Richard Biener  <rguenther@suse.de>
28436         PR tree-optimization/82819
28437         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
28438         code generating pluses that are no-ops in the target precision.
28440 2018-01-24  Richard Biener  <rguenther@suse.de>
28442         PR middle-end/84000
28443         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
28445 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
28447         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
28448         to merge probabilities.
28449         * predict.c (probably_never_executed): Also mark as cold functions
28450         with global 0 profile and guessed local profile.
28451         * profile-count.c (profile_probability::combine_with_count): New
28452         member function.
28453         * profile-count.h (profile_probability::operator*,
28454         profile_probability::operator*=, profile_probability::operator/,
28455         profile_probability::operator/=): Reduce precision to adjusted
28456         and set value to guessed on contradictory divisions.
28457         (profile_probability::combine_with_freq): Remove.
28458         (profile_probability::combine_wiht_count): Declare.
28459         (profile_count::force_nonzero):: Set to adjusted.
28460         (profile_count::probability_in):: Set quality to adjusted.
28461         * tree-ssa-tail-merge.c (replace_block_by): Use
28462         combine_with_count.
28464 2018-01-23  Andrew Waterman  <andrew@sifive.com>
28465             Jim Wilson  <jimw@sifive.com>
28467         * config/riscv/riscv.c (riscv_stack_boundary): New.
28468         (riscv_option_override): Set riscv_stack_boundary.  Handle
28469         riscv_preferred_stack_boundary_arg.
28470         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
28471         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
28472         (STACK_BOUNDARY): Set to riscv_stack_boundary.
28473         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
28474         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
28475         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
28477 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
28479         PR target/83905
28480         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
28481         of struct ix86_frame.
28482         (ix86_expand_epilogue): Likewise.  Add a local variable for
28483         the reg_save_offset field in struct ix86_frame.
28485 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
28487         PR tree-optimization/82604
28488         * tree-loop-distribution.c (enum partition_kind): New enum item
28489         PKIND_PARTIAL_MEMSET.
28490         (partition_builtin_p): Support above new enum item.
28491         (generate_code_for_partition): Ditto.
28492         (compute_access_range): Differentiate cases that equality can be
28493         proven at all loops, the innermost loops or no loops.
28494         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
28495         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
28496         (finalize_partitions, distribute_loop): Don't fuse partition of
28497         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
28498         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
28499         parloop is enabled.
28501 2018-01-23  Martin Liska  <mliska@suse.cz>
28503         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
28504         order to ignore the predictor.
28505         (PRED_POLYMORPHIC_CALL): Likewise.
28506         (PRED_RECURSIVE_CALL): Likewise.
28508 2018-01-23  Martin Liska  <mliska@suse.cz>
28510         * tree-profile.c (tree_profiling): Print function header to
28511         aware reader which function we are working on.
28512         * value-prof.c (gimple_find_values_to_profile): Do not print
28513         not interesting value histograms.
28515 2018-01-23  Martin Liska  <mliska@suse.cz>
28517         * profile-count.h (enum profile_quality): Add
28518         profile_uninitialized as the first value. Do not number values
28519         as they are zero based.
28520         (profile_count::verify): Update sanity check.
28521         (profile_probability::verify): Likewise.
28523 2018-01-23  Nathan Sidwell  <nathan@acm.org>
28525         * doc/invoke.texi (ffor-scope): Deprecate.
28527 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
28529         PR tree-optimization/83510
28530         * domwalk.c (set_all_edges_as_executable): New function.
28531         (dom_walker::dom_walker): Convert bool param
28532         "skip_unreachable_blocks" to enum reachability.  Move setup of
28533         edge flags to set_all_edges_as_executable and only do it when
28534         reachability is REACHABLE_BLOCKS.
28535         * domwalk.h (enum dom_walker::reachability): New enum.
28536         (dom_walker::dom_walker): Convert bool param
28537         "skip_unreachable_blocks" to enum reachability.
28538         (set_all_edges_as_executable): New decl.
28539         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
28540         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
28541         "reachability".
28542         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
28543         but converting true to REACHABLE_BLOCKS.
28544         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
28545         * tree-vrp.c
28546         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
28547         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
28548         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
28549         REACHABLE_BLOCKS.
28550         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
28551         if check_all_array_refs will be called.
28553 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
28555         * tree.c (selftest::test_location_wrappers): Add more test
28556         coverage.
28558 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
28560         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
28561         (selftest::test_bit_in_range): Likewise.
28563 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
28565         PR testsuite/83888
28566         * doc/sourcebuild.texi (vect_float): Say that the selector
28567         only describes the situation when -funsafe-math-optimizations is on.
28568         (vect_float_strict): Document.
28570 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
28572         PR tree-optimization/83965
28573         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
28574         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
28575         instead of checking only for a reduction.
28576         (vect_recog_widen_sum_pattern): Likewise.
28578 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
28580         * predict.c (probably_never_executed): Only use precise profile info.
28581         (compute_function_frequency): Skip after inlining hack since we now
28582         have quality checking.
28584 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
28586         * profile-count.h (profile_probability::very_unlikely,
28587         profile_probability::unlikely, profile_probability::even): Set
28588         precision to guessed.
28590 2018-01-23  Richard Biener  <rguenther@suse.de>
28592         PR tree-optimization/83963
28593         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
28594         Properly terminate dominator walk when crossing the exit edge not
28595         when visiting its source block.
28597 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
28599         PR c++/83918
28600         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
28601         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
28603 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
28605         PR tree-optimization/83957
28606         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
28607         semicolon after for body surrounded by braces.
28609         PR tree-optimization/83081
28610         * profile-count.h (profile_probability::split): New method.
28611         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
28612         Use profile_probability::split.
28613         (do_compare_rtx_and_jump): Fix adjustment of probabilities
28614         when splitting a single conditional jump into 2.
28616 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
28618         PR tree-optimization/69452
28619         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
28620         decl.
28622 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28624         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
28625         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
28626         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
28628 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28630         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
28631         declaration.
28632         * config/rl78/rl78.md (movdi): New define_expand.
28633         * config/rl78/rl78.c (rl78_split_movdi): New function.
28635 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
28637         PR target/83862
28638         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
28639         no longer used.
28640         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
28641         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
28642         128-bit to produce an UNSPEC move to get the double word with the
28643         signbit and then a shift directly to do signbit.
28644         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
28645         implementation with a new version that just does either a direct
28646         move or a regular move.  Move memory interface to separate insns.
28647         Move insns so they are next to the expander.
28648         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
28649         with signbit move.  Split big and little endian case.
28650         (signbit<mode>2_dm_mem_le): Likewise.
28651         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
28652         (signbit<mode>2_dm2): Likewise.
28654 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28656         * config/rl78/rl78.md (anddi3): New define_expand.
28658 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28660         * config/rl78/rl78.md (umindi3): New define_expand.
28662 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28664         * config/rl78/rl78.md (smindi3): New define_expand.
28666 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28668         * config/rl78/rl78.md (smaxdi3): New define_expand.
28670 2018-01-22  Carl Love  <cel@us.ibm.com>
28672         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
28673         LVX_V1TI): Add macro expansion.
28674         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
28675         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
28676         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
28677         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
28678         Change check to determine if the instruction is a byte reversing
28679         entry.  Fix typo in comment.
28680         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
28681         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
28682         Add def_builtin calls for new builtins.
28683         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
28684         Add define_insn expansion.
28686 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28688         * config/rl78/rl78.md (umaxdi3): New define_expand.
28690 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
28692         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
28693         for non-QImode registers.
28695 2018-01-22  Richard Biener  <rguenther@suse.de>
28697         PR tree-optimization/83963
28698         * graphite-scop-detection.c (scop_detection::get_sese): Delay
28699         including the loop exit block.
28700         (scop_detection::merge_sese): Likewise.
28701         (scop_detection::add_scop): Do it here instead.
28703 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28705         * doc/sourcebuild.texi (arm_softfloat): Document.
28707 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
28709         PR gcc/77734
28710         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
28711         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
28712         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
28714 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28715             David Edelsohn  <dje.gcc@gmail.com>
28717         PR target/83946
28718         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
28719         Change "crset eq" to "crset 2".
28720         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
28721         (*call_indirect_aix<mode>_nospec): Likewise.
28722         (*call_value_indirect_aix<mode>_nospec): Likewise.
28723         (*call_indirect_elfv2<mode>_nospec): Likewise.
28724         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
28725         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
28726         change assembly output from . to $.
28727         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
28728         (indirect_jump<mode>_nospec): Change assembly output from . to $.
28729         (*tablejump<mode>_internal1_nospec): Likewise.
28731 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
28733         PR target/80870
28734         * config/sh/sh_optimize_sett_clrt.cc:
28735         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
28737 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
28739         PR tree-optimization/83940
28740         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
28741         offset_dt to vect_constant_def rather than vect_unknown_def_type.
28742         (vect_check_load_store_mask): Add a mask_dt_out parameter and
28743         use it to pass back the definition type.
28744         (vect_check_store_rhs): Likewise rhs_dt_out.
28745         (vect_build_gather_load_calls): Add a mask_dt argument and use
28746         it instead of a call to vect_is_simple_use.
28747         (vectorizable_store): Update calls to vect_check_load_store_mask
28748         and vect_check_store_rhs.  Use the dt returned by the latter instead
28749         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
28750         instead of calls to vect_is_simple_use.  Pass the scalar rather
28751         than the vector operand to vect_is_simple_use when handling
28752         second and subsequent copies of an rhs value.
28753         (vectorizable_load): Update calls to vect_check_load_store_mask
28754         and vect_build_gather_load_calls.  Use the cached mask_dt and
28755         gs_info.offset_dt instead of calls to vect_is_simple_use.
28757 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
28759         PR middle-end/83945
28760         * tree-emutls.c: Include gimplify.h.
28761         (lower_emutls_2): New function.
28762         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
28763         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
28764         it before further processing.
28766         PR target/83930
28767         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
28768         UINTVAL (trueop1) instead of INTVAL (op1).
28770 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
28772         PR debug/81570
28773         PR debug/83728
28774         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
28775         INCOMING_FRAME_SP_OFFSET if not defined.
28776         (scan_trace): Add ENTRY argument.  If true and
28777         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
28778         emit a note to adjust the CFA offset.
28779         (create_cfi_notes): Adjust scan_trace callers.
28780         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
28781         INCOMING_FRAME_SP_OFFSET in the CIE.
28782         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
28783         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
28784         Likewise.
28785         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
28786         * doc/tm.texi: Regenerated.
28788 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
28790         PR rtl-optimization/83147
28791         * lra-constraints.c (remove_inheritance_pseudos): Use
28792         lra_substitute_pseudo_within_insn.
28794 2018-01-19  Tom de Vries  <tom@codesourcery.com>
28795             Cesar Philippidis  <cesar@codesourcery.com>
28797         PR target/83920
28798         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
28800 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
28802         PR target/83790
28803         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
28804         spaces for function labels.
28806 2018-01-19  Martin Liska  <mliska@suse.cz>
28808         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
28809         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
28810         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
28811         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
28812         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
28813         (PRED_CONST_RETURN): Change from 69 to 65.
28814         (PRED_NULL_RETURN): Change from 91 to 71.
28815         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
28816         (PRED_LOOP_GUARD): Change from 66 to 73.
28818 2018-01-19  Martin Liska  <mliska@suse.cz>
28820         * predict.c (predict_insn_def): Add new assert.
28821         (struct branch_predictor): Change type to signed integer.
28822         (test_prediction_value_range): Amend test to cover
28823         PROB_UNINITIALIZED.
28824         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
28825         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
28826         (PRED_LOOP_ITERATIONS_MAX): Likewise.
28827         (PRED_LOOP_IV_COMPARE): Likewise.
28828         * predict.h (PROB_UNINITIALIZED): Define new constant.
28830 2018-01-19  Martin Liska  <mliska@suse.cz>
28832         * predict.c (dump_prediction): Add new format for
28833         analyze_brprob.py script which is enabled with -details
28834         suboption.
28835         * profile-count.h (precise_p): New function.
28837 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
28839         PR tree-optimization/83922
28840         * tree-vect-loop.c (vect_verify_full_masking): Return false if
28841         there are no statements that need masking.
28842         (vect_active_double_reduction_p): New function.
28843         (vect_analyze_loop_operations): Use it when handling phis that
28844         are not in the loop header.
28846 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
28848         PR tree-optimization/83914
28849         * tree-vect-loop.c (vectorizable_induction): Don't convert
28850         init_expr or apply the peeling adjustment for inductions
28851         that are nested within the vectorized loop.
28853 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28855         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
28856         instead of NEG.
28858 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
28860         PR sanitizer/81715
28861         PR testsuite/83882
28862         * function.h (gimplify_parameters): Add gimple_seq * argument.
28863         * function.c: Include gimple.h and options.h.
28864         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
28865         for the added local temporaries if needed.
28866         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
28867         if there are any parameter cleanups, wrap whole body into a
28868         try/finally with the cleanups.
28870 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
28872         PR target/82964
28873         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
28874         Use GET_MODE_CLASS for scalar floating point.
28876 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
28878         PR ipa/82256
28879         patch by PaX Team
28880         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
28881         Fix call of call_cgraph_insertion_hooks.
28883 2018-01-18  Martin Sebor  <msebor@redhat.com>
28885         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
28887 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
28889         PR ipa/83619
28890         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
28891         frequencies.
28893 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
28895         PR other/70268
28896         * common.opt: (-ffile-prefix-map): New option.
28897         * opts.c (common_handle_option): Defer it.
28898         * opts-global.c (handle_common_deferred_options): Handle it.
28899         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
28900         * file-prefix-map.h: New file.
28901         (remap_debug_filename, add_debug_prefix_map): ...here.
28902         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
28903         * final.c (debug_prefix_map, add_debug_prefix_map
28904         remap_debug_filename): Move to...
28905         * file-prefix-map.c: New file.
28906         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
28907         generalize, get rid of alloca(), use strrchr() instead of strchr().
28908         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
28909         Implement in terms of add_prefix_map().
28910         (remap_macro_filename, remap_debug_filename): Implement in term of
28911         remap_filename().
28912         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
28913         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
28914         * dbxout.c: Include file-prefix-map.h.
28915         * varasm.c: Likewise.
28916         * vmsdbgout.c: Likewise.
28917         * xcoffout.c: Likewise.
28918         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
28919         * doc/cppopts.texi (-fmacro-prefix-map): Document.
28920         * doc/invoke.texi (-ffile-prefix-map): Document.
28921         (-fdebug-prefix-map): Update description.
28923 2018-01-18  Martin Liska  <mliska@suse.cz>
28925         * config/i386/i386.c (indirect_thunk_name): Document that also
28926         lfence is emitted.
28927         (output_indirect_thunk): Document why both instructions
28928         (pause and lfence) are generated.
28930 2018-01-18  Richard Biener  <rguenther@suse.de>
28932         PR tree-optimization/83887
28933         * graphite-scop-detection.c
28934         (scop_detection::get_nearest_dom_with_single_entry): Remove.
28935         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
28936         (scop_detection::merge_sese): Re-implement with a flood-fill
28937         algorithm that properly finds a SESE region if it exists.
28939 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
28941         PR c/61240
28942         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
28943         pointer_diff optimizations use view_convert instead of convert.
28945 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
28947         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
28948         Generate different code for -mno-speculate-indirect-jumps.
28949         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
28950         (*call_indirect_aix<mode>): Disable for
28951         -mno-speculate-indirect-jumps.
28952         (*call_indirect_aix<mode>_nospec): New define_insn.
28953         (*call_value_indirect_aix<mode>): Disable for
28954         -mno-speculate-indirect-jumps.
28955         (*call_value_indirect_aix<mode>_nospec): New define_insn.
28956         (*sibcall_nonlocal_sysv<mode>): Generate different code for
28957         -mno-speculate-indirect-jumps.
28958         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
28960 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
28962         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
28963         long double type, set the flags for noting the default long double
28964         type, even if we don't pass or return a long double type.
28966 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
28968         PR ipa/83051
28969         * ipa-inline.c (flatten_function): Do not overwrite final inlining
28970         failure.
28972 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
28974         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
28975         support for merge[hl].
28976         (fold_mergehl_helper): New helper function.
28977         (tree-vector-builder.h): New #include for tree_vector_builder usage.
28978         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
28979         (altivec_vmrglw_direct): Add xxmrglw insn.
28981 2018-01-17  Andrew Waterman  <andrew@sifive.com>
28983         * config/riscv/riscv.c (riscv_conditional_register_usage): If
28984         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
28986 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
28988         PR lto/83121
28989         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
28990         call the lto_location_cache before reading the
28991         DECL_SOURCE_LOCATION of the types.
28993 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
28994             Richard Sandiford  <richard.sandiford@linaro.org>
28996         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
28997         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
28998         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
28999         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
29000         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
29001         Add declaration.
29002         * config/aarch64/constraints.md (aarch64_movti_operand):
29003         Limit immediates.
29004         * config/aarch64/predicates.md (Uti): Add new constraint.
29006 2018-01-17  Carl Love  <cel@us.ibm.com>
29008         * config/rs6000/vsx.md (define_expand xl_len_r,
29009         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
29010         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
29011         lxvll.
29012         (define_expand, define_insn): Move the shift left from  the
29013         define_insn to the define_expand for lxvl and stxvl instructions.
29014         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
29015         and XL_LEN_R definitions to PURE.
29017 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
29019         * config/i386/i386.c (indirect_thunk_name): Declare regno
29020         as unsigned int.  Compare regno with INVALID_REGNUM.
29021         (output_indirect_thunk): Ditto.
29022         (output_indirect_thunk_function): Ditto.
29023         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
29024         in the call to output_indirect_thunk_function.
29026 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
29028         PR middle-end/83884
29029         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
29030         rather than the size of inner_type to determine the stack slot size
29031         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
29033 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
29035         PR target/83546
29036         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
29037         to PTA_SILVERMONT.
29039 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
29041         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
29042         endian Linux systems to optionally enable multilibs for selecting
29043         the long double type if the user configured an explicit type.
29044         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
29045         have no long double multilibs if not defined.
29046         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
29047         warn if the user used -mabi={ieee,ibm}longdouble and we built
29048         multilibs for long double.
29049         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
29050         appropriate multilib option.
29051         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
29052         multilib options.
29053         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
29054         for building long double multilibs.
29055         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
29057 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
29059         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
29060         copies.
29062         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
29063         64 bits.
29064         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
29065         128 bits.
29067         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
29068         variables.
29070         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
29071         return value.
29073 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
29075         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
29076         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
29078 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
29080         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
29081         different rtl trees depending on TARGET_64BIT.
29082         (rs6000_gen_lvx): Likewise.
29084 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
29086         * config/visium/visium.md (nop): Tweak comment.
29087         (hazard_nop): Likewise.
29089 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29091         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
29092         -mspeculate-indirect-jumps.
29093         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
29094         for -mno-speculate-indirect-jumps.
29095         (*call_indirect_elfv2<mode>_nospec): New define_insn.
29096         (*call_value_indirect_elfv2<mode>): Disable for
29097         -mno-speculate-indirect-jumps.
29098         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
29099         (indirect_jump): Emit different RTL for
29100         -mno-speculate-indirect-jumps.
29101         (*indirect_jump<mode>): Disable for
29102         -mno-speculate-indirect-jumps.
29103         (*indirect_jump<mode>_nospec): New define_insn.
29104         (tablejump): Emit different RTL for
29105         -mno-speculate-indirect-jumps.
29106         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
29107         (tablejumpsi_nospec): New define_expand.
29108         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
29109         (tablejumpdi_nospec): New define_expand.
29110         (*tablejump<mode>_internal1): Disable for
29111         -mno-speculate-indirect-jumps.
29112         (*tablejump<mode>_internal1_nospec): New define_insn.
29113         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
29114         option.
29116 2018-01-16  Artyom Skrobov tyomitch@gmail.com
29118         * caller-save.c (insert_save): Drop unnecessary parameter.  All
29119         callers updated.
29121 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
29122             Richard Biener  <rguenth@suse.de>
29124         PR libgomp/83590
29125         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
29126         return early, inline manually is_gimple_sizepos.  Make sure if we
29127         call gimplify_expr we don't end up with a gimple constant.
29128         * tree.c (variably_modified_type_p): Don't return true for
29129         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
29130         * gimplify.h (is_gimple_sizepos): Remove.
29132 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
29134         PR tree-optimization/83857
29135         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
29136         vectorizable_live_operation for pure SLP statements.
29137         (vectorizable_live_operation): Handle PHIs.
29139 2018-01-16  Richard Biener  <rguenther@suse.de>
29141         PR tree-optimization/83867
29142         * tree-vect-stmts.c (vect_transform_stmt): Precompute
29143         nested_in_vect_loop_p since the scalar stmt may get invalidated.
29145 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
29147         PR c/83844
29148         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
29149         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
29150         If off is not INTEGER_CST, issue a may not be aligned warning
29151         rather than isn't aligned.  Use isn%'t rather than isn't.
29152         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
29153         into MULT_EXPR.
29154         <case MULT_EXPR>: Improve the case when bottom and one of the
29155         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
29156         operand, in that case check if the other operand is multiple of
29157         bottom divided by the INTEGER_CST operand.
29159 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
29161         PR target/83858
29162         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
29163         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
29164         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
29165         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
29166         * config/pa/pa.c (pa_function_arg_advance): Likewise.
29167         (pa_function_arg, pa_arg_partial_bytes): Likewise.
29168         (pa_function_arg_size): New function.
29170 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
29172         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
29173         in a separate statement.
29175 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
29177         PR tree-optimization/83847
29178         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
29179         group gathers and scatters.
29181 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
29183         PR rtl-optimization/86620
29184         * params.def (max-sched-ready-insns): Bump minimum value to 1.
29186         PR rtl-optimization/83213
29187         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
29188         to last if both are JUMP_INSNs.
29190         PR tree-optimization/83843
29191         * gimple-ssa-store-merging.c
29192         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
29193         store_immediate_info for bswap/nop orig_stores.
29195 2018-01-15  Andrew Waterman  <andrew@sifive.com>
29197         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
29198         !TARGET_MUL.
29199         <UDIV>: Increase cost if !TARGET_DIV.
29201 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
29203         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
29204         (define_attr "cr_logical_3op"): New.
29205         (cceq_ior_compare): Adjust.
29206         (cceq_ior_compare_complement): Adjust.
29207         (*cceq_rev_compare): Adjust.
29208         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
29209         (is_cracked_insn): Adjust.
29210         (insn_must_be_first_in_group): Adjust.
29211         * config/rs6000/40x.md: Adjust.
29212         * config/rs6000/440.md: Adjust.
29213         * config/rs6000/476.md: Adjust.
29214         * config/rs6000/601.md: Adjust.
29215         * config/rs6000/603.md: Adjust.
29216         * config/rs6000/6xx.md: Adjust.
29217         * config/rs6000/7450.md: Adjust.
29218         * config/rs6000/7xx.md: Adjust.
29219         * config/rs6000/8540.md: Adjust.
29220         * config/rs6000/cell.md: Adjust.
29221         * config/rs6000/e300c2c3.md: Adjust.
29222         * config/rs6000/e500mc.md: Adjust.
29223         * config/rs6000/e500mc64.md: Adjust.
29224         * config/rs6000/e5500.md: Adjust.
29225         * config/rs6000/e6500.md: Adjust.
29226         * config/rs6000/mpc.md: Adjust.
29227         * config/rs6000/power4.md: Adjust.
29228         * config/rs6000/power5.md: Adjust.
29229         * config/rs6000/power6.md: Adjust.
29230         * config/rs6000/power7.md: Adjust.
29231         * config/rs6000/power8.md: Adjust.
29232         * config/rs6000/power9.md: Adjust.
29233         * config/rs6000/rs64.md: Adjust.
29234         * config/rs6000/titan.md: Adjust.
29236 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
29238         * config/i386/predicates.md (indirect_branch_operand): Rewrite
29239         ix86_indirect_branch_register logic.
29241 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
29243         * config/i386/constraints.md (Bs): Update
29244         ix86_indirect_branch_register check.  Don't check
29245         ix86_indirect_branch_register with GOT_memory_operand.
29246         (Bw): Likewise.
29247         * config/i386/predicates.md (GOT_memory_operand): Don't check
29248         ix86_indirect_branch_register here.
29249         (GOT32_symbol_operand): Likewise.
29251 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
29253         * config/i386/predicates.md (constant_call_address_operand):
29254         Rewrite ix86_indirect_branch_register logic.
29255         (sibcall_insn_operand): Likewise.
29257 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
29259         * config/i386/constraints.md (Bs): Replace
29260         ix86_indirect_branch_thunk_register with
29261         ix86_indirect_branch_register.
29262         (Bw): Likewise.
29263         * config/i386/i386.md (indirect_jump): Likewise.
29264         (tablejump): Likewise.
29265         (*sibcall_memory): Likewise.
29266         (*sibcall_value_memory): Likewise.
29267         Peepholes of indirect call and jump via memory: Likewise.
29268         * config/i386/i386.opt: Likewise.
29269         * config/i386/predicates.md (indirect_branch_operand): Likewise.
29270         (GOT_memory_operand): Likewise.
29271         (call_insn_operand): Likewise.
29272         (sibcall_insn_operand): Likewise.
29273         (GOT32_symbol_operand): Likewise.
29275 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
29277         PR middle-end/83837
29278         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
29279         type rather than type addr's type points to.
29280         (expand_omp_atomic_mutex): Likewise.
29281         (expand_omp_atomic): Likewise.
29283 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
29285         PR target/83839
29286         * config/i386/i386.c (output_indirect_thunk_function): Use
29287         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
29288         for  __x86_return_thunk.
29290 2018-01-15  Richard Biener  <rguenther@suse.de>
29292         PR middle-end/83850
29293         * expmed.c (extract_bit_field_1): Fix typo.
29295 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29297         PR target/83687
29298         * config/arm/iterators.md (VF): New mode iterator.
29299         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
29300         Remove integer-related logic from pattern.
29301         (neon_vabd<mode>_3): Likewise.
29303 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
29305         PR middle-end/82694
29306         * common.opt (fstrict-overflow): No longer an alias.
29307         (fwrapv-pointer): New option.
29308         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
29309         also for pointer types based on flag_wrapv_pointer.
29310         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
29311         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
29312         opts->x_flag_wrapv got set.
29313         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
29314         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
29315         POINTER_TYPE_OVERFLOW_UNDEFINED.
29316         * match.pd: Likewise in address comparison pattern.
29317         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
29319 2018-01-15  Richard Biener  <rguenther@suse.de>
29321         PR lto/83804
29322         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
29323         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
29324         Reset type names to their identifier if their TYPE_DECL doesn't
29325         have linkage (and thus is used for ODR and devirt).
29326         (save_debug_info_for_decl): Remove.
29327         (save_debug_info_for_type): Likewise.
29328         (add_tree_to_fld_list): Adjust.
29329         * tree-pretty-print.c (dump_generic_node): Make dumping of
29330         type names more robust.
29332 2018-01-15  Richard Biener  <rguenther@suse.de>
29334         * BASE-VER: Bump to 8.0.1.
29336 2018-01-14  Martin Sebor  <msebor@redhat.com>
29338         PR other/83508
29339         * builtins.c (check_access): Avoid warning when the no-warning bit
29340         is set.
29342 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
29344         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
29345         * ira-color (allocno_hard_regs_compare): Likewise.
29347 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
29349         PR target/83013
29350         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
29351         Use .pushsection/.popsection.
29353 2018-01-14  Martin Sebor  <msebor@redhat.com>
29355         PR c++/81327
29356         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
29358 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
29360         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
29361         entry from extra_headers.
29362         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
29363         extra_headers, make the list bitwise identical to the i?86-*-* one.
29365 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
29367         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
29368         -mcmodel=large with -mindirect-branch=thunk,
29369         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
29370         -mfunction-return=thunk-extern.
29371         * doc/invoke.texi: Document -mcmodel=large is incompatible with
29372         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
29373         -mfunction-return=thunk and -mfunction-return=thunk-extern.
29375 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
29377         * config/i386/i386.c (print_reg): Print the name of the full
29378         integer register without '%'.
29379         (ix86_print_operand): Handle 'V'.
29380         * doc/extend.texi: Document 'V' modifier.
29382 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
29384         * config/i386/constraints.md (Bs): Disallow memory operand for
29385         -mindirect-branch-register.
29386         (Bw): Likewise.
29387         * config/i386/predicates.md (indirect_branch_operand): Likewise.
29388         (GOT_memory_operand): Likewise.
29389         (call_insn_operand): Likewise.
29390         (sibcall_insn_operand): Likewise.
29391         (GOT32_symbol_operand): Likewise.
29392         * config/i386/i386.md (indirect_jump): Call convert_memory_address
29393         for -mindirect-branch-register.
29394         (tablejump): Likewise.
29395         (*sibcall_memory): Likewise.
29396         (*sibcall_value_memory): Likewise.
29397         Disallow peepholes of indirect call and jump via memory for
29398         -mindirect-branch-register.
29399         (*call_pop): Replace m with Bw.
29400         (*call_value_pop): Likewise.
29401         (*sibcall_pop_memory): Replace m with Bs.
29402         * config/i386/i386.opt (mindirect-branch-register): New option.
29403         * doc/invoke.texi: Document -mindirect-branch-register option.
29405 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
29407         * config/i386/i386-protos.h (ix86_output_function_return): New.
29408         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
29409         set function_return_type.
29410         (indirect_thunk_name): Add ret_p to indicate thunk for function
29411         return.
29412         (output_indirect_thunk_function): Pass false to
29413         indirect_thunk_name.
29414         (ix86_output_indirect_branch_via_reg): Likewise.
29415         (ix86_output_indirect_branch_via_push): Likewise.
29416         (output_indirect_thunk_function): Create alias for function
29417         return thunk if regno < 0.
29418         (ix86_output_function_return): New function.
29419         (ix86_handle_fndecl_attribute): Handle function_return.
29420         (ix86_attribute_table): Add function_return.
29421         * config/i386/i386.h (machine_function): Add
29422         function_return_type.
29423         * config/i386/i386.md (simple_return_internal): Use
29424         ix86_output_function_return.
29425         (simple_return_internal_long): Likewise.
29426         * config/i386/i386.opt (mfunction-return=): New option.
29427         (indirect_branch): Mention -mfunction-return=.
29428         * doc/extend.texi: Document function_return function attribute.
29429         * doc/invoke.texi: Document -mfunction-return= option.
29431 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
29433         * config/i386/i386-opts.h (indirect_branch): New.
29434         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
29435         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
29436         with local indirect jump when converting indirect call and jump.
29437         (ix86_set_indirect_branch_type): New.
29438         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
29439         (indirectlabelno): New.
29440         (indirect_thunk_needed): Likewise.
29441         (indirect_thunk_bnd_needed): Likewise.
29442         (indirect_thunks_used): Likewise.
29443         (indirect_thunks_bnd_used): Likewise.
29444         (INDIRECT_LABEL): Likewise.
29445         (indirect_thunk_name): Likewise.
29446         (output_indirect_thunk): Likewise.
29447         (output_indirect_thunk_function): Likewise.
29448         (ix86_output_indirect_branch_via_reg): Likewise.
29449         (ix86_output_indirect_branch_via_push): Likewise.
29450         (ix86_output_indirect_branch): Likewise.
29451         (ix86_output_indirect_jmp): Likewise.
29452         (ix86_code_end): Call output_indirect_thunk_function if needed.
29453         (ix86_output_call_insn): Call ix86_output_indirect_branch if
29454         needed.
29455         (ix86_handle_fndecl_attribute): Handle indirect_branch.
29456         (ix86_attribute_table): Add indirect_branch.
29457         * config/i386/i386.h (machine_function): Add indirect_branch_type
29458         and has_local_indirect_jump.
29459         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
29460         to true.
29461         (tablejump): Likewise.
29462         (*indirect_jump): Use ix86_output_indirect_jmp.
29463         (*tablejump_1): Likewise.
29464         (simple_return_indirect_internal): Likewise.
29465         * config/i386/i386.opt (mindirect-branch=): New option.
29466         (indirect_branch): New.
29467         (keep): Likewise.
29468         (thunk): Likewise.
29469         (thunk-inline): Likewise.
29470         (thunk-extern): Likewise.
29471         * doc/extend.texi: Document indirect_branch function attribute.
29472         * doc/invoke.texi: Document -mindirect-branch= option.
29474 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
29476         PR ipa/83051
29477         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
29479 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
29481         * ipa-inline.c (want_inline_small_function_p): Return false if
29482         inlining has already failed with CIF_FINAL_ERROR.
29483         (update_caller_keys): Call want_inline_small_function_p before
29484         can_inline_edge_p.
29485         (update_callee_keys): Likewise.
29487 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
29489         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
29490         New function.
29491         (rs6000_quadword_masked_address_p): Likewise.
29492         (quad_aligned_load_p): Likewise.
29493         (quad_aligned_store_p): Likewise.
29494         (const_load_sequence_p): Add comment to describe the outer-most loop.
29495         (mimic_memory_attributes_and_flags): New function.
29496         (rs6000_gen_stvx): Likewise.
29497         (replace_swapped_aligned_store): Likewise.
29498         (rs6000_gen_lvx): Likewise.
29499         (replace_swapped_aligned_load): Likewise.
29500         (replace_swapped_load_constant): Capitalize argument name in
29501         comment describing this function.
29502         (rs6000_analyze_swaps): Add a third pass to search for vector loads
29503         and stores that access quad-word aligned addresses and replace
29504         with stvx or lvx instructions when appropriate.
29505         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
29506         New function prototype.
29507         (rs6000_quadword_masked_address_p): Likewise.
29508         (rs6000_gen_lvx): Likewise.
29509         (rs6000_gen_stvx): Likewise.
29510         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
29511         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
29512         when memory address is aligned.
29513         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
29514         this split to select lvx instruction when memory address is aligned.
29515         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
29516         instruction when memory address is aligned.
29517         (*vsx_le_perm_load_v16qi): Likewise.
29518         (four unnamed splitters): Modify to select the stvx instruction
29519         when memory is aligned.
29521 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
29523         * predict.c (determine_unlikely_bbs): Handle correctly BBs
29524         which appears in the queue multiple times.
29526 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29527             Alan Hayward  <alan.hayward@arm.com>
29528             David Sherwood  <david.sherwood@arm.com>
29530         * tree-vectorizer.h (vec_lower_bound): New structure.
29531         (_loop_vec_info): Add check_nonzero and lower_bounds.
29532         (LOOP_VINFO_CHECK_NONZERO): New macro.
29533         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
29534         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
29535         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
29536         fields.  Make seg_len the distance travelled, not including the
29537         access size.
29538         (dr_direction_indicator): Declare.
29539         (dr_zero_step_indicator): Likewise.
29540         (dr_known_forward_stride_p): Likewise.
29541         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
29542         tree-ssanames.h.
29543         (runtime_alias_check_p): Allow runtime alias checks with
29544         variable strides.
29545         (operator ==): Compare access_size and align.
29546         (prune_runtime_alias_test_list): Rework for new distinction between
29547         the access_size and seg_len.
29548         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
29549         segment lengths.
29550         (get_segment_min_max): New function.
29551         (create_intersect_range_checks): Use it.
29552         (dr_step_indicator): New function.
29553         (dr_direction_indicator): Likewise.
29554         (dr_zero_step_indicator): Likewise.
29555         (dr_known_forward_stride_p): Likewise.
29556         * tree-loop-distribution.c (data_ref_segment_size): Return
29557         DR_STEP * (niters - 1).
29558         (compute_alias_check_pairs): Update call to the dr_with_seg_len
29559         constructor.
29560         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
29561         (vect_preserves_scalar_order_p): New function, split out from...
29562         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
29563         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
29564         (vect_vfa_access_size): New function.
29565         (vect_vfa_align): Likewise.
29566         (vect_compile_time_alias): Take access_size_a and access_b arguments.
29567         (dump_lower_bound): New function.
29568         (vect_check_lower_bound): Likewise.
29569         (vect_small_gap_p): Likewise.
29570         (vectorizable_with_step_bound_p): Likewise.
29571         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
29572         depencies if the vectorization factor is 1.  Convert the checks
29573         for nonzero steps into checks on the bounds of DR_STEP.  Try using
29574         a bunds check for variable steps if the minimum required step is
29575         relatively small. Update calls to the dr_with_seg_len
29576         constructor and to vect_compile_time_alias.
29577         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
29578         function.
29579         (vect_loop_versioning): Call it.
29580         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
29581         when retrying.
29582         (vect_estimate_min_profitable_iters): Account for any bounds checks.
29584 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29585             Alan Hayward  <alan.hayward@arm.com>
29586             David Sherwood  <david.sherwood@arm.com>
29588         * doc/sourcebuild.texi (vect_scatter_store): Document.
29589         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
29590         optabs.
29591         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
29592         Document.
29593         * genopinit.c (main): Add supports_vec_scatter_store and
29594         supports_vec_scatter_store_cached to target_optabs.
29595         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
29596         IFN_MASK_SCATTER_STORE.
29597         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
29598         functions.
29599         * internal-fn.h (internal_store_fn_p): Declare.
29600         (internal_fn_stored_value_index): Likewise.
29601         * internal-fn.c (scatter_store_direct): New macro.
29602         (expand_scatter_store_optab_fn): New function.
29603         (direct_scatter_store_optab_supported_p): New macro.
29604         (internal_store_fn_p): New function.
29605         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
29606         IFN_MASK_SCATTER_STORE.
29607         (internal_fn_mask_index): Likewise.
29608         (internal_fn_stored_value_index): New function.
29609         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
29610         for scatter stores.
29611         * optabs-query.h (supports_vec_scatter_store_p): Declare.
29612         * optabs-query.c (supports_vec_scatter_store_p): New function.
29613         * tree-vectorizer.h (vect_get_store_rhs): Declare.
29614         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
29615         true for scatter stores.
29616         (vect_gather_scatter_fn_p): Handle scatter stores too.
29617         (vect_check_gather_scatter): Consider using scatter stores if
29618         supports_vec_scatter_store_p.
29619         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
29620         scatter stores too.
29621         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
29622         internal_fn_stored_value_index.
29623         (check_load_store_masking): Handle scatter stores too.
29624         (vect_get_store_rhs): Make public.
29625         (vectorizable_call): Use internal_store_fn_p.
29626         (vectorizable_store): Handle scatter store internal functions.
29627         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
29628         when deciding whether the end of the group has been reached.
29629         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
29630         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
29631         (mask_scatter_store<mode>): New insns.
29633 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29634             Alan Hayward  <alan.hayward@arm.com>
29635             David Sherwood  <david.sherwood@arm.com>
29637         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
29638         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
29639         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
29640         function.
29641         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
29642         Use vect_truncate_gather_scatter_offset if we can't treat the
29643         operation as a normal gather load or scatter store.
29644         (get_group_load_store_type): Take the gather_scatter_info
29645         as argument.  Try using a gather load or scatter store for
29646         single-element groups.
29647         (get_load_store_type): Update calls to get_group_load_store_type
29648         and vect_use_strided_gather_scatters_p.
29650 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29651             Alan Hayward  <alan.hayward@arm.com>
29652             David Sherwood  <david.sherwood@arm.com>
29654         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
29655         optional tree argument.
29656         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
29657         null target hooks.
29658         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
29659         but continue to use the current value as a fallback.
29660         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
29661         to compare the updates.
29662         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
29663         (get_load_store_type): Use it when handling a strided access.
29664         (vect_get_strided_load_store_ops): New function.
29665         (vect_get_data_ptr_increment): Likewise.
29666         (vectorizable_load): Handle strided gather loads.  Always pass
29667         a step to vect_create_data_ref_ptr and bump_vector_ptr.
29669 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29670             Alan Hayward  <alan.hayward@arm.com>
29671             David Sherwood  <david.sherwood@arm.com>
29673         * doc/md.texi (gather_load@var{m}): Document.
29674         (mask_gather_load@var{m}): Likewise.
29675         * genopinit.c (main): Add supports_vec_gather_load and
29676         supports_vec_gather_load_cached to target_optabs.
29677         * optabs-tree.c (init_tree_optimization_optabs): Use
29678         ggc_cleared_alloc to allocate target_optabs.
29679         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
29680         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
29681         functions.
29682         * internal-fn.h (internal_load_fn_p): Declare.
29683         (internal_gather_scatter_fn_p): Likewise.
29684         (internal_fn_mask_index): Likewise.
29685         (internal_gather_scatter_fn_supported_p): Likewise.
29686         * internal-fn.c (gather_load_direct): New macro.
29687         (expand_gather_load_optab_fn): New function.
29688         (direct_gather_load_optab_supported_p): New macro.
29689         (direct_internal_fn_optab): New function.
29690         (internal_load_fn_p): Likewise.
29691         (internal_gather_scatter_fn_p): Likewise.
29692         (internal_fn_mask_index): Likewise.
29693         (internal_gather_scatter_fn_supported_p): Likewise.
29694         * optabs-query.c (supports_at_least_one_mode_p): New function.
29695         (supports_vec_gather_load_p): Likewise.
29696         * optabs-query.h (supports_vec_gather_load_p): Declare.
29697         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
29698         and memory_type field.
29699         (NUM_PATTERNS): Bump to 15.
29700         * tree-vect-data-refs.c: Include internal-fn.h.
29701         (vect_gather_scatter_fn_p): New function.
29702         (vect_describe_gather_scatter_call): Likewise.
29703         (vect_check_gather_scatter): Try using internal functions for
29704         gather loads.  Recognize existing calls to a gather load function.
29705         (vect_analyze_data_refs): Consider using gather loads if
29706         supports_vec_gather_load_p.
29707         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
29708         (vect_get_gather_scatter_offset_type): Likewise.
29709         (vect_convert_mask_for_vectype): Likewise.
29710         (vect_add_conversion_to_patterm): Likewise.
29711         (vect_try_gather_scatter_pattern): Likewise.
29712         (vect_recog_gather_scatter_pattern): New pattern recognizer.
29713         (vect_vect_recog_func_ptrs): Add it.
29714         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
29715         internal_fn_mask_index and internal_gather_scatter_fn_p.
29716         (check_load_store_masking): Take the gather_scatter_info as an
29717         argument and handle gather loads.
29718         (vect_get_gather_scatter_ops): New function.
29719         (vectorizable_call): Check internal_load_fn_p.
29720         (vectorizable_load): Likewise.  Handle gather load internal
29721         functions.
29722         (vectorizable_store): Update call to check_load_store_masking.
29723         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
29724         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
29725         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
29726         (aarch64_gather_scale_operand_d): New predicates.
29727         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
29728         (mask_gather_load<mode>): New insns.
29730 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29731             Alan Hayward  <alan.hayward@arm.com>
29732             David Sherwood  <david.sherwood@arm.com>
29734         * optabs.def (fold_left_plus_optab): New optab.
29735         * doc/md.texi (fold_left_plus_@var{m}): Document.
29736         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
29737         * internal-fn.c (fold_left_direct): Define.
29738         (expand_fold_left_optab_fn): Likewise.
29739         (direct_fold_left_optab_supported_p): Likewise.
29740         * fold-const-call.c (fold_const_fold_left): New function.
29741         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
29742         * tree-parloops.c (valid_reduction_p): New function.
29743         (gather_scalar_reductions): Use it.
29744         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
29745         (vect_finish_replace_stmt): Declare.
29746         * tree-vect-loop.c (fold_left_reduction_fn): New function.
29747         (needs_fold_left_reduction_p): New function, split out from...
29748         (vect_is_simple_reduction): ...here.  Accept reductions that
29749         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
29750         (vect_force_simple_reduction): Also store the reduction type in
29751         the assignment's STMT_VINFO_REDUC_TYPE.
29752         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
29753         (merge_with_identity): New function.
29754         (vect_expand_fold_left): Likewise.
29755         (vectorize_fold_left_reduction): Likewise.
29756         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
29757         scalar phi in place for it.  Check for target support and reject
29758         cases that would reassociate the operation.  Defer the transform
29759         phase to vectorize_fold_left_reduction.
29760         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
29761         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
29762         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
29764 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29766         * tree-if-conv.c (predicate_mem_writes): Remove redundant
29767         call to ifc_temp_var.
29769 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29770             Alan Hayward  <alan.hayward@arm.com>
29771             David Sherwood  <david.sherwood@arm.com>
29773         * target.def (legitimize_address_displacement): Take the original
29774         offset as a poly_int.
29775         * targhooks.h (default_legitimize_address_displacement): Update
29776         accordingly.
29777         * targhooks.c (default_legitimize_address_displacement): Likewise.
29778         * doc/tm.texi: Regenerate.
29779         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
29780         as an argument, moving assert of ad->disp == ad->disp_term to...
29781         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
29782         Try calling targetm.legitimize_address_displacement before expanding
29783         the address rather than afterwards, and adjust for the new interface.
29784         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
29785         Match the new hook interface.  Handle SVE addresses.
29786         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
29787         new hook interface.
29789 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29791         * Makefile.in (OBJS): Add early-remat.o.
29792         * target.def (select_early_remat_modes): New hook.
29793         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
29794         * doc/tm.texi: Regenerate.
29795         * targhooks.h (default_select_early_remat_modes): Declare.
29796         * targhooks.c (default_select_early_remat_modes): New function.
29797         * timevar.def (TV_EARLY_REMAT): New timevar.
29798         * passes.def (pass_early_remat): New pass.
29799         * tree-pass.h (make_pass_early_remat): Declare.
29800         * early-remat.c: New file.
29801         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
29802         function.
29803         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
29805 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29806             Alan Hayward  <alan.hayward@arm.com>
29807             David Sherwood  <david.sherwood@arm.com>
29809         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
29810         vfm1 with a bound_epilog parameter.
29811         (vect_do_peeling): Update calls accordingly, and move the prologue
29812         call earlier in the function.  Treat the base bound_epilog as 0 for
29813         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
29814         this base when peeling for gaps.
29815         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
29816         with fully-masked loops.
29817         (vect_estimate_min_profitable_iters): Handle the single peeled
29818         iteration in that case.
29820 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29821             Alan Hayward  <alan.hayward@arm.com>
29822             David Sherwood  <david.sherwood@arm.com>
29824         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
29825         single-element interleaving even if the size is not a power of 2.
29826         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
29827         accesses for single-element interleaving if the group size is
29828         not a power of 2.
29830 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29831             Alan Hayward  <alan.hayward@arm.com>
29832             David Sherwood  <david.sherwood@arm.com>
29834         * doc/md.texi (fold_extract_last_@var{m}): Document.
29835         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
29836         * optabs.def (fold_extract_last_optab): New optab.
29837         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
29838         * internal-fn.c (fold_extract_direct): New macro.
29839         (expand_fold_extract_optab_fn): Likewise.
29840         (direct_fold_extract_optab_supported_p): Likewise.
29841         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
29842         * tree-vect-loop.c (vect_model_reduction_cost): Handle
29843         EXTRACT_LAST_REDUCTION.
29844         (get_initial_def_for_reduction): Do not create an initial vector
29845         for EXTRACT_LAST_REDUCTION reductions.
29846         (vectorizable_reduction): Leave the scalar phi in place for
29847         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
29848         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
29849         epilogue code for EXTRACT_LAST_REDUCTION and defer the
29850         transform phase to vectorizable_condition.
29851         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
29852         split out from...
29853         (vect_finish_stmt_generation): ...here.
29854         (vect_finish_replace_stmt): New function.
29855         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
29856         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
29857         pattern.
29858         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
29860 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29861             Alan Hayward  <alan.hayward@arm.com>
29862             David Sherwood  <david.sherwood@arm.com>
29864         * doc/md.texi (extract_last_@var{m}): Document.
29865         * optabs.def (extract_last_optab): New optab.
29866         * internal-fn.def (EXTRACT_LAST): New internal function.
29867         * internal-fn.c (cond_unary_direct): New macro.
29868         (expand_cond_unary_optab_fn): Likewise.
29869         (direct_cond_unary_optab_supported_p): Likewise.
29870         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
29871         loops using EXTRACT_LAST.
29872         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
29873         (extract_last_<mode>): ...this optab.
29874         (vec_extract<mode><Vel>): Update accordingly.
29876 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29877             Alan Hayward  <alan.hayward@arm.com>
29878             David Sherwood  <david.sherwood@arm.com>
29880         * target.def (empty_mask_is_expensive): New hook.
29881         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
29882         * doc/tm.texi: Regenerate.
29883         * targhooks.h (default_empty_mask_is_expensive): Declare.
29884         * targhooks.c (default_empty_mask_is_expensive): New function.
29885         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
29886         if the target says that empty masks are expensive.
29887         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
29888         New function.
29889         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
29891 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29892             Alan Hayward  <alan.hayward@arm.com>
29893             David Sherwood  <david.sherwood@arm.com>
29895         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
29896         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
29897         (vect_use_loop_mask_for_alignment_p): New function.
29898         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
29899         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
29900         niters_skip argument.  Make sure that the first niters_skip elements
29901         of the first iteration are inactive.
29902         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
29903         Update call to vect_set_loop_masks_directly.
29904         (get_misalign_in_elems): New function, split out from...
29905         (vect_gen_prolog_loop_niters): ...here.
29906         (vect_update_init_of_dr): Take a code argument that specifies whether
29907         the adjustment should be added or subtracted.
29908         (vect_update_init_of_drs): Likewise.
29909         (vect_prepare_for_masked_peels): New function.
29910         (vect_do_peeling): Skip prologue peeling if we're using a mask
29911         instead.  Update call to vect_update_inits_of_drs.
29912         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
29913         mask_skip_niters.
29914         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
29915         alignment.  Do not include the number of peeled iterations in
29916         the minimum threshold in that case.
29917         (vectorizable_induction): Adjust the start value down by
29918         LOOP_VINFO_MASK_SKIP_NITERS iterations.
29919         (vect_transform_loop): Call vect_prepare_for_masked_peels.
29920         Take the number of skipped iterations into account when calculating
29921         the loop bounds.
29922         * tree-vect-stmts.c (vect_gen_while_not): New function.
29924 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29925             Alan Hayward  <alan.hayward@arm.com>
29926             David Sherwood  <david.sherwood@arm.com>
29928         * doc/sourcebuild.texi (vect_fully_masked): Document.
29929         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
29930         default value to 0.
29931         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
29932         split out from...
29933         (vect_analyze_loop_2): ...here. Don't check the vectorization
29934         factor against the number of loop iterations if the loop is
29935         fully-masked.
29937 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29938             Alan Hayward  <alan.hayward@arm.com>
29939             David Sherwood  <david.sherwood@arm.com>
29941         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
29942         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
29943         (dump_groups): Update accordingly.
29944         (iv_use::mem_type): New member variable.
29945         (address_p): New function.
29946         (record_use): Add a mem_type argument and initialize the new
29947         mem_type field.
29948         (record_group_use): Add a mem_type argument.  Use address_p.
29949         Remove obsolete null checks of base_object.  Update call to record_use.
29950         (find_interesting_uses_op): Update call to record_group_use.
29951         (find_interesting_uses_cond): Likewise.
29952         (find_interesting_uses_address): Likewise.
29953         (get_mem_type_for_internal_fn): New function.
29954         (find_address_like_use): Likewise.
29955         (find_interesting_uses_stmt): Try find_address_like_use before
29956         calling find_interesting_uses_op.
29957         (addr_offset_valid_p): Use the iv mem_type field as the type
29958         of the addressed memory.
29959         (add_autoinc_candidates): Likewise.
29960         (get_address_cost): Likewise.
29961         (split_small_address_groups_p): Use address_p.
29962         (split_address_groups): Likewise.
29963         (add_iv_candidate_for_use): Likewise.
29964         (autoinc_possible_for_pair): Likewise.
29965         (rewrite_groups): Likewise.
29966         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
29967         (determine_group_iv_cost): Update after split of USE_ADDRESS.
29968         (get_alias_ptr_type_for_ptr_address): New function.
29969         (rewrite_use_address): Rewrite address uses in calls that were
29970         identified by find_address_like_use.
29972 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29973             Alan Hayward  <alan.hayward@arm.com>
29974             David Sherwood  <david.sherwood@arm.com>
29976         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
29977         TARGET_MEM_REFs.
29978         * gimple-expr.h (is_gimple_addressable: Likewise.
29979         * gimple-expr.c (is_gimple_address): Likewise.
29980         * internal-fn.c (expand_call_mem_ref): New function.
29981         (expand_mask_load_optab_fn): Use it.
29982         (expand_mask_store_optab_fn): Likewise.
29984 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29985             Alan Hayward  <alan.hayward@arm.com>
29986             David Sherwood  <david.sherwood@arm.com>
29988         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
29989         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
29990         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
29991         (cond_umax@var{mode}): Document.
29992         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
29993         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
29994         (cond_umin_optab, cond_umax_optab): New optabs.
29995         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
29996         (COND_IOR, COND_XOR): New internal functions.
29997         * internal-fn.h (get_conditional_internal_fn): Declare.
29998         * internal-fn.c (cond_binary_direct): New macro.
29999         (expand_cond_binary_optab_fn): Likewise.
30000         (direct_cond_binary_optab_supported_p): Likewise.
30001         (get_conditional_internal_fn): New function.
30002         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
30003         Cope with reduction statements that are vectorized as calls rather
30004         than assignments.
30005         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
30006         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
30007         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
30008         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
30009         (UNSPEC_COND_EOR): New unspecs.
30010         (optab): Add mappings for them.
30011         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
30012         (sve_int_op, sve_fp_op): New int attributes.
30014 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30015             Alan Hayward  <alan.hayward@arm.com>
30016             David Sherwood  <david.sherwood@arm.com>
30018         * optabs.def (while_ult_optab): New optab.
30019         * doc/md.texi (while_ult@var{m}@var{n}): Document.
30020         * internal-fn.def (WHILE_ULT): New internal function.
30021         * internal-fn.h (direct_internal_fn_supported_p): New override
30022         that takes two types as argument.
30023         * internal-fn.c (while_direct): New macro.
30024         (expand_while_optab_fn): New function.
30025         (convert_optab_supported_p): Likewise.
30026         (direct_while_optab_supported_p): New macro.
30027         * wide-int.h (wi::udiv_ceil): New function.
30028         * tree-vectorizer.h (rgroup_masks): New structure.
30029         (vec_loop_masks): New typedef.
30030         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
30031         and fully_masked_p.
30032         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
30033         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
30034         (vect_max_vf): New function.
30035         (slpeel_make_loop_iterate_ntimes): Delete.
30036         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
30037         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
30038         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
30039         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
30040         internal-fn.h, stor-layout.h and optabs-query.h.
30041         (vect_set_loop_mask): New function.
30042         (add_preheader_seq): Likewise.
30043         (add_header_seq): Likewise.
30044         (interleave_supported_p): Likewise.
30045         (vect_maybe_permute_loop_masks): Likewise.
30046         (vect_set_loop_masks_directly): Likewise.
30047         (vect_set_loop_condition_masked): Likewise.
30048         (vect_set_loop_condition_unmasked): New function, split out from
30049         slpeel_make_loop_iterate_ntimes.
30050         (slpeel_make_loop_iterate_ntimes): Rename to..
30051         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
30052         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
30053         (vect_do_peeling): Update call accordingly.
30054         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
30055         loops.
30056         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
30057         mask_compare_type, can_fully_mask_p and fully_masked_p.
30058         (release_vec_loop_masks): New function.
30059         (_loop_vec_info): Use it to free the loop masks.
30060         (can_produce_all_loop_masks_p): New function.
30061         (vect_get_max_nscalars_per_iter): Likewise.
30062         (vect_verify_full_masking): Likewise.
30063         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
30064         retries, and free the mask rgroups before retrying.  Check loop-wide
30065         reasons for disallowing fully-masked loops.  Make the final decision
30066         about whether use a fully-masked loop or not.
30067         (vect_estimate_min_profitable_iters): Do not assume that peeling
30068         for the number of iterations will be needed for fully-masked loops.
30069         (vectorizable_reduction): Disable fully-masked loops.
30070         (vectorizable_live_operation): Likewise.
30071         (vect_halve_mask_nunits): New function.
30072         (vect_double_mask_nunits): Likewise.
30073         (vect_record_loop_mask): Likewise.
30074         (vect_get_loop_mask): Likewise.
30075         (vect_transform_loop): Handle the case in which the final loop
30076         iteration might handle a partial vector.  Call vect_set_loop_condition
30077         instead of slpeel_make_loop_iterate_ntimes.
30078         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
30079         (check_load_store_masking): New function.
30080         (prepare_load_store_mask): Likewise.
30081         (vectorizable_store): Handle fully-masked loops.
30082         (vectorizable_load): Likewise.
30083         (supportable_widening_operation): Use vect_halve_mask_nunits for
30084         booleans.
30085         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
30086         (vect_gen_while): New function.
30087         * config/aarch64/aarch64.md (umax<mode>3): New expander.
30088         (aarch64_uqdec<mode>): New insn.
30090 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30091             Alan Hayward  <alan.hayward@arm.com>
30092             David Sherwood  <david.sherwood@arm.com>
30094         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
30095         (reduc_xor_scal_optab): New optabs.
30096         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
30097         (reduc_xor_scal_@var{m}): Document.
30098         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
30099         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
30100         internal functions.
30101         * fold-const-call.c (fold_const_call): Handle them.
30102         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
30103         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
30104         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
30105         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
30106         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
30107         (UNSPEC_XORV): New unspecs.
30108         (optab): Add entries for them.
30109         (BITWISEV): New int iterator.
30110         (bit_reduc_op): New int attributes.
30112 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30113             Alan Hayward  <alan.hayward@arm.com>
30114             David Sherwood  <david.sherwood@arm.com>
30116         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
30117         * internal-fn.def (VEC_SHL_INSERT): New internal function.
30118         * optabs.def (vec_shl_insert_optab): New optab.
30119         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
30120         (duplicate_and_interleave): Likewise.
30121         * tree-vect-loop.c: Include internal-fn.h.
30122         (neutral_op_for_slp_reduction): New function, split out from
30123         get_initial_defs_for_reduction.
30124         (get_initial_def_for_reduction): Handle option 2 for variable-length
30125         vectors by loading the neutral value into a vector and then shifting
30126         the initial value into element 0.
30127         (get_initial_defs_for_reduction): Replace the code argument with
30128         the neutral value calculated by neutral_op_for_slp_reduction.
30129         Use gimple_build_vector for constant-length vectors.
30130         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
30131         but the first group_size elements have a neutral value.
30132         Use duplicate_and_interleave otherwise.
30133         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
30134         Update call to get_initial_defs_for_reduction.  Handle SLP
30135         reductions for variable-length vectors by creating one vector
30136         result for each scalar result, with the elements associated
30137         with other scalar results stubbed out with the neutral value.
30138         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
30139         Require IFN_VEC_SHL_INSERT for double reductions on
30140         variable-length vectors, or SLP reductions that have
30141         a neutral value.  Require can_duplicate_and_interleave_p
30142         support for variable-length unchained SLP reductions if there
30143         is no neutral value, such as for MIN/MAX reductions.  Also require
30144         the number of vector elements to be a multiple of the number of
30145         SLP statements when doing variable-length unchained SLP reductions.
30146         Update call to vect_create_epilog_for_reduction.
30147         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
30148         and remove initial values.
30149         (duplicate_and_interleave): Make public.
30150         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
30151         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
30153 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30154             Alan Hayward  <alan.hayward@arm.com>
30155             David Sherwood  <david.sherwood@arm.com>
30157         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
30158         (can_duplicate_and_interleave_p): New function.
30159         (vect_get_and_check_slp_defs): Take the vector of statements
30160         rather than just the current one.  Remove excess parentheses.
30161         Restriction rejectinon of vect_constant_def and vect_external_def
30162         for variable-length vectors to boolean types, or types for which
30163         can_duplicate_and_interleave_p is false.
30164         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
30165         (duplicate_and_interleave): New function.
30166         (vect_get_constant_vectors): Use gimple_build_vector for
30167         constant-length vectors and suitable variable-length constant
30168         vectors.  Use duplicate_and_interleave for other variable-length
30169         vectors.  Don't defer the update when inserting new statements.
30171 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30172             Alan Hayward  <alan.hayward@arm.com>
30173             David Sherwood  <david.sherwood@arm.com>
30175         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
30176         min_profitable_iters doesn't go negative.
30178 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30179             Alan Hayward  <alan.hayward@arm.com>
30180             David Sherwood  <david.sherwood@arm.com>
30182         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
30183         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
30184         * optabs.def (vec_mask_load_lanes_optab): New optab.
30185         (vec_mask_store_lanes_optab): Likewise.
30186         * internal-fn.def (MASK_LOAD_LANES): New internal function.
30187         (MASK_STORE_LANES): Likewise.
30188         * internal-fn.c (mask_load_lanes_direct): New macro.
30189         (mask_store_lanes_direct): Likewise.
30190         (expand_mask_load_optab_fn): Handle masked operations.
30191         (expand_mask_load_lanes_optab_fn): New macro.
30192         (expand_mask_store_optab_fn): Handle masked operations.
30193         (expand_mask_store_lanes_optab_fn): New macro.
30194         (direct_mask_load_lanes_optab_supported_p): Likewise.
30195         (direct_mask_store_lanes_optab_supported_p): Likewise.
30196         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
30197         parameter.
30198         (vect_load_lanes_supported): Likewise.
30199         * tree-vect-data-refs.c (strip_conversion): New function.
30200         (can_group_stmts_p): Likewise.
30201         (vect_analyze_data_ref_accesses): Use it instead of checking
30202         for a pair of assignments.
30203         (vect_store_lanes_supported): Take a masked_p parameter.
30204         (vect_load_lanes_supported): Likewise.
30205         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
30206         vect_store_lanes_supported and vect_load_lanes_supported.
30207         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
30208         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
30209         parameter.  Don't allow gaps for masked accesses.
30210         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
30211         and vect_load_lanes_supported.
30212         (get_load_store_type): Take a masked_p parameter and update
30213         call to get_group_load_store_type.
30214         (vectorizable_store): Update call to get_load_store_type.
30215         Handle IFN_MASK_STORE_LANES.
30216         (vectorizable_load): Update call to get_load_store_type.
30217         Handle IFN_MASK_LOAD_LANES.
30219 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30220             Alan Hayward  <alan.hayward@arm.com>
30221             David Sherwood  <david.sherwood@arm.com>
30223         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
30224         modes for SVE.
30225         * config/aarch64/aarch64-protos.h
30226         (aarch64_sve_struct_memory_operand_p): Declare.
30227         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
30228         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
30229         (VPRED, vpred): Handle SVE structure modes.
30230         * config/aarch64/constraints.md (Utx): New constraint.
30231         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
30232         (aarch64_sve_struct_nonimmediate_operand): New predicates.
30233         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
30234         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
30235         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
30236         structure modes.  Split into pieces after RA.
30237         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
30238         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
30239         New patterns.
30240         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
30241         SVE structure modes.
30242         (aarch64_classify_address): Likewise.
30243         (sizetochar): Move earlier in file.
30244         (aarch64_print_operand): Handle SVE register lists.
30245         (aarch64_array_mode): New function.
30246         (aarch64_sve_struct_memory_operand_p): Likewise.
30247         (TARGET_ARRAY_MODE): Redefine.
30249 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30250             Alan Hayward  <alan.hayward@arm.com>
30251             David Sherwood  <david.sherwood@arm.com>
30253         * target.def (array_mode): New target hook.
30254         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
30255         * doc/tm.texi: Regenerate.
30256         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
30257         * hooks.c (hook_optmode_mode_uhwi_none): New function.
30258         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
30259         targetm.array_mode.
30260         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
30261         type sizes.
30263 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30264             Alan Hayward  <alan.hayward@arm.com>
30265             David Sherwood  <david.sherwood@arm.com>
30267         * fold-const.c (fold_binary_loc): Check the argument types
30268         rather than the result type when testing for a vector operation.
30270 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30272         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
30273         * doc/tm.texi: Regenerate.
30275 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30276             Alan Hayward  <alan.hayward@arm.com>
30277             David Sherwood  <david.sherwood@arm.com>
30279         * doc/invoke.texi (-msve-vector-bits=): Document new option.
30280         (sve): Document new AArch64 extension.
30281         * doc/md.texi (w): Extend the description of the AArch64
30282         constraint to include SVE vectors.
30283         (Upl, Upa): Document new AArch64 predicate constraints.
30284         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
30285         enum.
30286         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
30287         (msve-vector-bits=): New option.
30288         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
30289         SVE when these are disabled.
30290         (sve): New extension.
30291         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
30292         modes.  Adjust their number of units based on aarch64_sve_vg.
30293         (MAX_BITSIZE_MODE_ANY_MODE): Define.
30294         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
30295         aarch64_addr_query_type.
30296         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
30297         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
30298         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
30299         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
30300         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
30301         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
30302         (aarch64_simd_imm_zero_p): Delete.
30303         (aarch64_check_zero_based_sve_index_immediate): Declare.
30304         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
30305         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
30306         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
30307         (aarch64_sve_float_mul_immediate_p): Likewise.
30308         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
30309         rather than an rtx.
30310         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
30311         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
30312         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
30313         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
30314         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
30315         (aarch64_regmode_natural_size): Likewise.
30316         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
30317         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
30318         left one place.
30319         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
30320         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
30321         for VG and the SVE predicate registers.
30322         (V_ALIASES): Add a "z"-prefixed alias.
30323         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
30324         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
30325         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
30326         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
30327         (REG_CLASS_NAMES): Add entries for them.
30328         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
30329         and the predicate registers.
30330         (aarch64_sve_vg): Declare.
30331         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
30332         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
30333         (REGMODE_NATURAL_SIZE): Define.
30334         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
30335         SVE macros.
30336         * config/aarch64/aarch64.c: Include cfgrtl.h.
30337         (simd_immediate_info): Add a constructor for series vectors,
30338         and an associated step field.
30339         (aarch64_sve_vg): New variable.
30340         (aarch64_dbx_register_number): Handle VG and the predicate registers.
30341         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
30342         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
30343         (VEC_ANY_DATA, VEC_STRUCT): New constants.
30344         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
30345         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
30346         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
30347         (aarch64_get_mask_mode): New functions.
30348         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
30349         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
30350         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
30351         predicate modes and predicate registers.  Explicitly restrict
30352         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
30353         to store a vector mode if it is recognized by
30354         aarch64_classify_vector_mode.
30355         (aarch64_regmode_natural_size): New function.
30356         (aarch64_hard_regno_caller_save_mode): Return the original mode
30357         for predicates.
30358         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
30359         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
30360         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
30361         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
30362         functions.
30363         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
30364         does not overlap dest if the function is frame-related.  Handle
30365         SVE constants.
30366         (aarch64_split_add_offset): New function.
30367         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
30368         them aarch64_add_offset.
30369         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
30370         and update call to aarch64_sub_sp.
30371         (aarch64_add_cfa_expression): New function.
30372         (aarch64_expand_prologue): Pass extra temporary registers to the
30373         functions above.  Handle the case in which we need to emit new
30374         DW_CFA_expressions for registers that were originally saved
30375         relative to the stack pointer, but now have to be expressed
30376         relative to the frame pointer.
30377         (aarch64_output_mi_thunk): Pass extra temporary registers to the
30378         functions above.
30379         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
30380         IP0 and IP1 values for SVE frames.
30381         (aarch64_expand_vec_series): New function.
30382         (aarch64_expand_sve_widened_duplicate): Likewise.
30383         (aarch64_expand_sve_const_vector): Likewise.
30384         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
30385         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
30386         into the register, rather than emitting a SET directly.
30387         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
30388         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
30389         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
30390         (offset_9bit_signed_scaled_p): New functions.
30391         (aarch64_replicate_bitmask_imm): New function.
30392         (aarch64_bitmask_imm): Use it.
30393         (aarch64_cannot_force_const_mem): Reject expressions involving
30394         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
30395         (aarch64_classify_index): Handle SVE indices, by requiring
30396         a plain register index with a scale that matches the element size.
30397         (aarch64_classify_address): Handle SVE addresses.  Assert that
30398         the mode of the address is VOIDmode or an integer mode.
30399         Update call to aarch64_classify_symbol.
30400         (aarch64_classify_symbolic_expression): Update call to
30401         aarch64_classify_symbol.
30402         (aarch64_const_vec_all_in_range_p): New function.
30403         (aarch64_print_vector_float_operand): Likewise.
30404         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
30405         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
30406         and the FP immediates 1.0 and 0.5.
30407         (aarch64_print_address_internal): Handle SVE addresses.
30408         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
30409         (aarch64_regno_regclass): Handle predicate registers.
30410         (aarch64_secondary_reload): Handle big-endian reloads of SVE
30411         data modes.
30412         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
30413         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
30414         (aarch64_convert_sve_vector_bits): New function.
30415         (aarch64_override_options): Use it to handle -msve-vector-bits=.
30416         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
30417         rather than an rtx.
30418         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
30419         Handle SVE vector and predicate modes.  Accept VL-based constants
30420         that need only one temporary register, and VL offsets that require
30421         no temporary registers.
30422         (aarch64_conditional_register_usage): Mark the predicate registers
30423         as fixed if SVE isn't available.
30424         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
30425         Return true for SVE vector and predicate modes.
30426         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
30427         rather than an unsigned int.  Handle SVE modes.
30428         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
30429         SVE modes.
30430         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
30431         if SVE is enabled.
30432         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
30433         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
30434         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
30435         (aarch64_sve_float_mul_immediate_p): New functions.
30436         (aarch64_sve_valid_immediate): New function.
30437         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
30438         Explicitly reject structure modes.  Check for INDEX constants.
30439         Handle PTRUE and PFALSE constants.
30440         (aarch64_check_zero_based_sve_index_immediate): New function.
30441         (aarch64_simd_imm_zero_p): Delete.
30442         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
30443         vector modes.  Accept constants in the range of CNT[BHWD].
30444         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
30445         ask for an Advanced SIMD mode.
30446         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
30447         (aarch64_simd_vector_alignment): Handle SVE predicates.
30448         (aarch64_vectorize_preferred_vector_alignment): New function.
30449         (aarch64_simd_vector_alignment_reachable): Use it instead of
30450         the vector size.
30451         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
30452         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
30453         functions.
30454         (MAX_VECT_LEN): Delete.
30455         (expand_vec_perm_d): Add a vec_flags field.
30456         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
30457         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
30458         (aarch64_evpc_ext): Don't apply a big-endian lane correction
30459         for SVE modes.
30460         (aarch64_evpc_rev): Rename to...
30461         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
30462         (aarch64_evpc_rev_global): New function.
30463         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
30464         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
30465         MAX_VECT_LEN.
30466         (aarch64_evpc_sve_tbl): New function.
30467         (aarch64_expand_vec_perm_const_1): Update after rename of
30468         aarch64_evpc_rev.  Handle SVE permutes too, trying
30469         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
30470         than aarch64_evpc_tbl.
30471         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
30472         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
30473         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
30474         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
30475         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
30476         (aarch64_expand_sve_vcond): New functions.
30477         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
30478         of aarch64_vector_mode_p.
30479         (aarch64_dwarf_poly_indeterminate_value): New function.
30480         (aarch64_compute_pressure_classes): Likewise.
30481         (aarch64_can_change_mode_class): Likewise.
30482         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
30483         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
30484         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
30485         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
30486         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
30487         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
30488         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
30489         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
30490         constraints.
30491         (Dn, Dl, Dr): Accept const as well as const_vector.
30492         (Dz): Likewise.  Compare against CONST0_RTX.
30493         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
30494         of "vector" where appropriate.
30495         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
30496         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
30497         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
30498         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
30499         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
30500         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
30501         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
30502         (v_int_equiv): Extend to SVE modes.
30503         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
30504         mode attributes.
30505         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
30506         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
30507         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
30508         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
30509         (SVE_COND_FP_CMP): New int iterators.
30510         (perm_hilo): Handle the new unpack unspecs.
30511         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
30512         attributes.
30513         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
30514         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
30515         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
30516         (aarch64_equality_operator, aarch64_constant_vector_operand)
30517         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
30518         (aarch64_sve_nonimmediate_operand): Likewise.
30519         (aarch64_sve_general_operand): Likewise.
30520         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
30521         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
30522         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
30523         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
30524         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
30525         (aarch64_sve_float_arith_immediate): Likewise.
30526         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
30527         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
30528         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
30529         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
30530         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
30531         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
30532         (aarch64_sve_float_arith_operand): Likewise.
30533         (aarch64_sve_float_arith_with_sub_operand): Likewise.
30534         (aarch64_sve_float_mul_operand): Likewise.
30535         (aarch64_sve_vec_perm_operand): Likewise.
30536         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
30537         (aarch64_mov_operand): Accept const_poly_int and const_vector.
30538         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
30539         as well as const_vector.
30540         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
30541         in file.  Use CONST0_RTX and CONSTM1_RTX.
30542         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
30543         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
30544         Use aarch64_simd_imm_zero.
30545         * config/aarch64/aarch64-sve.md: New file.
30546         * config/aarch64/aarch64.md: Include it.
30547         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
30548         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
30549         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
30550         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
30551         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
30552         (sve): New attribute.
30553         (enabled): Disable instructions with the sve attribute unless
30554         TARGET_SVE.
30555         (movqi, movhi): Pass CONST_POLY_INT operaneds through
30556         aarch64_expand_mov_immediate.
30557         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
30558         CNT[BHSD] immediates.
30559         (movti): Split CONST_POLY_INT moves into two halves.
30560         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
30561         Split additions that need a temporary here if the destination
30562         is the stack pointer.
30563         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
30564         (*add<mode>3_poly_1): New instruction.
30565         (set_clobber_cc): New expander.
30567 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30569         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
30570         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
30571         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
30572         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
30573         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
30574         Change innermode from fixed_mode_size to machine_mode.
30575         (simplify_subreg): Update call accordingly.  Handle a constant-sized
30576         subreg of a variable-length CONST_VECTOR.
30578 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
30579             Alan Hayward  <alan.hayward@arm.com>
30580             David Sherwood  <david.sherwood@arm.com>
30582         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
30583         (add_offset_to_base): New function, split out from...
30584         (create_mem_ref): ...here.  When handling a scale other than 1,
30585         check first whether the address is valid without the offset.
30586         Add it into the base if so, leaving the index and scale as-is.
30588 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
30590         PR c++/83778
30591         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
30592         fold_for_warn before checking if arg2 is INTEGER_CST.
30594 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
30596         * config/rs6000/predicates.md (load_multiple_operation): Delete.
30597         (store_multiple_operation): Delete.
30598         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
30599         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
30600         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
30601         guarded by TARGET_STRING.
30602         (rs6000_output_load_multiple): Delete.
30603         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30604         OPTION_MASK_STRING / TARGET_STRING handling.
30605         (print_operand) <'N', 'O'>: Add comment that these are unused now.
30606         (const rs6000_opt_masks) <"string">: Change mask to 0.
30607         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
30608         (MASK_STRING): Delete.
30609         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
30610         parts.  Simplify.
30611         (load_multiple): Delete.
30612         (*ldmsi8): Delete.
30613         (*ldmsi7): Delete.
30614         (*ldmsi6): Delete.
30615         (*ldmsi5): Delete.
30616         (*ldmsi4): Delete.
30617         (*ldmsi3): Delete.
30618         (store_multiple): Delete.
30619         (*stmsi8): Delete.
30620         (*stmsi7): Delete.
30621         (*stmsi6): Delete.
30622         (*stmsi5): Delete.
30623         (*stmsi4): Delete.
30624         (*stmsi3): Delete.
30625         (movmemsi_8reg): Delete.
30626         (corresponding unnamed define_insn): Delete.
30627         (movmemsi_6reg): Delete.
30628         (corresponding unnamed define_insn): Delete.
30629         (movmemsi_4reg): Delete.
30630         (corresponding unnamed define_insn): Delete.
30631         (movmemsi_2reg): Delete.
30632         (corresponding unnamed define_insn): Delete.
30633         (movmemsi_1reg): Delete.
30634         (corresponding unnamed define_insn): Delete.
30635         * config/rs6000/rs6000.opt (mno-string): New.
30636         (mstring): Replace by deprecation warning stub.
30637         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
30639 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
30641         * regrename.c (regrename_do_replace): If replacing the same
30642         reg multiple times, try to reuse last created gen_raw_REG.
30644         PR debug/81155
30645         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
30646         main to workaround a bug in GDB.
30648 2018-01-12  Tom de Vries  <tom@codesourcery.com>
30650         PR target/83737
30651         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
30653 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
30655         PR rtl-optimization/80481
30656         * ira-color.c (get_cap_member): New function.
30657         (allocnos_conflict_by_live_ranges_p): Use it.
30658         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
30659         (setup_slot_coalesced_allocno_live_ranges): Ditto.
30661 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
30663         PR target/83628
30664         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
30665         (*saddl_se_1): Ditto.
30666         (*ssubsi_1): Ditto.
30667         (*ssubl_se_1): Ditto.
30669 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
30671         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
30672         rather than wi::to_widest for DR_INITs.
30673         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
30674         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
30675         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
30676         INTEGER_CSTs.
30677         (vect_analyze_group_access_1): Note that here.
30679 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
30681         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
30682         polynomial type sizes.
30684 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
30686         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
30687         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
30688         (gimple_add_tmp_var): Likewise.
30690 2018-01-12  Martin Liska  <mliska@suse.cz>
30692         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
30693         (gimple_alloc_sizes): Likewise.
30694         (dump_gimple_statistics): Use PRIu64 in printf format.
30695         * gimple.h: Change uint64_t to int.
30697 2018-01-12  Martin Liska  <mliska@suse.cz>
30699         * tree-core.h: Use uint64_t instead of int.
30700         * tree.c (tree_node_counts): Likewise.
30701         (tree_node_sizes): Likewise.
30702         (dump_tree_statistics): Use PRIu64 in printf format.
30704 2018-01-12  Martin Liska  <mliska@suse.cz>
30706         * Makefile.in: As qsort_chk is implemented in vec.c, add
30707         vec.o to linkage of gencfn-macros.
30708         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
30709         passing the info to record_node_allocation_statistics.
30710         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
30711         and pass the info.
30712         * ggc-common.c (struct ggc_usage): Add operator== and use
30713         it in operator< and compare function.
30714         * mem-stats.h (struct mem_usage): Likewise.
30715         * vec.c (struct vec_usage): Remove operator< and compare
30716         function. Can be simply inherited.
30718 2018-01-12  Martin Jambor  <mjambor@suse.cz>
30720         PR target/81616
30721         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
30722         * tree-ssa-math-opts.c: Include domwalk.h.
30723         (convert_mult_to_fma_1): New function.
30724         (fma_transformation_info): New type.
30725         (fma_deferring_state): Likewise.
30726         (cancel_fma_deferring): New function.
30727         (result_of_phi): Likewise.
30728         (last_fma_candidate_feeds_initial_phi): Likewise.
30729         (convert_mult_to_fma): Added deferring logic, split actual
30730         transformation to convert_mult_to_fma_1.
30731         (math_opts_dom_walker): New type.
30732         (math_opts_dom_walker::after_dom_children): New method, body moved
30733         here from pass_optimize_widening_mul::execute, added deferring logic
30734         bits.
30735         (pass_optimize_widening_mul::execute): Moved most of code to
30736         math_opts_dom_walker::after_dom_children.
30737         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
30738         * config/i386/i386.c (ix86_option_override_internal): Added
30739         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
30741 2018-01-12  Richard Biener  <rguenther@suse.de>
30743         PR debug/83157
30744         * dwarf2out.c (gen_variable_die): Do not reset old_die for
30745         inline instance vars.
30747 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
30749         PR target/81819
30750         * config/rx/rx.c (rx_is_restricted_memory_address):
30751         Handle SUBREG case.
30753 2018-01-12  Richard Biener  <rguenther@suse.de>
30755         PR tree-optimization/80846
30756         * target.def (split_reduction): New target hook.
30757         * targhooks.c (default_split_reduction): New function.
30758         * targhooks.h (default_split_reduction): Declare.
30759         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
30760         target requests first reduce vectors by combining low and high
30761         parts.
30762         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
30763         (get_vectype_for_scalar_type_and_size): Export.
30764         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
30765         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
30766         * doc/tm.texi: Regenerate.
30767         * config/i386/i386.c (ix86_split_reduction): Implement
30768         TARGET_VECTORIZE_SPLIT_REDUCTION.
30770 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
30772         PR target/83368
30773         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
30774         in PIC mode except for TARGET_VXWORKS_RTP.
30775         * config/sparc/sparc.c: Include cfgrtl.h.
30776         (TARGET_INIT_PIC_REG): Define.
30777         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
30778         (sparc_pic_register_p): New predicate.
30779         (sparc_legitimate_address_p): Use it.
30780         (sparc_legitimize_pic_address): Likewise.
30781         (sparc_delegitimize_address): Likewise.
30782         (sparc_mode_dependent_address_p): Likewise.
30783         (gen_load_pcrel_sym): Remove 4th parameter.
30784         (load_got_register): Adjust call to above.  Remove obsolete stuff.
30785         (sparc_expand_prologue): Do not call load_got_register here.
30786         (sparc_flat_expand_prologue): Likewise.
30787         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
30788         (sparc_use_pseudo_pic_reg): New function.
30789         (sparc_init_pic_reg): Likewise.
30790         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
30791         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
30793 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
30795         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
30796         Add item for branch_cost.
30798 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
30800         PR rtl-optimization/83565
30801         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
30802         not extend the result to a larger mode for rotate operations.
30803         (num_sign_bit_copies1): Likewise.
30805 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30807         PR target/40411
30808         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
30809         -symbolic.
30810         Use values-Xc.o for -pedantic.
30811         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
30813 2018-01-12  Martin Liska  <mliska@suse.cz>
30815         PR ipa/83054
30816         * ipa-devirt.c (final_warning_record::grow_type_warnings):
30817         New function.
30818         (possible_polymorphic_call_targets): Use it.
30819         (ipa_devirt): Likewise.
30821 2018-01-12  Martin Liska  <mliska@suse.cz>
30823         * profile-count.h (enum profile_quality): Use 0 as invalid
30824         enum value of profile_quality.
30826 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
30828         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
30829         -mext-string options.
30831 2018-01-12  Richard Biener  <rguenther@suse.de>
30833         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
30834         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
30835         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
30836         Likewise.
30837         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
30839 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
30841         * configure.ac (--with-long-double-format): Add support for the
30842         configuration option to change the default long double format on
30843         PowerPC systems.
30844         * config.gcc (powerpc*-linux*-*): Likewise.
30845         * configure: Regenerate.
30846         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
30847         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
30848         used without modification.
30850 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30852         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
30853         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
30854         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
30855         MISC_BUILTIN_SPEC_BARRIER.
30856         (rs6000_init_builtins): Likewise.
30857         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
30858         enum value.
30859         (speculation_barrier): New define_insn.
30860         * doc/extend.texi: Document __builtin_speculation_barrier.
30862 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
30864         PR target/83203
30865         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
30866         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
30867         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
30868         iterators.
30869         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
30870         integral modes instead of "ss" and "sd".
30871         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
30872         vectors with 32-bit and 64-bit elements.
30873         (vecdupssescalarmodesuffix): New mode attribute.
30874         (vec_dup<mode>): Use it.
30876 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
30878         PR target/83330
30879         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
30880         frame if argument is passed on stack.
30882 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
30884         PR target/82682
30885         * ree.c (combine_reaching_defs): Optimize also
30886         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
30887         reg2=any_extend(exp); reg1=reg2;, formatting fix.
30889 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
30891         PR middle-end/83189
30892         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
30894 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
30896         PR middle-end/83718
30897         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
30898         after they are computed.
30900 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
30902         PR tree-optimization/83695
30903         * gimple-loop-linterchange.cc
30904         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
30905         reset cached scev information after interchange.
30906         (pass_linterchange::execute): Remove call to scev_reset_htab.
30908 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30910         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
30911         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
30912         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
30913         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
30914         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
30915         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
30916         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
30917         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
30918         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
30919         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
30920         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
30921         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
30922         (V_lane_reg): Likewise.
30923         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
30924         New define_expand.
30925         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
30926         (vfmal_lane_low<mode>_intrinsic,
30927         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
30928         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
30929         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
30930         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
30931         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
30932         vfmsl_lane_high<mode>_intrinsic): New define_insns.
30934 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30936         * config/arm/arm-cpus.in (fp16fml): New feature.
30937         (ALL_SIMD): Add fp16fml.
30938         (armv8.2-a): Add fp16fml as an option.
30939         (armv8.3-a): Likewise.
30940         (armv8.4-a): Add fp16fml as part of fp16.
30941         * config/arm/arm.h (TARGET_FP16FML): Define.
30942         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
30943         when appropriate.
30944         * config/arm/arm-modes.def (V2HF): Define.
30945         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
30946         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
30947         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
30948         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
30949         vfmsl_low, vfmsl_high): New set of builtins.
30950         * config/arm/iterators.md (PLUSMINUS): New code iterator.
30951         (vfml_op): New code attribute.
30952         (VFMLHALVES): New int iterator.
30953         (VFML, VFMLSEL): New mode attributes.
30954         (V_reg): Define mapping for V2HF.
30955         (V_hi, V_lo): New mode attributes.
30956         (VF_constraint): Likewise.
30957         (vfml_half, vfml_half_selector): New int attributes.
30958         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
30959         define_expand.
30960         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
30961         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
30962         New define_insn.
30963         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
30964         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
30965         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
30966         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
30967         documentation.
30968         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
30969         Document new effective target and option set.
30971 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30973         * config/arm/arm-cpus.in (armv8_4): New feature.
30974         (ARMv8_4a): New fgroup.
30975         (armv8.4-a): New arch.
30976         * config/arm/arm-tables.opt: Regenerate.
30977         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
30978         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
30979         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
30980         Add matching rules for -march=armv8.4-a and extensions.
30981         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
30983 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
30985         PR target/81821
30986         * config/rx/rx.md (BW): New mode attribute.
30987         (sync_lock_test_and_setsi): Add mode suffix to insn output.
30989 2018-01-11  Richard Biener  <rguenther@suse.de>
30991         PR tree-optimization/83435
30992         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
30993         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
30994         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
30996 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
30997             Alan Hayward  <alan.hayward@arm.com>
30998             David Sherwood  <david.sherwood@arm.com>
31000         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
31001         field.
31002         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
31003         (aarch64_print_address_internal): Use it to check for a zero offset.
31005 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
31006             Alan Hayward  <alan.hayward@arm.com>
31007             David Sherwood  <david.sherwood@arm.com>
31009         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
31010         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
31011         Return a poly_int64 rather than a HOST_WIDE_INT.
31012         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
31013         rather than a HOST_WIDE_INT.
31014         * config/aarch64/aarch64.h (aarch64_frame): Protect with
31015         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
31016         hard_fp_offset, frame_size, initial_adjust, callee_offset and
31017         final_offset from HOST_WIDE_INT to poly_int64.
31018         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
31019         to_constant when getting the number of units in an Advanced SIMD
31020         mode.
31021         (aarch64_builtin_vectorized_function): Check for a constant number
31022         of units.
31023         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
31024         GET_MODE_SIZE.
31025         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
31026         attribute instead of GET_MODE_NUNITS.
31027         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
31028         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
31029         GET_MODE_SIZE for fixed-size registers.
31030         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
31031         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
31032         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
31033         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
31034         (aarch64_print_operand, aarch64_print_address_internal)
31035         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
31036         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
31037         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
31038         Handle polynomial GET_MODE_SIZE.
31039         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
31040         wider than SImode without modification.
31041         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
31042         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
31043         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
31044         passing and returning SVE modes.
31045         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
31046         rather than GEN_INT.
31047         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
31048         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
31049         (aarch64_allocate_and_probe_stack_space): Likewise.
31050         (aarch64_layout_frame): Cope with polynomial offsets.
31051         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
31052         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
31053         polynomial offsets.
31054         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
31055         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
31056         poly_int64 rather than a HOST_WIDE_INT.
31057         (aarch64_get_separate_components, aarch64_process_components)
31058         (aarch64_expand_prologue, aarch64_expand_epilogue)
31059         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
31060         (aarch64_anchor_offset): New function, split out from...
31061         (aarch64_legitimize_address): ...here.
31062         (aarch64_builtin_vectorization_cost): Handle polynomial
31063         TYPE_VECTOR_SUBPARTS.
31064         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
31065         GET_MODE_NUNITS.
31066         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
31067         number of elements from the PARALLEL rather than the mode.
31068         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
31069         rather than GET_MODE_BITSIZE.
31070         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
31071         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
31072         (aarch64_expand_vec_perm_const_1): Handle polynomial
31073         d->perm.length () and d->perm elements.
31074         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
31075         Apply to_constant to d->perm elements.
31076         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
31077         polynomial CONST_VECTOR_NUNITS.
31078         (aarch64_move_pointer): Take amount as a poly_int64 rather
31079         than an int.
31080         (aarch64_progress_pointer): Avoid temporary variable.
31081         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
31082         the mode attribute instead of GET_MODE.
31084 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
31085             Alan Hayward  <alan.hayward@arm.com>
31086             David Sherwood  <david.sherwood@arm.com>
31088         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
31089         x exists before using it.
31090         (aarch64_add_constant_internal): Rename to...
31091         (aarch64_add_offset_1): ...this.  Replace regnum with separate
31092         src and dest rtxes.  Handle the case in which they're different,
31093         including when the offset is zero.  Replace scratchreg with an rtx.
31094         Use 2 additions if there is no spare register into which we can
31095         move a 16-bit constant.
31096         (aarch64_add_constant): Delete.
31097         (aarch64_add_offset): Replace reg with separate src and dest
31098         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
31099         Use aarch64_add_offset_1.
31100         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
31101         an rtx rather than an int.  Take the delta as a poly_int64
31102         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
31103         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
31104         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
31105         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
31106         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
31107         and aarch64_add_sp.
31108         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
31109         aarch64_add_constant.
31111 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
31113         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
31114         Use scalar_float_mode.
31116 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
31118         * config/aarch64/aarch64-simd.md
31119         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
31120         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
31121         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
31122         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
31123         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
31124         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
31125         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
31126         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
31127         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
31128         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
31130 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
31132         PR target/83514
31133         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
31134         targ_options->x_arm_arch_string is non NULL.
31136 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
31138         * config/aarch64/aarch64.h
31139         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
31141 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
31143         PR target/82096
31144         * expmed.c (emit_store_flag_force): Swap if const op0
31145         and change VOIDmode to mode of op0.
31147 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
31149         PR rtl-optimization/83761
31150         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
31151         than bytes to mode_for_size.
31153 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
31155         PR middle-end/83189
31156         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
31157         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
31158         profile.
31160 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
31162         PR middle-end/83575
31163         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
31164         when in layout mode.
31165         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
31166         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
31167         partition fixup.
31169 2018-01-10  Michael Collison  <michael.collison@arm.com>
31171         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
31172         * config/aarch64/aarch64-option-extension.def: Add
31173         AARCH64_OPT_EXTENSION of 'fp16fml'.
31174         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
31175         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
31176         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
31177         * config/aarch64/constraints.md (Ui7): New constraint.
31178         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
31179         (VFMLA_SEL_W): Ditto.
31180         (f16quad): Ditto.
31181         (f16mac1): Ditto.
31182         (VFMLA16_LOW): New int iterator.
31183         (VFMLA16_HIGH): Ditto.
31184         (UNSPEC_FMLAL): New unspec.
31185         (UNSPEC_FMLSL): Ditto.
31186         (UNSPEC_FMLAL2): Ditto.
31187         (UNSPEC_FMLSL2): Ditto.
31188         (f16mac): New code attribute.
31189         * config/aarch64/aarch64-simd-builtins.def
31190         (aarch64_fmlal_lowv2sf): Ditto.
31191         (aarch64_fmlsl_lowv2sf): Ditto.
31192         (aarch64_fmlalq_lowv4sf): Ditto.
31193         (aarch64_fmlslq_lowv4sf): Ditto.
31194         (aarch64_fmlal_highv2sf): Ditto.
31195         (aarch64_fmlsl_highv2sf): Ditto.
31196         (aarch64_fmlalq_highv4sf): Ditto.
31197         (aarch64_fmlslq_highv4sf): Ditto.
31198         (aarch64_fmlal_lane_lowv2sf): Ditto.
31199         (aarch64_fmlsl_lane_lowv2sf): Ditto.
31200         (aarch64_fmlal_laneq_lowv2sf): Ditto.
31201         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
31202         (aarch64_fmlalq_lane_lowv4sf): Ditto.
31203         (aarch64_fmlsl_lane_lowv4sf): Ditto.
31204         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
31205         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
31206         (aarch64_fmlal_lane_highv2sf): Ditto.
31207         (aarch64_fmlsl_lane_highv2sf): Ditto.
31208         (aarch64_fmlal_laneq_highv2sf): Ditto.
31209         (aarch64_fmlsl_laneq_highv2sf): Ditto.
31210         (aarch64_fmlalq_lane_highv4sf): Ditto.
31211         (aarch64_fmlsl_lane_highv4sf): Ditto.
31212         (aarch64_fmlalq_laneq_highv4sf): Ditto.
31213         (aarch64_fmlsl_laneq_highv4sf): Ditto.
31214         * config/aarch64/aarch64-simd.md:
31215         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
31216         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
31217         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
31218         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
31219         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
31220         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
31221         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
31222         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
31223         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
31224         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
31225         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
31226         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
31227         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
31228         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
31229         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
31230         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
31231         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
31232         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
31233         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
31234         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
31235         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
31236         (vfmlsl_low_u32): Ditto.
31237         (vfmlalq_low_u32): Ditto.
31238         (vfmlslq_low_u32): Ditto.
31239         (vfmlal_high_u32): Ditto.
31240         (vfmlsl_high_u32): Ditto.
31241         (vfmlalq_high_u32): Ditto.
31242         (vfmlslq_high_u32): Ditto.
31243         (vfmlal_lane_low_u32): Ditto.
31244         (vfmlsl_lane_low_u32): Ditto.
31245         (vfmlal_laneq_low_u32): Ditto.
31246         (vfmlsl_laneq_low_u32): Ditto.
31247         (vfmlalq_lane_low_u32): Ditto.
31248         (vfmlslq_lane_low_u32): Ditto.
31249         (vfmlalq_laneq_low_u32): Ditto.
31250         (vfmlslq_laneq_low_u32): Ditto.
31251         (vfmlal_lane_high_u32): Ditto.
31252         (vfmlsl_lane_high_u32): Ditto.
31253         (vfmlal_laneq_high_u32): Ditto.
31254         (vfmlsl_laneq_high_u32): Ditto.
31255         (vfmlalq_lane_high_u32): Ditto.
31256         (vfmlslq_lane_high_u32): Ditto.
31257         (vfmlalq_laneq_high_u32): Ditto.
31258         (vfmlslq_laneq_high_u32): Ditto.
31259         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
31260         (AARCH64_FL_FOR_ARCH8_4): New.
31261         (AARCH64_ISA_F16FML): New ISA flag.
31262         (TARGET_F16FML): New feature flag for fp16fml.
31263         (doc/invoke.texi): Document new fp16fml option.
31265 2018-01-10  Michael Collison  <michael.collison@arm.com>
31267         * config/aarch64/aarch64-builtins.c:
31268         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
31269         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
31270         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
31271         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
31272         (AARCH64_ISA_SHA3): New ISA flag.
31273         (TARGET_SHA3): New feature flag for sha3.
31274         * config/aarch64/iterators.md (sha512_op): New int attribute.
31275         (CRYPTO_SHA512): New int iterator.
31276         (UNSPEC_SHA512H): New unspec.
31277         (UNSPEC_SHA512H2): Ditto.
31278         (UNSPEC_SHA512SU0): Ditto.
31279         (UNSPEC_SHA512SU1): Ditto.
31280         * config/aarch64/aarch64-simd-builtins.def
31281         (aarch64_crypto_sha512hqv2di): New builtin.
31282         (aarch64_crypto_sha512h2qv2di): Ditto.
31283         (aarch64_crypto_sha512su0qv2di): Ditto.
31284         (aarch64_crypto_sha512su1qv2di): Ditto.
31285         (aarch64_eor3qv8hi): Ditto.
31286         (aarch64_rax1qv2di): Ditto.
31287         (aarch64_xarqv2di): Ditto.
31288         (aarch64_bcaxqv8hi): Ditto.
31289         * config/aarch64/aarch64-simd.md:
31290         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
31291         (aarch64_crypto_sha512su0qv2di): Ditto.
31292         (aarch64_crypto_sha512su1qv2di): Ditto.
31293         (aarch64_eor3qv8hi): Ditto.
31294         (aarch64_rax1qv2di): Ditto.
31295         (aarch64_xarqv2di): Ditto.
31296         (aarch64_bcaxqv8hi): Ditto.
31297         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
31298         (vsha512h2q_u64): Ditto.
31299         (vsha512su0q_u64): Ditto.
31300         (vsha512su1q_u64): Ditto.
31301         (veor3q_u16): Ditto.
31302         (vrax1q_u64): Ditto.
31303         (vxarq_u64): Ditto.
31304         (vbcaxq_u16): Ditto.
31305         * config/arm/types.md (crypto_sha512): New type attribute.
31306         (crypto_sha3): Ditto.
31307         (doc/invoke.texi): Document new sha3 option.
31309 2018-01-10  Michael Collison  <michael.collison@arm.com>
31311         * config/aarch64/aarch64-builtins.c:
31312         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
31313         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
31314         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
31315         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
31316         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
31317         (AARCH64_ISA_SM4): New ISA flag.
31318         (TARGET_SM4): New feature flag for sm4.
31319         * config/aarch64/aarch64-simd-builtins.def
31320         (aarch64_sm3ss1qv4si): Ditto.
31321         (aarch64_sm3tt1aq4si): Ditto.
31322         (aarch64_sm3tt1bq4si): Ditto.
31323         (aarch64_sm3tt2aq4si): Ditto.
31324         (aarch64_sm3tt2bq4si): Ditto.
31325         (aarch64_sm3partw1qv4si): Ditto.
31326         (aarch64_sm3partw2qv4si): Ditto.
31327         (aarch64_sm4eqv4si): Ditto.
31328         (aarch64_sm4ekeyqv4si): Ditto.
31329         * config/aarch64/aarch64-simd.md:
31330         (aarch64_sm3ss1qv4si): Ditto.
31331         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
31332         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
31333         (aarch64_sm4eqv4si): Ditto.
31334         (aarch64_sm4ekeyqv4si): Ditto.
31335         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
31336         (sm3part_op): Ditto.
31337         (CRYPTO_SM3TT): Ditto.
31338         (CRYPTO_SM3PART): Ditto.
31339         (UNSPEC_SM3SS1): New unspec.
31340         (UNSPEC_SM3TT1A): Ditto.
31341         (UNSPEC_SM3TT1B): Ditto.
31342         (UNSPEC_SM3TT2A): Ditto.
31343         (UNSPEC_SM3TT2B): Ditto.
31344         (UNSPEC_SM3PARTW1): Ditto.
31345         (UNSPEC_SM3PARTW2): Ditto.
31346         (UNSPEC_SM4E): Ditto.
31347         (UNSPEC_SM4EKEY): Ditto.
31348         * config/aarch64/constraints.md (Ui2): New constraint.
31349         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
31350         * config/arm/types.md (crypto_sm3): New type attribute.
31351         (crypto_sm4): Ditto.
31352         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
31353         (vsm3tt1aq_u32): Ditto.
31354         (vsm3tt1bq_u32): Ditto.
31355         (vsm3tt2aq_u32): Ditto.
31356         (vsm3tt2bq_u32): Ditto.
31357         (vsm3partw1q_u32): Ditto.
31358         (vsm3partw2q_u32): Ditto.
31359         (vsm4eq_u32): Ditto.
31360         (vsm4ekeyq_u32): Ditto.
31361         (doc/invoke.texi): Document new sm4 option.
31363 2018-01-10  Michael Collison  <michael.collison@arm.com>
31365         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
31366         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
31367         (AARCH64_FL_FOR_ARCH8_4): New.
31368         (AARCH64_FL_V8_4): New flag.
31369         (doc/invoke.texi): Document new armv8.4-a option.
31371 2018-01-10  Michael Collison  <michael.collison@arm.com>
31373         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
31374         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
31375         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
31376         * config/aarch64/aarch64-option-extension.def: Add
31377         AARCH64_OPT_EXTENSION of 'sha2'.
31378         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
31379         (crypto): Disable sha2 and aes if crypto disabled.
31380         (crypto): Enable aes and sha2 if enabled.
31381         (simd): Disable sha2 and aes if simd disabled.
31382         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
31383         New flags.
31384         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
31385         (TARGET_SHA2): New feature flag for sha2.
31386         (TARGET_AES): New feature flag for aes.
31387         * config/aarch64/aarch64-simd.md:
31388         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
31389         conditional on TARGET_AES.
31390         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
31391         (aarch64_crypto_sha1hsi): Make pattern conditional
31392         on TARGET_SHA2.
31393         (aarch64_crypto_sha1hv4si): Ditto.
31394         (aarch64_be_crypto_sha1hv4si): Ditto.
31395         (aarch64_crypto_sha1su1v4si): Ditto.
31396         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
31397         (aarch64_crypto_sha1su0v4si): Ditto.
31398         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
31399         (aarch64_crypto_sha256su0v4si): Ditto.
31400         (aarch64_crypto_sha256su1v4si): Ditto.
31401         (doc/invoke.texi): Document new aes and sha2 options.
31403 2018-01-10  Martin Sebor  <msebor@redhat.com>
31405         PR tree-optimization/83781
31406         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
31407         as string arrays.
31409 2018-01-11  Martin Sebor  <msebor@gmail.com>
31410             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
31412         PR tree-optimization/83501
31413         PR tree-optimization/81703
31415         * tree-ssa-strlen.c (get_string_cst): Rename...
31416         (get_string_len): ...to this.  Handle global constants.
31417         (handle_char_store): Adjust.
31419 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
31420             Jim Wilson  <jimw@sifive.com>
31422         * config/riscv/riscv-protos.h (riscv_output_return): New.
31423         * config/riscv/riscv.c (struct machine_function): New naked_p field.
31424         (riscv_attribute_table, riscv_output_return),
31425         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
31426         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
31427         (riscv_compute_frame_info): Only compute frame->mask if not a naked
31428         function.
31429         (riscv_expand_prologue): Add early return for naked function.
31430         (riscv_expand_epilogue): Likewise.
31431         (riscv_function_ok_for_sibcall): Return false for naked function.
31432         (riscv_set_current_function): New.
31433         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
31434         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
31435         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
31436         * doc/extend.texi (RISC-V Function Attributes): New.
31438 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
31440         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
31441         check for 128-bit long double before checking TCmode.
31442         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
31443         128-bit long doubles before checking TFmode or TCmode.
31444         (FLOAT128_IBM_P): Likewise.
31446 2018-01-10  Martin Sebor  <msebor@redhat.com>
31448         PR tree-optimization/83671
31449         * builtins.c (c_strlen): Unconditionally return zero for the empty
31450         string.
31451         Use -Warray-bounds for warnings.
31452         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
31453         for non-constant array indices with COMPONENT_REF, arrays of
31454         arrays, and pointers to arrays.
31455         (gimple_fold_builtin_strlen): Determine and set length range for
31456         non-constant character arrays.
31458 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
31460         PR middle-end/81897
31461         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
31462         empty blocks.
31464 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
31466         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
31468 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
31470         PR target/83399
31471         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
31472         VECTOR_MEM_ALTIVEC_OR_VSX_P.
31473         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
31474         indexed_or_indirect_operand predicate.
31475         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
31476         (*vsx_le_perm_load_v8hi): Likewise.
31477         (*vsx_le_perm_load_v16qi): Likewise.
31478         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
31479         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
31480         (*vsx_le_perm_store_v8hi): Likewise.
31481         (*vsx_le_perm_store_v16qi): Likewise.
31482         (eight unnamed splitters): Likewise.
31484 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
31486         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
31487         * config/rs6000/emmintrin.h: Likewise.
31488         * config/rs6000/mmintrin.h: Likewise.
31489         * config/rs6000/xmmintrin.h: Likewise.
31491 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
31493         PR c++/43486
31494         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
31495         "public_flag".
31496         * tree.c (tree_nop_conversion): Return true for location wrapper
31497         nodes.
31498         (maybe_wrap_with_location): New function.
31499         (selftest::check_strip_nops): New function.
31500         (selftest::test_location_wrappers): New function.
31501         (selftest::tree_c_tests): Call it.
31502         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
31503         (maybe_wrap_with_location): New decl.
31504         (EXPR_LOCATION_WRAPPER_P): New macro.
31505         (location_wrapper_p): New inline function.
31506         (tree_strip_any_location_wrapper): New inline function.
31508 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
31510         PR target/83735
31511         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
31512         stack_realign_offset for the largest alignment of stack slot
31513         actually used.
31514         (ix86_find_max_used_stack_alignment): New function.
31515         (ix86_finalize_stack_frame_flags): Use it.  Set
31516         max_used_stack_alignment if we don't realign stack.
31517         * config/i386/i386.h (machine_function): Add
31518         max_used_stack_alignment.
31520 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
31522         * config/arm/arm.opt (-mbranch-cost): New option.
31523         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
31524         account.
31526 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
31528         PR target/83629
31529         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
31530         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
31532 2018-01-10  Richard Biener  <rguenther@suse.de>
31534         PR debug/83765
31535         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
31536         early out so it also covers the case where we have a non-NULL
31537         origin.
31539 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
31541         PR tree-optimization/83753
31542         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
31543         for non-strided grouped accesses if the number of elements is 1.
31545 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
31547         PR target/81616
31548         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
31549         * i386.h (TARGET_USE_GATHER): Define.
31550         * x86-tune.def (X86_TUNE_USE_GATHER): New.
31552 2018-01-10  Martin Liska  <mliska@suse.cz>
31554         PR bootstrap/82831
31555         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
31556         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
31557         partitioning.
31558         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
31559         CLEANUP_NO_PARTITIONING is not set.
31561 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
31563         * doc/rtl.texi: Remove documentation of (const ...) wrappers
31564         for vectors, as a partial revert of r254296.
31565         * rtl.h (const_vec_p): Delete.
31566         (const_vec_duplicate_p): Don't test for vector CONSTs.
31567         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
31568         * expmed.c (make_tree): Likewise.
31570         Revert:
31571         * common.md (E, F): Use CONSTANT_P instead of checking for
31572         CONST_VECTOR.
31573         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
31574         checking for CONST_VECTOR.
31576 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
31578         PR middle-end/83575
31579         * predict.c (force_edge_cold): Handle in more sane way edges
31580         with no prediction.
31582 2018-01-09  Carl Love  <cel@us.ibm.com>
31584         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
31585         V4SI, V4SF types.
31586         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
31587         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
31588         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
31589         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
31590         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
31591         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
31592         * config/rs6000/rs6000-protos.h: Add extern defition for
31593         rs6000_generate_float2_double_code.
31594         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
31595         function.
31596         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
31597         (float2_v2df): Add define_expand.
31599 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
31601         PR target/83628
31602         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
31603         op_mode in the force_to_mode call.
31605 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
31607         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
31608         instead of checking each element individually.
31609         (aarch64_evpc_uzp): Likewise.
31610         (aarch64_evpc_zip): Likewise.
31611         (aarch64_evpc_ext): Likewise.
31612         (aarch64_evpc_rev): Likewise.
31613         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
31614         instead of checking each element individually.  Return true without
31615         generating rtl if
31616         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
31617         whether all selected elements come from the same input, instead of
31618         checking each element individually.  Remove calls to gen_rtx_REG,
31619         start_sequence and end_sequence and instead assert that no rtl is
31620         generated.
31622 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
31624         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
31625         order of HIGH and CONST checks.
31627 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
31629         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
31630         if the destination isn't an SSA_NAME.
31632 2018-01-09  Richard Biener  <rguenther@suse.de>
31634         PR tree-optimization/83668
31635         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
31636         move prologue...
31637         (canonicalize_loop_form): ... here, renamed from ...
31638         (canonicalize_loop_closed_ssa_form): ... this and amended to
31639         swap successor edges for loop exit blocks to make us use
31640         the RPO order we need for initial schedule generation.
31642 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
31644         PR tree-optimization/64811
31645         * match.pd: When optimizing comparisons with Inf, avoid
31646         introducing or losing exceptions from comparisons with NaN.
31648 2018-01-09  Martin Liska  <mliska@suse.cz>
31650         PR sanitizer/82517
31651         * asan.c (shadow_mem_size): Add gcc_assert.
31653 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
31655         Don't save registers in main().
31657         PR target/83738
31658         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
31659         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
31660         * config/avr/avr.c (avr_set_current_function): Don't error if
31661         naked, OS_task or OS_main are specified at the same time.
31662         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
31663         OS_main.
31664         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
31665         attribute.
31666         * common/config/avr/avr-common.c (avr_option_optimization_table):
31667         Switch on -mmain-is-OS_task for optimizing compilations.
31669 2018-01-09  Richard Biener  <rguenther@suse.de>
31671         PR tree-optimization/83572
31672         * graphite.c: Include cfganal.h.
31673         (graphite_transform_loops): Connect infinite loops to exit
31674         and remove fake edges at the end.
31676 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
31678         * ipa-inline.c (edge_badness): Revert accidental checkin.
31680 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
31682         PR ipa/80763
31683         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
31684         symbols; not inline clones.
31686 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
31688         PR target/83507
31689         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
31690         hard registers.  Formatting fixes.
31692         PR preprocessor/83722
31693         * gcc.c (try_generate_repro): Pass
31694         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
31695         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
31696         do_report_bug.
31698 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
31699             Kito Cheng  <kito.cheng@gmail.com>
31701         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
31702         (riscv_leaf_function_p): Delete.
31703         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
31705 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31707         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
31708         function.
31709         (do_ifelse): New function.
31710         (do_isel): New function.
31711         (do_sub3): New function.
31712         (do_add3): New function.
31713         (do_load_mask_compare): New function.
31714         (do_overlap_load_compare): New function.
31715         (expand_compare_loop): New function.
31716         (expand_block_compare): Call expand_compare_loop() when appropriate.
31717         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
31718         option description.
31719         (-mblock-compare-inline-loop-limit): New option.
31721 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31723         PR target/83677
31724         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
31725         Reverse order of second and third operands in first alternative.
31726         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
31727         of first and second elements in UNSPEC_VPERMR vector.
31728         (altivec_expand_vec_perm_le): Likewise.
31730 2018-01-08  Jeff Law  <law@redhat.com>
31732         PR rtl-optimizatin/81308
31733         * tree-switch-conversion.c (cfg_altered): New file scoped static.
31734         (process_switch): If group_case_labels makes a change, then set
31735         cfg_altered.
31736         (pass_convert_switch::execute): If a switch is converted, then
31737         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
31739         PR rtl-optimization/81308
31740         * recog.c (split_all_insns): Conditionally cleanup the CFG after
31741         splitting insns.
31743 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
31745         PR target/83663 - Revert r255946
31746         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
31747         generation for cases where splatting a value is not useful.
31748         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
31749         across a vec_duplicate and a paradoxical subreg forming a vector
31750         mode to a vec_concat.
31752 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
31754         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
31755         -march=armv8.3-a variants.
31756         * config/arm/t-multilib: Likewise.
31757         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
31759 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
31761         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
31762         to generate rtl.
31763         (cceq_ior_compare_complement): Give it a name so I can use it, and
31764         change boolean_or_operator predicate to boolean_operator so it can
31765         be used to generate a crand.
31766         (eqne): New code iterator.
31767         (bd/bd_neg): New code_attrs.
31768         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
31769         a single define_insn.
31770         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
31771         decrement (bdnzt/bdnzf/bdzt/bdzf).
31772         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
31773         with the new names of the branch decrement patterns, and added the
31774         names of the branch decrement conditional patterns.
31776 2018-01-08  Richard Biener  <rguenther@suse.de>
31778         PR tree-optimization/83563
31779         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
31780         cache.
31782 2018-01-08  Richard Biener  <rguenther@suse.de>
31784         PR middle-end/83713
31785         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
31787 2018-01-08  Richard Biener  <rguenther@suse.de>
31789         PR tree-optimization/83685
31790         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
31791         references to abnormals.
31793 2018-01-08  Richard Biener  <rguenther@suse.de>
31795         PR lto/83719
31796         * dwarf2out.c (output_indirect_strings): Handle empty
31797         skeleton_debug_str_hash.
31798         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
31800 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
31802         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
31803         (emit_store_direct): Likewise.
31804         (arc_trampoline_adjust_address): Likewise.
31805         (arc_asm_trampoline_template): New function.
31806         (arc_initialize_trampoline): Use asm_trampoline_template.
31807         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
31808         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
31809         * config/arc/arc.md (flush_icache): Delete pattern.
31811 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
31813         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
31814         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
31815         munaligned-access.
31817 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
31819         PR target/83681
31820         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
31821         by not USED_FOR_TARGET.
31822         (make_pass_resolve_sw_modes): Likewise.
31824 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
31826         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
31827         USED_FOR_TARGET.
31829 2018-01-08  Richard Biener  <rguenther@suse.de>
31831         PR middle-end/83580
31832         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
31834 2018-01-08  Richard Biener  <rguenther@suse.de>
31836         PR middle-end/83517
31837         * match.pd ((t * 2) / 2) -> t): Add missing :c.
31839 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
31841         PR middle-end/81897
31842         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
31843         basic blocks with a small number of successors.
31844         (convert_control_dep_chain_into_preds): Improve handling of
31845         forwarder blocks.
31846         (dump_predicates): Split apart into...
31847         (dump_pred_chain): ...here...
31848         (dump_pred_info): ...and here.
31849         (can_one_predicate_be_invalidated_p): Add debugging printfs.
31850         (can_chain_union_be_invalidated_p): Improve check for invalidation
31851         of paths.
31852         (uninit_uses_cannot_happen): Avoid unnecessary if
31853         convert_control_dep_chain_into_preds yielded nothing.
31855 2018-01-06  Martin Sebor  <msebor@redhat.com>
31857         PR tree-optimization/83640
31858         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
31859         subtracting negative offset from size.
31860         (builtin_access::overlap): Adjust offset bounds of the access to fall
31861         within the size of the object if possible.
31863 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
31865         PR rtl-optimization/83699
31866         * expmed.c (extract_bit_field_1): Restrict the vector usage of
31867         extract_bit_field_as_subreg to cases in which the extracted
31868         value is also a vector.
31870         * lra-constraints.c (process_alt_operands): Test for the equivalence
31871         substitutions when detecting a possible reload cycle.
31873 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
31875         PR debug/83480
31876         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
31877         by default if flag_selective_schedling{,2}.  Formatting fixes.
31879         PR rtl-optimization/83682
31880         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
31881         if it has non-VECTOR_MODE element mode.
31882         (vec_duplicate_p): Likewise.
31884         PR middle-end/83694
31885         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
31886         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
31888 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
31890         PR target/83604
31891         * config/i386/i386-builtin.def
31892         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
31893         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
31894         Require also OPTION_MASK_ISA_AVX512F in addition to
31895         OPTION_MASK_ISA_GFNI.
31896         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
31897         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
31898         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
31899         to OPTION_MASK_ISA_GFNI.
31900         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
31901         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
31902         OPTION_MASK_ISA_AVX512BW.
31903         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
31904         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
31905         addition to OPTION_MASK_ISA_GFNI.
31906         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
31907         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
31908         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
31909         to OPTION_MASK_ISA_GFNI.
31910         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
31911         a requirement for all ISAs rather than any of them with a few
31912         exceptions.
31913         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
31914         processing.
31915         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
31916         bitmasks to be enabled with 3 exceptions, instead of requiring any
31917         enabled ISA with lots of exceptions.
31918         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
31919         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
31920         Change avx512bw in isa attribute to avx512f.
31921         * config/i386/sgxintrin.h: Add license boilerplate.
31922         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
31923         to __AVX512F__ and __AVX512VL to __AVX512VL__.
31924         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
31925         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
31926         defined.
31927         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
31928         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
31929         temporarily sse2 rather than sse if not enabled already.
31931         PR target/83604
31932         * config/i386/sse.md (VI248_VLBW): Rename to ...
31933         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
31934         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
31935         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
31936         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
31937         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
31938         mode iterator instead of VI248_VLBW.
31940 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
31942         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
31943         (record_modified): Skip clobbers; add debug output.
31944         (param_change_prob): Use sreal frequencies.
31946 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
31948         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
31949         punt for user-aligned variables.
31951 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
31953         * tree-chrec.c (chrec_contains_symbols): Return true for
31954         POLY_INT_CST.
31956 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
31958         PR target/82439
31959         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
31960         of (x|y) == x for BICS pattern.
31962 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
31964         PR tree-optimization/83605
31965         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
31966         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
31967         can throw.
31969 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
31971         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
31972         * config/epiphany/rtems.h: New file.
31974 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
31975             Uros Bizjak  <ubizjak@gmail.com>
31977         PR target/83554
31978         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
31979         QIreg_operand instead of register_operand predicate.
31980         * config/i386/i386.c (ix86_rop_should_change_byte_p,
31981         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
31982         comments instead of -fmitigate[-_]rop.
31984 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
31986         PR bootstrap/81926
31987         * cgraphunit.c (symbol_table::compile): Switch to text_section
31988         before calling assembly_start debug hook.
31989         * run-rtl-passes.c (run_rtl_passes): Likewise.
31990         Include output.h.
31992 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
31994         * tree-vrp.c (extract_range_from_binary_expr_1): Check
31995         range_int_cst_p rather than !symbolic_range_p before calling
31996         extract_range_from_multiplicative_op_1.
31998 2018-01-04  Jeff Law  <law@redhat.com>
32000         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
32001         redundant test in assertion.
32003 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
32005         * doc/rtl.texi: Document machine_mode wrapper classes.
32007 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
32009         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
32010         using tree_to_uhwi.
32012 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
32014         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
32015         the VEC_PERM_EXPR fold to fail.
32017 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
32019         PR debug/83585
32020         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
32021         to switched_sections.
32023 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
32025         PR target/83680
32026         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
32027         test for d.testing.
32029 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
32031         PR target/83387
32032         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
32033         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
32035 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
32037         PR debug/83666
32038         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
32039         is BLKmode and bitpos not zero or mode change is needed.
32041 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
32043         PR target/83675
32044         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
32045         TARGET_VIS2.
32047 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
32049         PR target/83628
32050         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
32051         instead of MULT rtx.  Update all corresponding splitters.
32052         (*saddl_se): Ditto.
32053         (*ssub<modesuffix>): Ditto.
32054         (*ssubl_se): Ditto.
32055         (*cmp_sadd_di): Update split patterns.
32056         (*cmp_sadd_si): Ditto.
32057         (*cmp_sadd_sidi): Ditto.
32058         (*cmp_ssub_di): Ditto.
32059         (*cmp_ssub_si): Ditto.
32060         (*cmp_ssub_sidi): Ditto.
32061         * config/alpha/predicates.md (const23_operand): New predicate.
32062         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
32063         Look for ASHIFT, not MULT inner operand.
32064         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
32066 2018-01-04  Martin Liska  <mliska@suse.cz>
32068         PR gcov-profile/83669
32069         * gcov.c (output_intermediate_file): Add version to intermediate
32070         gcov file.
32071         * doc/gcov.texi: Document new field 'version' in intermediate
32072         file format. Fix location of '-k' option of gcov command.
32074 2018-01-04  Martin Liska  <mliska@suse.cz>
32076         PR ipa/82352
32077         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
32079 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
32081         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
32083 2018-01-03  Martin Sebor  <msebor@redhat.com>
32085         PR tree-optimization/83655
32086         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
32087         checking calls with invalid arguments.
32089 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32091         * tree-vect-stmts.c (vect_get_store_rhs): New function.
32092         (vectorizable_mask_load_store): Delete.
32093         (vectorizable_call): Return false for masked loads and stores.
32094         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
32095         instead of gimple_assign_rhs1.
32096         (vectorizable_load): Handle IFN_MASK_LOAD.
32097         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
32099 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32101         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
32102         split out from..,
32103         (vectorizable_mask_load_store): ...here.
32104         (vectorizable_load): ...and here.
32106 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32108         * tree-vect-stmts.c (vect_build_all_ones_mask)
32109         (vect_build_zero_merge_argument): New functions, split out from...
32110         (vectorizable_load): ...here.
32112 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32114         * tree-vect-stmts.c (vect_check_store_rhs): New function,
32115         split out from...
32116         (vectorizable_mask_load_store): ...here.
32117         (vectorizable_store): ...and here.
32119 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32121         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
32122         split out from...
32123         (vectorizable_mask_load_store): ...here.
32125 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32127         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
32128         (vect_model_store_cost): Take a vec_load_store_type instead of a
32129         vect_def_type.
32130         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
32131         (vect_model_store_cost): Take a vec_load_store_type instead of a
32132         vect_def_type.
32133         (vectorizable_mask_load_store): Update accordingly.
32134         (vectorizable_store): Likewise.
32135         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
32137 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32139         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
32140         IFN_MASK_LOAD calls here rather than...
32141         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
32143 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32144             Alan Hayward  <alan.hayward@arm.com>
32145             David Sherwood  <david.sherwood@arm.com>
32147         * expmed.c (extract_bit_field_1): For vector extracts,
32148         fall back to extract_bit_field_as_subreg if vec_extract
32149         isn't available.
32151 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32152             Alan Hayward  <alan.hayward@arm.com>
32153             David Sherwood  <david.sherwood@arm.com>
32155         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
32156         they are variable or constant sized.
32157         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
32158         slots for constant-sized data.
32160 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32161             Alan Hayward  <alan.hayward@arm.com>
32162             David Sherwood  <david.sherwood@arm.com>
32164         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
32165         handling COND_EXPRs with boolean comparisons, try to find a better
32166         basis for the mask type than the boolean itself.
32168 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32170         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
32171         is calculated and how it can be overridden.
32172         * genmodes.c (max_bitsize_mode_any_mode): New variable.
32173         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
32174         if defined.
32175         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
32176         if nonzero.
32178 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32179             Alan Hayward  <alan.hayward@arm.com>
32180             David Sherwood  <david.sherwood@arm.com>
32182         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
32183         Remove the mode argument.
32184         (aarch64_simd_valid_immediate): Remove the mode and inverse
32185         arguments.
32186         * config/aarch64/iterators.md (bitsize): New iterator.
32187         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
32188         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
32189         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
32190         aarch64_simd_valid_immediate.
32191         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
32192         (aarch64_reg_or_bic_imm): Likewise.
32193         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
32194         with an insn_type enum and msl with a modifier_type enum.
32195         Replace element_width with a scalar_mode.  Change the shift
32196         to unsigned int.  Add constructors for scalar_float_mode and
32197         scalar_int_mode elements.
32198         (aarch64_vect_float_const_representable_p): Delete.
32199         (aarch64_can_const_movi_rtx_p)
32200         (aarch64_simd_scalar_immediate_valid_for_move)
32201         (aarch64_simd_make_constant): Update call to
32202         aarch64_simd_valid_immediate.
32203         (aarch64_advsimd_valid_immediate_hs): New function.
32204         (aarch64_advsimd_valid_immediate): Likewise.
32205         (aarch64_simd_valid_immediate): Remove mode and inverse
32206         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
32207         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
32208         and aarch64_float_const_representable_p on the result.
32209         (aarch64_output_simd_mov_immediate): Remove mode argument.
32210         Update call to aarch64_simd_valid_immediate and use of
32211         simd_immediate_info.
32212         (aarch64_output_scalar_simd_mov_immediate): Update call
32213         accordingly.
32215 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32216             Alan Hayward  <alan.hayward@arm.com>
32217             David Sherwood  <david.sherwood@arm.com>
32219         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
32220         (mode_nunits): Likewise CONST_MODE_NUNITS.
32221         * machmode.def (ADJUST_NUNITS): Document.
32222         * genmodes.c (mode_data::need_nunits_adj): New field.
32223         (blank_mode): Update accordingly.
32224         (adj_nunits): New variable.
32225         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
32226         parameter.
32227         (emit_mode_size_inline): Set need_bytesize_adj for all modes
32228         listed in adj_nunits.
32229         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
32230         listed in adj_nunits.  Don't emit case statements for such modes.
32231         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
32232         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
32233         nothing if adj_nunits is nonnull.
32234         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
32235         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
32236         (emit_mode_fbit): Update use of print_maybe_const_decl.
32237         (emit_move_size): Likewise.  Treat the array as non-const
32238         if adj_nunits.
32239         (emit_mode_adjustments): Handle adj_nunits.
32241 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32243         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
32244         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
32245         (VECTOR_MODES): Use it.
32246         (make_vector_modes): Take the prefix as an argument.
32248 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32249             Alan Hayward  <alan.hayward@arm.com>
32250             David Sherwood  <david.sherwood@arm.com>
32252         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
32253         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
32254         for MODE_VECTOR_BOOL.
32255         * machmode.def (VECTOR_BOOL_MODE): Document.
32256         * genmodes.c (VECTOR_BOOL_MODE): New macro.
32257         (make_vector_bool_mode): New function.
32258         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
32259         MODE_VECTOR_BOOL.
32260         * lto-streamer-in.c (lto_input_mode_table): Likewise.
32261         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
32262         Likewise.
32263         * stor-layout.c (int_mode_for_mode): Likewise.
32264         * tree.c (build_vector_type_for_mode): Likewise.
32265         * varasm.c (output_constant_pool_2): Likewise.
32266         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
32267         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
32268         for MODE_VECTOR_BOOL.
32269         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
32270         of mode class checks.
32271         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
32272         instead of a list of mode class checks.
32273         (expand_vector_scalar_condition): Likewise.
32274         (type_for_widest_vector_mode): Handle BImode as an inner mode.
32276 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32277             Alan Hayward  <alan.hayward@arm.com>
32278             David Sherwood  <david.sherwood@arm.com>
32280         * machmode.h (mode_size): Change from unsigned short to
32281         poly_uint16_pod.
32282         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
32283         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
32284         or if measurement_type is not polynomial.
32285         (fixed_size_mode::includes_p): Check for constant-sized modes.
32286         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
32287         return a poly_uint16 rather than an unsigned short.
32288         (emit_mode_size): Change the type of mode_size from unsigned short
32289         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
32290         (emit_mode_adjustments): Cope with polynomial vector sizes.
32291         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
32292         for GET_MODE_SIZE.
32293         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
32294         for GET_MODE_SIZE.
32295         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
32296         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
32297         * caller-save.c (setup_save_areas): Likewise.
32298         (replace_reg_with_saved_mem): Likewise.
32299         * calls.c (emit_library_call_value_1): Likewise.
32300         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
32301         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
32302         (gen_lowpart_for_combine): Likewise.
32303         * convert.c (convert_to_integer_1): Likewise.
32304         * cse.c (equiv_constant, cse_insn): Likewise.
32305         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
32306         (cselib_subst_to_values): Likewise.
32307         * dce.c (word_dce_process_block): Likewise.
32308         * df-problems.c (df_word_lr_mark_ref): Likewise.
32309         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
32310         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
32311         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
32312         (rtl_for_decl_location): Likewise.
32313         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
32314         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
32315         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
32316         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
32317         (expand_expr_real_1): Likewise.
32318         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
32319         (pad_below): Likewise.
32320         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
32321         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
32322         * ira.c (get_subreg_tracking_sizes): Likewise.
32323         * ira-build.c (ira_create_allocno_objects): Likewise.
32324         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
32325         (ira_sort_regnos_for_alter_reg): Likewise.
32326         * ira-costs.c (record_operand_costs): Likewise.
32327         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
32328         (resolve_simple_move): Likewise.
32329         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
32330         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
32331         (lra_constraints): Likewise.
32332         (CONST_POOL_OK_P): Reject variable-sized modes.
32333         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
32334         (add_pseudo_to_slot, lra_spill): Likewise.
32335         * omp-low.c (omp_clause_aligned_alignment): Likewise.
32336         * optabs-query.c (get_best_extraction_insn): Likewise.
32337         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
32338         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
32339         (expand_mult_highpart, valid_multiword_target_p): Likewise.
32340         * recog.c (offsettable_address_addr_space_p): Likewise.
32341         * regcprop.c (maybe_mode_change): Likewise.
32342         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
32343         * regrename.c (build_def_use): Likewise.
32344         * regstat.c (dump_reg_info): Likewise.
32345         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
32346         (find_reloads, find_reloads_subreg_address): Likewise.
32347         * reload1.c (eliminate_regs_1): Likewise.
32348         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
32349         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
32350         (simplify_binary_operation_1, simplify_subreg): Likewise.
32351         * targhooks.c (default_function_arg_padding): Likewise.
32352         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
32353         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
32354         (verify_gimple_assign_ternary): Likewise.
32355         * tree-inline.c (estimate_move_cost): Likewise.
32356         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32357         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
32358         (get_address_cost_ainc): Likewise.
32359         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
32360         (vect_supportable_dr_alignment): Likewise.
32361         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
32362         (vectorizable_reduction): Likewise.
32363         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
32364         (vectorizable_operation, vectorizable_load): Likewise.
32365         * tree.c (build_same_sized_truth_vector_type): Likewise.
32366         * valtrack.c (cleanup_auto_inc_dec): Likewise.
32367         * var-tracking.c (emit_note_insn_var_location): Likewise.
32368         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
32369         (ADDR_VEC_ALIGN): Likewise.
32371 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32372             Alan Hayward  <alan.hayward@arm.com>
32373             David Sherwood  <david.sherwood@arm.com>
32375         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
32376         unsigned short.
32377         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
32378         or if measurement_type is polynomial.
32379         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
32380         * combine.c (make_extraction): Likewise.
32381         * dse.c (find_shift_sequence): Likewise.
32382         * dwarf2out.c (mem_loc_descriptor): Likewise.
32383         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
32384         (extract_bit_field, extract_low_bits): Likewise.
32385         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
32386         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
32387         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
32388         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
32389         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
32390         * reload.c (find_reloads): Likewise.
32391         * reload1.c (alter_reg): Likewise.
32392         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
32393         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
32394         * tree-if-conv.c (predicate_mem_writes): Likewise.
32395         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
32396         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
32397         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
32398         * valtrack.c (dead_debug_insert_temp): Likewise.
32399         * varasm.c (mergeable_constant_section): Likewise.
32400         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
32402 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32403             Alan Hayward  <alan.hayward@arm.com>
32404             David Sherwood  <david.sherwood@arm.com>
32406         * expr.c (expand_assignment): Cope with polynomial mode sizes
32407         when assigning to a CONCAT.
32409 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32410             Alan Hayward  <alan.hayward@arm.com>
32411             David Sherwood  <david.sherwood@arm.com>
32413         * machmode.h (mode_precision): Change from unsigned short to
32414         poly_uint16_pod.
32415         (mode_to_precision): Return a poly_uint16 rather than an unsigned
32416         short.
32417         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
32418         or if measurement_type is not polynomial.
32419         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
32420         in which the mode is already known to be a scalar_int_mode.
32421         * genmodes.c (emit_mode_precision): Change the type of mode_precision
32422         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
32423         initializer.
32424         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
32425         for GET_MODE_PRECISION.
32426         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
32427         for GET_MODE_PRECISION.
32428         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
32429         as polynomial.
32430         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
32431         (expand_field_assignment, make_extraction): Likewise.
32432         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
32433         (get_last_value): Likewise.
32434         * convert.c (convert_to_integer_1): Likewise.
32435         * cse.c (cse_insn): Likewise.
32436         * expr.c (expand_expr_real_1): Likewise.
32437         * lra-constraints.c (simplify_operand_subreg): Likewise.
32438         * optabs-query.c (can_atomic_load_p): Likewise.
32439         * optabs.c (expand_atomic_load): Likewise.
32440         (expand_atomic_store): Likewise.
32441         * ree.c (combine_reaching_defs): Likewise.
32442         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
32443         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
32444         * tree.h (type_has_mode_precision_p): Likewise.
32445         * ubsan.c (instrument_si_overflow): Likewise.
32447 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32448             Alan Hayward  <alan.hayward@arm.com>
32449             David Sherwood  <david.sherwood@arm.com>
32451         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
32452         polynomial numbers of units.
32453         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
32454         (valid_vector_subparts_p): New function.
32455         (build_vector_type): Remove temporary shim and take the number
32456         of units as a poly_uint64 rather than an int.
32457         (build_opaque_vector_type): Take the number of units as a
32458         poly_uint64 rather than an int.
32459         * tree.c (build_vector_from_ctor): Handle polynomial
32460         TYPE_VECTOR_SUBPARTS.
32461         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
32462         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
32463         (build_vector_from_val): If the number of units is variable,
32464         use build_vec_duplicate_cst for constant operands and
32465         VEC_DUPLICATE_EXPR otherwise.
32466         (make_vector_type): Remove temporary is_constant ().
32467         (build_vector_type, build_opaque_vector_type): Take the number of
32468         units as a poly_uint64 rather than an int.
32469         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
32470         VECTOR_CST_NELTS.
32471         * cfgexpand.c (expand_debug_expr): Likewise.
32472         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
32473         (store_constructor, expand_expr_real_1): Likewise.
32474         (const_scalar_mask_from_tree): Likewise.
32475         * fold-const-call.c (fold_const_reduction): Likewise.
32476         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
32477         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
32478         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
32479         (fold_relational_const): Likewise.
32480         (native_interpret_vector): Likewise.  Change the size from an
32481         int to an unsigned int.
32482         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
32483         TYPE_VECTOR_SUBPARTS.
32484         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
32485         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
32486         duplicating a non-constant operand into a variable-length vector.
32487         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
32488         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
32489         * ipa-icf.c (sem_variable::equals): Likewise.
32490         * match.pd: Likewise.
32491         * omp-simd-clone.c (simd_clone_subparts): Likewise.
32492         * print-tree.c (print_node): Likewise.
32493         * stor-layout.c (layout_type): Likewise.
32494         * targhooks.c (default_builtin_vectorization_cost): Likewise.
32495         * tree-cfg.c (verify_gimple_comparison): Likewise.
32496         (verify_gimple_assign_binary): Likewise.
32497         (verify_gimple_assign_ternary): Likewise.
32498         (verify_gimple_assign_single): Likewise.
32499         * tree-pretty-print.c (dump_generic_node): Likewise.
32500         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32501         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
32502         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
32503         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
32504         (vect_shift_permute_load_chain): Likewise.
32505         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
32506         (expand_vector_condition, optimize_vector_constructor): Likewise.
32507         (lower_vec_perm, get_compute_type): Likewise.
32508         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
32509         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
32510         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
32511         (vect_recog_mask_conversion_pattern): Likewise.
32512         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
32513         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
32514         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32515         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
32516         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
32517         (vectorizable_shift, vectorizable_operation, vectorizable_store)
32518         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
32519         (supportable_widening_operation): Likewise.
32520         (supportable_narrowing_operation): Likewise.
32521         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
32522         Likewise.
32523         * varasm.c (output_constant): Likewise.
32525 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32526             Alan Hayward  <alan.hayward@arm.com>
32527             David Sherwood  <david.sherwood@arm.com>
32529         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
32530         so that both the length == 3 and length != 3 cases set up their
32531         own permute vectors.  Add comments explaining why we know the
32532         number of elements is constant.
32533         (vect_permute_load_chain): Likewise.
32535 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32536             Alan Hayward  <alan.hayward@arm.com>
32537             David Sherwood  <david.sherwood@arm.com>
32539         * machmode.h (mode_nunits): Change from unsigned char to
32540         poly_uint16_pod.
32541         (ONLY_FIXED_SIZE_MODES): New macro.
32542         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
32543         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
32544         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
32545         New typedefs.
32546         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
32547         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
32548         or if measurement_type is not polynomial.
32549         * genmodes.c (ZERO_COEFFS): New macro.
32550         (emit_mode_nunits_inline): Make mode_nunits_inline return a
32551         poly_uint16.
32552         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
32553         Use ZERO_COEFFS when emitting initializers.
32554         * data-streamer.h (bp_pack_poly_value): New function.
32555         (bp_unpack_poly_value): Likewise.
32556         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
32557         for GET_MODE_NUNITS.
32558         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
32559         for GET_MODE_NUNITS.
32560         * tree.c (make_vector_type): Remove temporary shim and make
32561         the real function take the number of units as a poly_uint64
32562         rather than an int.
32563         (build_vector_type_for_mode): Handle polynomial nunits.
32564         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
32565         * emit-rtl.c (const_vec_series_p_1): Likewise.
32566         (gen_rtx_CONST_VECTOR): Likewise.
32567         * fold-const.c (test_vec_duplicate_folding): Likewise.
32568         * genrecog.c (validate_pattern): Likewise.
32569         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
32570         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
32571         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
32572         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
32573         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
32574         * rtlanal.c (subreg_get_info): Likewise.
32575         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32576         (vect_grouped_load_supported): Likewise.
32577         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
32578         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
32579         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
32580         (simplify_const_unary_operation, simplify_binary_operation_1)
32581         (simplify_const_binary_operation, simplify_ternary_operation)
32582         (test_vector_ops_duplicate, test_vector_ops): Likewise.
32583         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
32584         instead of CONST_VECTOR_NUNITS.
32585         * varasm.c (output_constant_pool_2): Likewise.
32586         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
32587         explicit-encoded elements in the XVEC for variable-length vectors.
32589 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32591         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
32593 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32594             Alan Hayward  <alan.hayward@arm.com>
32595             David Sherwood  <david.sherwood@arm.com>
32597         * coretypes.h (fixed_size_mode): Declare.
32598         (fixed_size_mode_pod): New typedef.
32599         * builtins.h (target_builtins::x_apply_args_mode)
32600         (target_builtins::x_apply_result_mode): Change type to
32601         fixed_size_mode_pod.
32602         * builtins.c (apply_args_size, apply_result_size, result_vector)
32603         (expand_builtin_apply_args_1, expand_builtin_apply)
32604         (expand_builtin_return): Update accordingly.
32606 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32608         * cse.c (hash_rtx_cb): Hash only the encoded elements.
32609         * cselib.c (cselib_hash_rtx): Likewise.
32610         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
32611         CONST_VECTOR encoding.
32613 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
32614             Jeff Law  <law@redhat.com>
32616         PR target/83641
32617         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
32618         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
32619         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
32620         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
32622         PR target/83641
32623         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
32624         explicitly probe *sp in a noreturn function if there were any callee
32625         register saves or frame pointer is needed.
32627 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
32629         PR debug/83621
32630         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
32631         BLKmode for ternary, binary or unary expressions.
32633         PR debug/83645
32634         * var-tracking.c (delete_vta_debug_insn): New inline function.
32635         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
32636         insns from get_insns () to NULL instead of each bb separately.
32637         Use delete_vta_debug_insn.  No longer static.
32638         (vt_debug_insns_local, variable_tracking_main_1): Adjust
32639         delete_vta_debug_insns callers.
32640         * rtl.h (delete_vta_debug_insns): Declare.
32641         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
32642         instead of variable_tracking_main.
32644 2018-01-03  Martin Sebor  <msebor@redhat.com>
32646         PR tree-optimization/83603
32647         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
32648         arguments past the endof the argument list in functions declared
32649         without a prototype.
32650         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
32651         Avoid checking when arguments are null.
32653 2018-01-03  Martin Sebor  <msebor@redhat.com>
32655         PR c/83559
32656         * doc/extend.texi (attribute const): Fix a typo.
32657         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
32658         issuing -Wsuggest-attribute for void functions.
32660 2018-01-03  Martin Sebor  <msebor@redhat.com>
32662         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
32663         offset_int::from instead of wide_int::to_shwi.
32664         (maybe_diag_overlap): Remove assertion.
32665         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
32666         * gimple-ssa-sprintf.c (format_directive): Same.
32667         (parse_directive): Same.
32668         (sprintf_dom_walker::compute_format_length): Same.
32669         (try_substitute_return_value): Same.
32671 2018-01-03  Jeff Law  <law@redhat.com>
32673         PR middle-end/83654
32674         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
32675         non-constant residual for zero at runtime and avoid probing in
32676         that case.  Reorganize code for trailing problem to mirror handling
32677         of the residual.
32679 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32681         PR tree-optimization/83501
32682         * tree-ssa-strlen.c (get_string_cst): New.
32683         (handle_char_store): Call get_string_cst.
32685 2018-01-03  Martin Liska  <mliska@suse.cz>
32687         PR tree-optimization/83593
32688         * tree-ssa-strlen.c: Include tree-cfg.h.
32689         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
32690         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
32691         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
32692         to false.
32693         (strlen_dom_walker::before_dom_children): Call
32694         gimple_purge_dead_eh_edges. Dump tranformation with details
32695         dump flags.
32696         (strlen_dom_walker::before_dom_children): Update call by adding
32697         new argument cleanup_eh.
32698         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
32700 2018-01-03  Martin Liska  <mliska@suse.cz>
32702         PR ipa/83549
32703         * cif-code.def (VARIADIC_THUNK): New enum value.
32704         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
32705         thunks.
32707 2018-01-03  Jan Beulich  <jbeulich@suse.com>
32709         * sse.md (mov<mode>_internal): Tighten condition for when to use
32710         vmovdqu<ssescalarsize> for TI and OI modes.
32712 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
32714         Update copyright years.
32716 2018-01-03  Martin Liska  <mliska@suse.cz>
32718         PR ipa/83594
32719         * ipa-visibility.c (function_and_variable_visibility): Skip
32720         functions with noipa attribure.
32722 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
32724         * gcc.c (process_command): Update copyright notice dates.
32725         * gcov-dump.c (print_version): Ditto.
32726         * gcov.c (print_version): Ditto.
32727         * gcov-tool.c (print_version): Ditto.
32728         * gengtype.c (create_file): Ditto.
32729         * doc/cpp.texi: Bump @copying's copyright year.
32730         * doc/cppinternals.texi: Ditto.
32731         * doc/gcc.texi: Ditto.
32732         * doc/gccint.texi: Ditto.
32733         * doc/gcov.texi: Ditto.
32734         * doc/install.texi: Ditto.
32735         * doc/invoke.texi: Ditto.
32737 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32739         * vector-builder.h (vector_builder::m_full_nelts): Change from
32740         unsigned int to poly_uint64.
32741         (vector_builder::full_nelts): Update prototype accordingly.
32742         (vector_builder::new_vector): Likewise.
32743         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
32744         (vector_builder::operator ==): Likewise.
32745         (vector_builder::finalize): Likewise.
32746         * int-vector-builder.h (int_vector_builder::int_vector_builder):
32747         Take the number of elements as a poly_uint64 rather than an
32748         unsigned int.
32749         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
32750         from unsigned int to poly_uint64.
32751         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
32752         (vec_perm_indices::new_vector): Likewise.
32753         (vec_perm_indices::length): Likewise.
32754         (vec_perm_indices::nelts_per_input): Likewise.
32755         (vec_perm_indices::input_nelts): Likewise.
32756         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
32757         number of elements per input as a poly_uint64 rather than an
32758         unsigned int.  Use the original encoding for variable-length
32759         vectors, rather than clamping each individual element.
32760         For the second and subsequent elements in each pattern,
32761         clamp the step and base before clamping their sum.
32762         (vec_perm_indices::series_p): Handle polynomial element counts.
32763         (vec_perm_indices::all_in_range_p): Likewise.
32764         (vec_perm_indices_to_tree): Likewise.
32765         (vec_perm_indices_to_rtx): Likewise.
32766         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
32767         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
32768         (tree_vector_builder::new_binary_operation): Handle polynomial
32769         element counts.  Return false if we need to know the number
32770         of elements at compile time.
32771         * fold-const.c (fold_vec_perm): Punt if the number of elements
32772         isn't known at compile time.
32774 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32776         * vec-perm-indices.h (vec_perm_builder): Change element type
32777         from HOST_WIDE_INT to poly_int64.
32778         (vec_perm_indices::element_type): Update accordingly.
32779         (vec_perm_indices::clamp): Handle polynomial element_types.
32780         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
32781         (vec_perm_indices::all_in_range_p): Likewise.
32782         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
32783         than shwi trees.
32784         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
32785         polynomial vec_perm_indices element types.
32786         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
32787         * fold-const.c (fold_vec_perm): Likewise.
32788         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
32789         * tree-vect-generic.c (lower_vec_perm): Likewise.
32790         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
32791         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
32792         element type to HOST_WIDE_INT.
32794 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32795             Alan Hayward  <alan.hayward@arm.com>
32796             David Sherwood  <david.sherwood@arm.com>
32798         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
32799         rather than an int.  Use plus_constant.
32800         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
32801         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
32803 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32804             Alan Hayward  <alan.hayward@arm.com>
32805             David Sherwood  <david.sherwood@arm.com>
32807         * calls.c (emit_call_1, expand_call): Change struct_value_size from
32808         a HOST_WIDE_INT to a poly_int64.
32810 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32811             Alan Hayward  <alan.hayward@arm.com>
32812             David Sherwood  <david.sherwood@arm.com>
32814         * calls.c (load_register_parameters): Cope with polynomial
32815         mode sizes.  Require a constant size for BLKmode parameters
32816         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
32817         forces a parameter to be padded at the lsb end in order to
32818         fill a complete number of words, require the parameter size
32819         to be ordered wrt UNITS_PER_WORD.
32821 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32822             Alan Hayward  <alan.hayward@arm.com>
32823             David Sherwood  <david.sherwood@arm.com>
32825         * reload1.c (spill_stack_slot_width): Change element type
32826         from unsigned int to poly_uint64_pod.
32827         (alter_reg): Treat mode sizes as polynomial.
32829 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32830             Alan Hayward  <alan.hayward@arm.com>
32831             David Sherwood  <david.sherwood@arm.com>
32833         * reload.c (complex_word_subreg_p): New function.
32834         (reload_inner_reg_of_subreg, push_reload): Use it.
32836 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32837             Alan Hayward  <alan.hayward@arm.com>
32838             David Sherwood  <david.sherwood@arm.com>
32840         * lra-constraints.c (process_alt_operands): Reject matched
32841         operands whose sizes aren't ordered.
32842         (match_reload): Refer to this check here.
32844 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32845             Alan Hayward  <alan.hayward@arm.com>
32846             David Sherwood  <david.sherwood@arm.com>
32848         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
32849         that the mode size is in the set {1, 2, 4, 8, 16}.
32851 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32852             Alan Hayward  <alan.hayward@arm.com>
32853             David Sherwood  <david.sherwood@arm.com>
32855         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
32856         Use plus_constant instead of gen_rtx_PLUS.
32858 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32859             Alan Hayward  <alan.hayward@arm.com>
32860             David Sherwood  <david.sherwood@arm.com>
32862         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
32863         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
32864         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
32865         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
32866         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
32867         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
32868         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
32869         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
32870         * config/i386/i386.c (ix86_push_rounding): ...this new function.
32871         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
32872         a poly_int64.
32873         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
32874         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
32875         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
32876         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
32877         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
32878         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
32879         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
32880         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
32881         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
32882         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
32883         function.
32884         * expr.c (emit_move_resolve_push): Treat the input and result
32885         of PUSH_ROUNDING as a poly_int64.
32886         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
32887         (emit_push_insn): Likewise.
32888         * lra-eliminations.c (mark_not_eliminable): Likewise.
32889         * recog.c (push_operand): Likewise.
32890         * reload1.c (elimination_effects): Likewise.
32891         * rtlanal.c (nonzero_bits1): Likewise.
32892         * calls.c (store_one_arg): Likewise.  Require the padding to be
32893         known at compile time.
32895 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32896             Alan Hayward  <alan.hayward@arm.com>
32897             David Sherwood  <david.sherwood@arm.com>
32899         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
32900         Use plus_constant instead of gen_rtx_PLUS.
32902 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32903             Alan Hayward  <alan.hayward@arm.com>
32904             David Sherwood  <david.sherwood@arm.com>
32906         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
32907         rather than an int.
32909 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32910             Alan Hayward  <alan.hayward@arm.com>
32911             David Sherwood  <david.sherwood@arm.com>
32913         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
32914         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
32915         via stack temporaries.  Treat the mode size as polynomial too.
32917 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32918             Alan Hayward  <alan.hayward@arm.com>
32919             David Sherwood  <david.sherwood@arm.com>
32921         * expr.c (expand_expr_real_2): When handling conversions involving
32922         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
32923         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
32924         as a poly_uint64 too.
32926 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32927             Alan Hayward  <alan.hayward@arm.com>
32928             David Sherwood  <david.sherwood@arm.com>
32930         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
32932 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32933             Alan Hayward  <alan.hayward@arm.com>
32934             David Sherwood  <david.sherwood@arm.com>
32936         * combine.c (can_change_dest_mode): Handle polynomial
32937         REGMODE_NATURAL_SIZE.
32938         * expmed.c (store_bit_field_1): Likewise.
32939         * expr.c (store_constructor): Likewise.
32940         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
32941         and polynomial REGMODE_NATURAL_SIZE.
32942         (gen_lowpart_common): Likewise.
32943         * reginfo.c (record_subregs_of_mode): Likewise.
32944         * rtlanal.c (read_modify_subreg_p): Likewise.
32946 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32947             Alan Hayward  <alan.hayward@arm.com>
32948             David Sherwood  <david.sherwood@arm.com>
32950         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
32951         numbers of elements.
32953 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32954             Alan Hayward  <alan.hayward@arm.com>
32955             David Sherwood  <david.sherwood@arm.com>
32957         * match.pd: Cope with polynomial numbers of vector elements.
32959 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32960             Alan Hayward  <alan.hayward@arm.com>
32961             David Sherwood  <david.sherwood@arm.com>
32963         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
32964         in a POINTER_PLUS_EXPR.
32966 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32967             Alan Hayward  <alan.hayward@arm.com>
32968             David Sherwood  <david.sherwood@arm.com>
32970         * omp-simd-clone.c (simd_clone_subparts): New function.
32971         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
32972         (ipa_simd_modify_function_body): Likewise.
32974 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32975             Alan Hayward  <alan.hayward@arm.com>
32976             David Sherwood  <david.sherwood@arm.com>
32978         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
32979         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
32980         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
32981         (expand_vector_condition, vector_element): Likewise.
32982         (subparts_gt): New function.
32983         (get_compute_type): Use subparts_gt.
32984         (count_type_subparts): Delete.
32985         (expand_vector_operations_1): Use subparts_gt instead of
32986         count_type_subparts.
32988 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32989             Alan Hayward  <alan.hayward@arm.com>
32990             David Sherwood  <david.sherwood@arm.com>
32992         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
32993         (vect_compile_time_alias): ...this new function.  Do the calculation
32994         on poly_ints rather than trees.
32995         (vect_prune_runtime_alias_test_list): Update call accordingly.
32997 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32998             Alan Hayward  <alan.hayward@arm.com>
32999             David Sherwood  <david.sherwood@arm.com>
33001         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
33002         numbers of units.
33003         (vect_schedule_slp_instance): Likewise.
33005 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33006             Alan Hayward  <alan.hayward@arm.com>
33007             David Sherwood  <david.sherwood@arm.com>
33009         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
33010         constant and extern definitions for variable-length vectors.
33011         (vect_get_constant_vectors): Note that the number of units
33012         is known to be constant.
33014 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33015             Alan Hayward  <alan.hayward@arm.com>
33016             David Sherwood  <david.sherwood@arm.com>
33018         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
33019         of units as polynomial.  Choose between WIDE and NARROW based
33020         on multiple_p.
33022 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33023             Alan Hayward  <alan.hayward@arm.com>
33024             David Sherwood  <david.sherwood@arm.com>
33026         * tree-vect-stmts.c (simd_clone_subparts): New function.
33027         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
33029 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33030             Alan Hayward  <alan.hayward@arm.com>
33031             David Sherwood  <david.sherwood@arm.com>
33033         * tree-vect-stmts.c (vectorizable_call): Treat the number of
33034         vectors as polynomial.  Use build_index_vector for
33035         IFN_GOMP_SIMD_LANE.
33037 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33038             Alan Hayward  <alan.hayward@arm.com>
33039             David Sherwood  <david.sherwood@arm.com>
33041         * tree-vect-stmts.c (get_load_store_type): Treat the number of
33042         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
33043         for variable-length vectors.
33044         (vectorizable_mask_load_store): Treat the number of units as
33045         polynomial, asserting that it is constant if the condition has
33046         already been enforced.
33047         (vectorizable_store, vectorizable_load): Likewise.
33049 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33050             Alan Hayward  <alan.hayward@arm.com>
33051             David Sherwood  <david.sherwood@arm.com>
33053         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
33054         of units as polynomial.  Punt if we can't tell at compile time
33055         which vector contains the final result.
33057 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33058             Alan Hayward  <alan.hayward@arm.com>
33059             David Sherwood  <david.sherwood@arm.com>
33061         * tree-vect-loop.c (vectorizable_induction): Treat the number
33062         of units as polynomial.  Punt on SLP inductions.  Use an integer
33063         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
33064         cast of such a series for variable-length floating-point
33065         reductions.
33067 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33068             Alan Hayward  <alan.hayward@arm.com>
33069             David Sherwood  <david.sherwood@arm.com>
33071         * tree.h (build_index_vector): Declare.
33072         * tree.c (build_index_vector): New function.
33073         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
33074         of units as polynomial, forcibly converting it to a constant if
33075         vectorizable_reduction has already enforced the condition.
33076         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
33077         to create a {1,2,3,...} vector.
33078         (vectorizable_reduction): Treat the number of units as polynomial.
33079         Choose vectype_in based on the largest scalar element size rather
33080         than the smallest number of units.  Enforce the restrictions
33081         relied on above.
33083 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33084             Alan Hayward  <alan.hayward@arm.com>
33085             David Sherwood  <david.sherwood@arm.com>
33087         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
33088         number of units as polynomial.
33090 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33091             Alan Hayward  <alan.hayward@arm.com>
33092             David Sherwood  <david.sherwood@arm.com>
33094         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
33095         * target.def (autovectorize_vector_sizes): Return the vector sizes
33096         by pointer, using vector_sizes rather than a bitmask.
33097         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
33098         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
33099         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
33100         Likewise.
33101         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
33102         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
33103         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
33104         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
33105         * omp-general.c (omp_max_vf): Likewise.
33106         * omp-low.c (omp_clause_aligned_alignment): Likewise.
33107         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
33108         * tree-vect-loop.c (vect_analyze_loop): Likewise.
33109         * tree-vect-slp.c (vect_slp_bb): Likewise.
33110         * doc/tm.texi: Regenerate.
33111         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
33112         to a poly_uint64.
33113         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
33114         the vector size as a poly_uint64 rather than an unsigned int.
33115         (current_vector_size): Change from an unsigned int to a poly_uint64.
33116         (get_vectype_for_scalar_type): Update accordingly.
33117         * tree.h (build_truth_vector_type): Take the size and number of
33118         units as a poly_uint64 rather than an unsigned int.
33119         (build_vector_type): Add a temporary overload that takes
33120         the number of units as a poly_uint64 rather than an unsigned int.
33121         * tree.c (make_vector_type): Likewise.
33122         (build_truth_vector_type): Take the number of units as a poly_uint64
33123         rather than an unsigned int.
33125 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33126             Alan Hayward  <alan.hayward@arm.com>
33127             David Sherwood  <david.sherwood@arm.com>
33129         * target.def (get_mask_mode): Take the number of units and length
33130         as poly_uint64s rather than unsigned ints.
33131         * targhooks.h (default_get_mask_mode): Update accordingly.
33132         * targhooks.c (default_get_mask_mode): Likewise.
33133         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
33134         * doc/tm.texi: Regenerate.
33136 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33137             Alan Hayward  <alan.hayward@arm.com>
33138             David Sherwood  <david.sherwood@arm.com>
33140         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
33141         * omp-general.c (omp_max_vf): Likewise.
33142         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
33143         (expand_omp_simd): Handle polynomial safelen.
33144         * omp-low.c (omplow_simd_context): Add a default constructor.
33145         (omplow_simd_context::max_vf): Change from int to poly_uint64.
33146         (lower_rec_simd_input_clauses): Update accordingly.
33147         (lower_rec_input_clauses): Likewise.
33149 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33150             Alan Hayward  <alan.hayward@arm.com>
33151             David Sherwood  <david.sherwood@arm.com>
33153         * tree-vectorizer.h (vect_nunits_for_cost): New function.
33154         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
33155         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
33156         (vect_analyze_slp_cost): Likewise.
33157         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
33158         (vect_model_load_cost): Likewise.
33160 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33161             Alan Hayward  <alan.hayward@arm.com>
33162             David Sherwood  <david.sherwood@arm.com>
33164         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
33165         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
33166         from an unsigned int * to a poly_uint64_pod *.
33167         (calculate_unrolling_factor): New function.
33168         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
33170 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33171             Alan Hayward  <alan.hayward@arm.com>
33172             David Sherwood  <david.sherwood@arm.com>
33174         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
33175         from an unsigned int to a poly_uint64.
33176         (_loop_vec_info::slp_unrolling_factor): Likewise.
33177         (_loop_vec_info::vectorization_factor): Change from an int
33178         to a poly_uint64.
33179         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
33180         (vect_get_num_vectors): New function.
33181         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
33182         (vect_get_num_copies): Use vect_get_num_vectors.
33183         (vect_analyze_data_ref_dependences): Change max_vf from an int *
33184         to an unsigned int *.
33185         (vect_analyze_data_refs): Change min_vf from an int * to a
33186         poly_uint64 *.
33187         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
33188         than an unsigned HOST_WIDE_INT.
33189         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
33190         (vect_analyze_data_ref_dependence): Change max_vf from an int *
33191         to an unsigned int *.
33192         (vect_analyze_data_ref_dependences): Likewise.
33193         (vect_compute_data_ref_alignment): Handle polynomial vf.
33194         (vect_enhance_data_refs_alignment): Likewise.
33195         (vect_prune_runtime_alias_test_list): Likewise.
33196         (vect_shift_permute_load_chain): Likewise.
33197         (vect_supportable_dr_alignment): Likewise.
33198         (dependence_distance_ge_vf): Take the vectorization factor as a
33199         poly_uint64 rather than an unsigned HOST_WIDE_INT.
33200         (vect_analyze_data_refs): Change min_vf from an int * to a
33201         poly_uint64 *.
33202         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
33203         vfm1 as a poly_uint64 rather than an int.  Make the same change
33204         for the returned bound_scalar.
33205         (vect_gen_vector_loop_niters): Handle polynomial vf.
33206         (vect_do_peeling): Likewise.  Update call to
33207         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
33208         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
33209         be constant.
33210         * tree-vect-loop.c (vect_determine_vectorization_factor)
33211         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
33212         (vect_get_known_peeling_cost): Likewise.
33213         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
33214         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
33215         (vect_transform_loop): Likewise.  Use the lowest possible VF when
33216         updating the upper bounds of the loop.
33217         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
33218         rather than an int.
33219         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
33220         polynomial unroll factors.
33221         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
33222         (vect_make_slp_decision): Likewise.
33223         (vect_supported_load_permutation_p): Likewise, and polynomial
33224         vf too.
33225         (vect_analyze_slp_cost): Handle polynomial vf.
33226         (vect_slp_analyze_node_operations): Likewise.
33227         (vect_slp_analyze_bb_1): Likewise.
33228         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
33229         than an unsigned HOST_WIDE_INT.
33230         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
33231         (vectorizable_load): Handle polynomial vf.
33232         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
33233         a poly_uint64.
33234         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
33236 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33237             Alan Hayward  <alan.hayward@arm.com>
33238             David Sherwood  <david.sherwood@arm.com>
33240         * match.pd: Handle bit operations involving three constants
33241         and try to fold one pair.
33243 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
33245         * tree-vect-loop-manip.c: Include gimple-fold.h.
33246         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
33247         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
33248         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
33249         Add a path that uses a step of VF instead of 1, but disable it
33250         for now.
33251         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
33252         and niters_no_overflow parameters.  Update calls to
33253         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
33254         Create a new SSA name if the latter choses to use a ste other
33255         than zero, and return it via niters_vector_mult_vf_var.
33256         * tree-vect-loop.c (vect_transform_loop): Update calls to
33257         vect_do_peeling, vect_gen_vector_loop_niters and
33258         slpeel_make_loop_iterate_ntimes.
33259         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
33260         (vect_gen_vector_loop_niters): Update declarations after above changes.
33262 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
33264         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
33265         128-bit round to integer instructions.
33266         (ceil<mode>2): Likewise.
33267         (btrunc<mode>2): Likewise.
33268         (round<mode>2): Likewise.
33270 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
33272         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
33273         unaligned VSX load/store on P8/P9.
33274         (expand_block_clear): Allow the use of unaligned VSX
33275         load/store on P8/P9.
33277 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
33279         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
33280         New function.
33281         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
33282         swap associated with both a load and a store.
33284 2018-01-02  Andrew Waterman  <andrew@sifive.com>
33286         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
33287         * config/riscv/riscv.md (clear_cache): Use it.
33289 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
33291         * web.c: Remove out-of-date comment.
33293 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33295         * expr.c (fixup_args_size_notes): Check that any existing
33296         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
33297         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
33298         (emit_single_push_insn): ...here.
33300 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33302         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
33303         (const_vector_encoded_nelts): New function.
33304         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
33305         (const_vector_int_elt, const_vector_elt): Declare.
33306         * emit-rtl.c (const_vector_int_elt_1): New function.
33307         (const_vector_elt): Likewise.
33308         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
33309         of CONST_VECTOR_ELT.
33311 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33313         * expr.c: Include rtx-vector-builder.h.
33314         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
33315         directly on the tree encoding.
33316         (const_vector_from_tree): Likewise.
33317         * optabs.c: Include rtx-vector-builder.h.
33318         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
33319         sequence of "u" values.
33320         * vec-perm-indices.c: Include rtx-vector-builder.h.
33321         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
33322         directly on the vec_perm_indices encoding.
33324 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33326         * doc/rtl.texi (const_vector): Describe new encoding scheme.
33327         * Makefile.in (OBJS): Add rtx-vector-builder.o.
33328         * rtx-vector-builder.h: New file.
33329         * rtx-vector-builder.c: Likewise.
33330         * rtl.h (rtx_def::u2): Add a const_vector field.
33331         (CONST_VECTOR_NPATTERNS): New macro.
33332         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
33333         (CONST_VECTOR_DUPLICATE_P): Likewise.
33334         (CONST_VECTOR_STEPPED_P): Likewise.
33335         (CONST_VECTOR_ENCODED_ELT): Likewise.
33336         (const_vec_duplicate_p): Check for a duplicated vector encoding.
33337         (unwrap_const_vec_duplicate): Likewise.
33338         (const_vec_series_p): Check for a non-duplicated vector encoding.
33339         Say that the function only returns true for integer vectors.
33340         * emit-rtl.c: Include rtx-vector-builder.h.
33341         (gen_const_vec_duplicate_1): Delete.
33342         (gen_const_vector): Call gen_const_vec_duplicate instead of
33343         gen_const_vec_duplicate_1.
33344         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
33345         (gen_const_vec_duplicate): Use rtx_vector_builder.
33346         (gen_const_vec_series): Likewise.
33347         (gen_rtx_CONST_VECTOR): Likewise.
33348         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
33349         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
33350         Build a new vector rather than modifying a CONST_VECTOR in-place.
33351         (handle_special_swappables): Update call accordingly.
33352         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
33353         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
33354         Build a new vector rather than modifying a CONST_VECTOR in-place.
33355         (handle_special_swappables): Update call accordingly.
33357 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33359         * simplify-rtx.c (simplify_const_binary_operation): Use
33360         CONST_VECTOR_ELT instead of XVECEXP.
33362 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33364         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
33365         the selector elements to be different from the data elements
33366         if the selector is a VECTOR_CST.
33367         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
33368         ssizetype for the selector.
33370 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33372         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
33373         before testing each element individually.
33374         * tree-vect-generic.c (lower_vec_perm): Likewise.
33376 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33378         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
33379         * selftest-run-tests.c (selftest::run_tests): Call it.
33380         * vector-builder.h (vector_builder::operator ==): New function.
33381         (vector_builder::operator !=): Likewise.
33382         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
33383         (vec_perm_indices::all_from_input_p): New function.
33384         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
33385         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
33386         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
33387         instead of reading the VECTOR_CST directly.  Detect whether both
33388         vector inputs are the same before constructing the vec_perm_indices,
33389         and update the number of inputs argument accordingly.  Use the
33390         utility functions added above.  Only construct sel2 if we need to.
33392 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33394         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
33395         the broadcast of the low byte.
33396         (expand_mult_highpart): Use an explicit encoding for the permutes.
33397         * optabs-query.c (can_mult_highpart_p): Likewise.
33398         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
33399         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
33400         (vectorizable_bswap): Likewise.
33401         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
33402         explicit encoding for the power-of-2 permutes.
33403         (vect_permute_store_chain): Likewise.
33404         (vect_grouped_load_supported): Likewise.
33405         (vect_permute_load_chain): Likewise.
33407 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33409         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
33410         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
33411         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
33412         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
33413         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
33414         (vect_gen_perm_mask_any): Likewise.
33416 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33418         * int-vector-builder.h: New file.
33419         * vec-perm-indices.h: Include int-vector-builder.h.
33420         (vec_perm_indices): Redefine as an int_vector_builder.
33421         (auto_vec_perm_indices): Delete.
33422         (vec_perm_builder): Redefine as a stand-alone class.
33423         (vec_perm_indices::vec_perm_indices): New function.
33424         (vec_perm_indices::clamp): Likewise.
33425         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
33426         (vec_perm_indices::new_vector): New function.
33427         (vec_perm_indices::new_expanded_vector): Update for new
33428         vec_perm_indices class.
33429         (vec_perm_indices::rotate_inputs): New function.
33430         (vec_perm_indices::all_in_range_p): Operate directly on the
33431         encoded form, without computing elided elements.
33432         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
33433         encoding.  Update for new vec_perm_indices class.
33434         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
33435         the given vec_perm_builder.
33436         (expand_vec_perm_var): Update vec_perm_builder constructor.
33437         (expand_mult_highpart): Use vec_perm_builder instead of
33438         auto_vec_perm_indices.
33439         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
33440         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
33441         or double series encoding as appropriate.
33442         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
33443         vec_perm_indices instead of auto_vec_perm_indices.
33444         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
33445         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
33446         (vect_permute_store_chain): Likewise.
33447         (vect_grouped_load_supported): Likewise.
33448         (vect_permute_load_chain): Likewise.
33449         (vect_shift_permute_load_chain): Likewise.
33450         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
33451         (vect_transform_slp_perm_load): Likewise.
33452         (vect_schedule_slp_instance): Likewise.
33453         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
33454         (vectorizable_mask_load_store): Likewise.
33455         (vectorizable_bswap): Likewise.
33456         (vectorizable_store): Likewise.
33457         (vectorizable_load): Likewise.
33458         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
33459         vec_perm_indices instead of auto_vec_perm_indices.  Use
33460         tree_to_vec_perm_builder to read the vector from a tree.
33461         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
33462         vec_perm_builder instead of a vec_perm_indices.
33463         (have_whole_vector_shift): Use vec_perm_builder and
33464         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
33465         truncation to calc_vec_perm_mask_for_shift.
33466         (vect_create_epilog_for_reduction): Likewise.
33467         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
33468         from auto_vec_perm_indices to vec_perm_indices.
33469         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
33470         instead of changing individual elements.
33471         (aarch64_vectorize_vec_perm_const): Use new_vector to install
33472         the vector in d.perm.
33473         * config/arm/arm.c (expand_vec_perm_d::perm): Change
33474         from auto_vec_perm_indices to vec_perm_indices.
33475         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
33476         instead of changing individual elements.
33477         (arm_vectorize_vec_perm_const): Use new_vector to install
33478         the vector in d.perm.
33479         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
33480         Update vec_perm_builder constructor.
33481         (rs6000_expand_interleave): Likewise.
33482         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
33483         (rs6000_expand_interleave): Likewise.
33485 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33487         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
33488         to qimode could truncate the indices.
33489         * optabs.c (expand_vec_perm_var): Likewise.
33491 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33493         * Makefile.in (OBJS): Add vec-perm-indices.o.
33494         * vec-perm-indices.h: New file.
33495         * vec-perm-indices.c: Likewise.
33496         * target.h (vec_perm_indices): Replace with a forward class
33497         declaration.
33498         (auto_vec_perm_indices): Move to vec-perm-indices.h.
33499         * optabs.h: Include vec-perm-indices.h.
33500         (expand_vec_perm): Delete.
33501         (selector_fits_mode_p, expand_vec_perm_var): Declare.
33502         (expand_vec_perm_const): Declare.
33503         * target.def (vec_perm_const_ok): Replace with...
33504         (vec_perm_const): ...this new hook.
33505         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
33506         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
33507         * doc/tm.texi: Regenerate.
33508         * optabs.def (vec_perm_const): Delete.
33509         * doc/md.texi (vec_perm_const): Likewise.
33510         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
33511         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
33512         expand_vec_perm for constant permutation vectors.  Assert that
33513         the mode of variable permutation vectors is the integer equivalent
33514         of the mode that is being permuted.
33515         * optabs-query.h (selector_fits_mode_p): Declare.
33516         * optabs-query.c: Include vec-perm-indices.h.
33517         (selector_fits_mode_p): New function.
33518         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
33519         is defined, instead of checking whether the vec_perm_const_optab
33520         exists.  Use targetm.vectorize.vec_perm_const instead of
33521         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
33522         fit in the vector mode before using a variable permute.
33523         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
33524         vec_perm_indices instead of an rtx.
33525         (expand_vec_perm): Replace with...
33526         (expand_vec_perm_const): ...this new function.  Take the selector
33527         as a vec_perm_indices rather than an rtx.  Also take the mode of
33528         the selector.  Update call to shift_amt_for_vec_perm_mask.
33529         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
33530         Use vec_perm_indices::new_expanded_vector to expand the original
33531         selector into bytes.  Check whether the indices fit in the vector
33532         mode before using a variable permute.
33533         (expand_vec_perm_var): Make global.
33534         (expand_mult_highpart): Use expand_vec_perm_const.
33535         * fold-const.c: Includes vec-perm-indices.h.
33536         * tree-ssa-forwprop.c: Likewise.
33537         * tree-vect-data-refs.c: Likewise.
33538         * tree-vect-generic.c: Likewise.
33539         * tree-vect-loop.c: Likewise.
33540         * tree-vect-slp.c: Likewise.
33541         * tree-vect-stmts.c: Likewise.
33542         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
33543         Delete.
33544         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
33545         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
33546         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
33547         (aarch64_vectorize_vec_perm_const): ...this new function.
33548         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
33549         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33550         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
33551         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
33552         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
33553         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33554         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
33555         into...
33556         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
33557         check for NEON modes.
33558         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
33559         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
33560         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
33561         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
33562         into...
33563         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
33564         the old VEC_PERM_CONST conditions.
33565         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
33566         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
33567         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
33568         (ia64_vectorize_vec_perm_const_ok): Merge into...
33569         (ia64_vectorize_vec_perm_const): ...this new function.
33570         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
33571         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
33572         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
33573         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
33574         * config/mips/mips.c (mips_expand_vec_perm_const)
33575         (mips_vectorize_vec_perm_const_ok): Merge into...
33576         (mips_vectorize_vec_perm_const): ...this new function.
33577         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
33578         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
33579         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
33580         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
33581         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
33582         (rs6000_expand_vec_perm_const): Delete.
33583         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
33584         Delete.
33585         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33586         (altivec_expand_vec_perm_const_le): Take each operand individually.
33587         Operate on constant selectors rather than rtxes.
33588         (altivec_expand_vec_perm_const): Likewise.  Update call to
33589         altivec_expand_vec_perm_const_le.
33590         (rs6000_expand_vec_perm_const): Delete.
33591         (rs6000_vectorize_vec_perm_const_ok): Delete.
33592         (rs6000_vectorize_vec_perm_const): New function.
33593         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
33594         an element count and rtx array.
33595         (rs6000_expand_extract_even): Update call accordingly.
33596         (rs6000_expand_interleave): Likewise.
33597         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
33598         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
33599         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
33600         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
33601         (rs6000_expand_vec_perm_const): Delete.
33602         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
33603         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33604         (altivec_expand_vec_perm_const_le): Take each operand individually.
33605         Operate on constant selectors rather than rtxes.
33606         (altivec_expand_vec_perm_const): Likewise.  Update call to
33607         altivec_expand_vec_perm_const_le.
33608         (rs6000_expand_vec_perm_const): Delete.
33609         (rs6000_vectorize_vec_perm_const_ok): Delete.
33610         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
33611         reference to the SPE evmerge intructions.
33612         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
33613         an element count and rtx array.
33614         (rs6000_expand_extract_even): Update call accordingly.
33615         (rs6000_expand_interleave): Likewise.
33616         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
33617         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
33618         new function.
33619         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33621 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33623         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
33624         vector mode and that that mode matches the mode of the data
33625         being permuted.
33626         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
33627         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
33628         directly using expand_vec_perm_1 when forcing selectors into
33629         registers.
33630         (expand_vec_perm_var): New function, split out from expand_vec_perm.
33632 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33634         * optabs-query.h (can_vec_perm_p): Delete.
33635         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
33636         * optabs-query.c (can_vec_perm_p): Split into...
33637         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
33638         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
33639         particular selector is valid.
33640         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
33641         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
33642         (vect_grouped_load_supported): Likewise.
33643         (vect_shift_permute_load_chain): Likewise.
33644         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
33645         (vect_transform_slp_perm_load): Likewise.
33646         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
33647         (vectorizable_bswap): Likewise.
33648         (vect_gen_perm_mask_checked): Likewise.
33649         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
33650         implementations of variable permutation vectors into account
33651         when deciding which selector to use.
33652         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
33653         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
33654         with a false third argument.
33655         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
33656         to test whether the constant selector is valid and can_vec_perm_var_p
33657         to test whether a variable selector is valid.
33659 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33661         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
33662         * optabs-query.c (can_vec_perm_p): Likewise.
33663         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
33664         instead of vec_perm_indices.
33665         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
33666         (vect_gen_perm_mask_checked): Likewise,
33667         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
33668         (vect_gen_perm_mask_checked): Likewise,
33670 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
33672         * optabs-query.h (qimode_for_vec_perm): Declare.
33673         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
33674         (qimode_for_vec_perm): ...this new function.
33675         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
33677 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
33679         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
33680         does not have a conditional at the top.
33682 2018-01-02  Richard Biener  <rguenther@suse.de>
33684         * ipa-inline.c (big_speedup_p): Fix expression.
33686 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
33688         PR target/81616
33689         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
33690         for generic 4->6.
33692 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
33694         PR target/81616
33695         Generic tuning.
33696         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
33697         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
33698         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
33699         cond_taken_branch_cost 3->4.
33701 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
33703         PR tree-optimization/83581
33704         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
33705         TODO_cleanup_cfg if any changes have been made.
33707         PR middle-end/83608
33708         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
33709         convert_modes if target mode has the right side, but different mode
33710         class.
33712         PR middle-end/83609
33713         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
33714         last argument when extracting from CONCAT.  If either from_real or
33715         from_imag is NULL, use expansion through memory.  If result is not
33716         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
33717         the parts directly to inner mode, if even that fails, use expansion
33718         through memory.
33720         PR middle-end/83623
33721         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
33722         check for bswap in mode rather than HImode and use that in expand_unop
33723         too.
33725 Copyright (C) 2018 Free Software Foundation, Inc.
33727 Copying and distribution of this file, with or without modification,
33728 are permitted in any medium without royalty provided the copyright
33729 notice and this notice are preserved.