1 2018-11-29 Martin Liska <mliska@suse.cz>
4 * builtins.c (expand_movstr): Fix thinko introduced
5 when switching to the new enum.
7 2018-11-29 qing zhao <qing.zhao@oracle.com>
9 * cif-code.def (EXTERN_LIVE_ONLY_STATIC): New CIF code.
10 * common.opt: Add -flive-patching flag.
11 * doc/invoke.texi: Document -flive-patching.
12 * flag-types.h (enum live_patching_level): New enum.
13 * ipa-inline.c (can_inline_edge_p): Disable external functions from
14 inlining when flag_live_patching is LIVE_PATCHING_INLINE_ONLY_STATIC.
15 * opts.c (control_options_for_live_patching): New function.
16 (finish_options): Make flag_live_patching incompatible with flag_lto.
17 Control IPA optimizations based on different levels of
20 2018-11-29 Giuliano Belinassi <giuliano.belinassi@usp.br>
22 * match.pd (sinh (atanh (x))): New simplification rules.
23 (cosh (atanh (x))): Likewise.
25 2018-11-29 Jakub Jelinek <jakub@redhat.com>
28 * config/i386/sse.md (ssebytemode): Add V16SI, V8SI and V4SI entries.
29 (ssefltmodesuffix, ssefltvecmode): New define_mode_attrs.
30 (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
31 *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
32 *<sse4_1_avx2>_pblendvb_lt): New define_insns.
35 * tree.h (build_uniform_cst, uniform_integer_cst_p): Declare.
36 * tree.c (build_uniform_cst, uniform_integer_cst_p): New functions.
37 * match.pd (define_predicates): Add uniform_integer_cst_p.
38 (cmp @0 INTEGER_CST@1, cmp (convert?@2 @0) INTEGER_CST@1): Adjust
39 so that it works also for vector comparisons with uniform constants
40 with INTEGER_CST element.
43 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): For
44 vec_add and vec_sub builtins, perform PLUS_EXPR or MINUS_EXPR
45 in unsigned_type_for instead of vector integral type where overflow
48 2018-11-29 Michael Ploujnikov <michael.ploujnikov@oracle.com>
50 There can be at most one .resolver clone per function
51 * config/rs6000/rs6000.c (make_resolver_func): Generate
52 resolver symbol with clone_function_name instead of
53 clone_function_name_numbered.
55 2018-11-29 Richard Biener <rguenther@suse.de>
57 PR tree-optimization/88243
58 * tree-vect-patterns.c (vect_mark_pattern_stmts): Set the def
59 type of all pattern-sequence stmts to vect_internal_def.
61 2018-11-29 Andre Vieira <andre.simoesdiasvieira@arm.com>
64 * config/arm/arm-cpus.in (armv7-r): Add FP16conv configurations.
65 (cortex-r7, cortex-r8): Update default and add new configuration.
66 * doc/invoke.texi (armv7-r): Add two new vfp options.
67 (nofp.dp): Add cortex-r7 and cortex-r8 to the list of targets that
70 2018-11-29 Alan Modra <amodra@gmail.com>
72 * config/rs6000/rs6000.c (rs6000_emit_move): Disable long
73 double split for targets other than Darwin.
75 2018-11-29 Alan Modra <amodra@gmail.com>
77 * config/rs6000/rs6000.md (movdi_internal64): Correct lengths.
79 2018-11-29 Alan Modra <amodra@gmail.com>
81 * config/rs6000/predicates.md (easy_fp_constant): Remove code
82 dealing with integers in const_double. Assert on unexpected
83 modes. Delete superfluous ABI_V4 test.
84 * config/rs6000/rs6000.c (rs6000_emit_move): Comment fixes.
85 Don't call easy_fp_constant for Pmode.
86 (rs6000_hash_constant): Remove code dealing with integers in
88 (rs6000_legitimate_constant_p): Likewise.
89 (output_toc): Formatting, use CONST_DOUBLE_P.
91 2018-11-29 Alan Modra <amodra@gmail.com>
93 * config/rs6000/constraints.md (G, H): Comment on purpose of
94 constraint. Correct mode comments and "G" mode test.
95 * config/rs6000/rs6000.md (movdi_internal32): Remove "GH" from
96 alternative handling "F". Add length attr. Formatting.
97 (mov<mode>_softfloat <FMOVE128>): Delete "GH" from
98 alternative, and move "F" to separate alternative. Correct
101 2018-11-29 Alan Modra <amodra@gmail.com>
103 * config/rs6000/predicates.md (easy_vector_constant): Don't call
106 2018-11-29 Alan Modra <amodra@gmail.com>
108 * config.in (HAVE_AS_PLTSEQ): Add.
109 * config/rs6000/predicates.md (indirect_call_operand): New.
110 * config/rs6000/rs6000-protos.h (rs6000_pltseq_template),
111 (rs6000_sibcall_sysv): Declare.
112 * config/rs6000/rs6000.c (init_cumulative_args): Set cookie
113 CALL_LONG for -fno-plt.
114 (print_operand <T, z, 0>): Handle UNSPEC_PLTSEQ.
115 (rs6000_indirect_call_template_1): Emit .reloc directives for
117 (rs6000_pltseq_template): New function.
118 (rs6000_longcall_ref): Add arg parameter. Use PLT16 insns if
119 relocs supported by assembler. Move SYMBOL_REF test to callers.
120 (rs6000_call_aix): Adjust rs6000_longcall_ref call. Package
121 insns in UNSPEC_PLTSEQ, preserving original func_desc.
122 (rs6000_call_sysv): Likewise.
123 (rs6000_sibcall_sysv): New function.
124 * config/rs6000/rs6000.h (HAVE_AS_PLTSEQ): Provide default.
125 * config/rs6000/rs6000.md (UNSPEC_PLTSEQ, UNSPEC_PLT16_HA,
126 UNSPEC_PLT16_LO): New.
127 (pltseq_tocsave, pltseq_plt16_ha, pltseq_plt16_lo, pltseq_mtctr): New.
128 (call_indirect_nonlocal_sysv): Don't differentiate zero from non-zero
129 cookie in constraints. Test explicitly for flags in length attr.
130 Handle unspec operand 1.
131 (call_value_indirect_nonlocal_sysv): Likewise.
132 (call_indirect_aix, call_value_indirect_aix): Handle unspec operand 1.
133 (call_indirect_elfv2, call_value_indirect_elfv2): Likewise.
134 (sibcall, sibcall_value): Use rs6000_sibcall_sysv.
135 (sibcall_indirect_nonlocal_sysv): New pattern.
136 (sibcall_value_indirect_nonlocal_sysv): Likewise.
137 (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv): Remove indirect
139 * configure.ac: Check for gas plt sequence marker support.
140 * configure: Regenerate.
142 2018-11-29 Alan Modra <amodra@gmail.com>
144 * config/rs6000/predicates.md (unspec_tls): New.
145 * config/rs6000/rs6000-protos.h (rs6000_call_template),
146 (rs6000_sibcall_template): Update prototype.
147 (rs6000_longcall_ref): Delete.
148 (rs6000_call_sysv): Declare.
149 * config/rs6000/rs6000.c (edit_tls_call_insn): New function.
150 (global_tlsarg): New variable.
151 (rs6000_legitimize_tls_address): Rewrite __tls_get_addr call
153 (print_operand): Extract UNSPEC_TLSGD address operand.
154 (rs6000_call_template, rs6000_sibcall_template): Remove arg
155 parameter, extract from second call operand instead.
156 (rs6000_longcall_ref): Make static, localize vars.
157 (rs6000_call_aix): Rename parameter to reflect new usage. Take
158 tlsarg from global_tlsarg. Don't create unused rtl or nop insns.
159 (rs6000_sibcall_aix): Rename parameter to reflect new usage. Take
160 tlsarg from global_tlsarg.
161 (rs6000_call_sysv): New function.
162 * config/rs6000/rs6000.md: Adjust rs6000_call_template and
163 rs6000_sibcall_template throughout.
164 (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix, tls_gd_call_sysv): Delete.
165 (tls_ld_aix, tls_ld_sysv, tls_ld_call_aix, tls_ld_call_sysv): Delete.
166 (tls_gdld_nomark): New insn.
167 (tls_gd): Swap operand order. Simplify mode selection.
168 (tls_gd_high, tls_gd_low): Swap operand order.
169 (tls_ld): Remove const_int 0 vector element from UNSPEC_TLSLD.
170 Simplify mode selection.
171 (tls_ld_high, tls_ld_low): Similarly adjust UNSPEC_TLSLD.
172 (call, call_value): Don't assert for second call operand.
173 Use rs6000_call_sysv.
175 2018-11-29 Alan Modra <amodra@gmail.com>
177 * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64),
178 (call_nonlocal_darwin64, call_value_indirect_nonlocal_darwin64),
179 (call_value_nonlocal_darwin64): Remove constraints from second call
180 arg, the rounded_stack_size_rtx arg.
181 * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix),
182 (tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv, tls_ld_call_aix),
183 (tls_ld_call_sysv, call_local32, call_local64, call_value_local32),
184 (call_value_local64, call_indirect_nonlocal_sysv),
185 (call_nonlocal_sysv, call_nonlocal_sysv_secure),
186 (call_value_indirect_nonlocal_sysv, call_value_nonlocal_sysv),
187 (call_value_nonlocal_sysv_secure, call_local_aix),
188 (call_value_local_aix, call_nonlocal_aix, call_value_nonlocal_aix),
189 (call_indirect_aix, call_value_indirect_aix, call_indirect_elfv2),
190 (call_value_indirect_elfv2, sibcall_local32, sibcall_local64),
191 (sibcall_value_local32, sibcall_value_local64, sibcall_aix),
192 (sibcall_value_aix): Likewise.
194 2018-11-29 Alan Modra <amodra@gmail.com>
196 * config/rs6000/rs6000.md (TLSmode): Delete mode iterator. Replace
197 with P throughout except for call mems which should use SI.
198 (tls_abi_suffix, tls_sysv_suffix, tls_insn_suffix): Delete mode
199 attributes. Replace with bits, mode and ptrload respectively.
201 2018-11-29 Alan Modra <amodra@gmail.com>
203 * config/rs6000/rs6000-protos.h (rs6000_indirect_call_template),
204 (rs6000_indirect_sibcall_template): Declare.
205 * config/rs6000/rs6000.c (rs6000_indirect_call_template_1),
206 (rs6000_indirect_call_template, rs6000_indirect_sibcall_template):
208 * config/rs6000/rs6000.md (call_indirect_nonlocal_sysv),
209 (call_value_indirect_nonlocal_sysv, sibcall_nonlocal_sysv),
210 (call_indirect_aix, call_value_indirect_aix): Use
211 rs6000_indirect_call_template and rs6000_indirect_sibcall_template.
212 call_indirect_elfv2, call_value_indirect_elfv2): Likewise, and
213 handle both speculation and non-speculation cases.
214 (call_indirect_aix_nospec, call_value_indirect_aix_nospec): Delete.
215 (call_indirect_elfv2_nospec, call_value_indirect_elfv2_nospec): Delete.
217 2018-11-29 Alan Modra <amodra@gmail.com>
219 * config/rs6000/rs6000-protos.h (rs6000_call_template): Declare.
220 (rs6000_sibcall_template): Declare.
221 (macho_call_template): Rename from output_call.
222 * config/rs6000/rs6000.c (rs6000_call_template_1): New function.
223 (rs6000_call_template, rs6000_sibcall_template): Likewise.
224 (macho_call_template): Rename from output_call.
225 * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv),
226 (tls_gd_call_aix, tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv),
227 (tls_ld_call_aix, tls_ld_call_sysv, call_nonlocal_sysv),
228 (call_nonlocal_sysv_secure, call_value_nonlocal_sysv),
229 (call_value_nonlocal_sysv_secure, call_nonlocal_aix),
230 (call_value_nonlocal_aix): Use rs6000_call_template and update
231 occurrences of output_call to macho_call_template.
232 (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv, sibcall_aix),
233 (sibcall_value_aix): Use rs6000_sibcall_template.
235 2018-11-28 Aaron Sawdey <acsawdey@linux.ibm.com>
237 * config/rs6000/rs6000-string.c (expand_block_clear): Change how
238 we determine if unaligned vsx is ok.
240 2018-11-28 Jan Hubicka <jh@suse.cz>
242 * predict.c (determine_unlikely_bbs): Forward declare; also determine
243 edges that are always known to be taken because it is only likely
244 edge out of given BB.
245 (tree_estimate_probability): Call before profile guessing to get bit
248 2018-11-28 Jan Hubicka <jh@suse.cz>
250 * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Handle
251 profile_probability::always better.
253 2018-11-28 Jan Hubicka <jh@suse.cz>
255 * profile-count.h (profile_count::split): Give better result when
256 splitting profile_probability::always.
258 2018-11-28 Vladimir Makarov <vmakarov@redhat.com>
261 * ira-costs.c (scan_one_insn): Process subregs when updating costs
262 for pseudos and allocnos from insn.
264 2018-11-28 David Edelsohn <dje.gcc@gmail.com>
266 * config/rs6000/aix72.h: Update to match aix71.h changes.
268 2018-11-28 Jeff Law <law@redhat.com>
270 * config/h8300/constraints.md: Add "C" constraint for call insns.
271 * config/h8300/h8300.md (call, call_value): Turn into a define_expand
272 and define_insn pair. Move invalid call targets into a register in
273 the expander and fix constraints in the matching pattern.
274 * config/h8300/predicates.md (call_expander_operand): Renamed from
275 call_insn_operand. Reject things we shouldn't be trying to handle.
276 (call_insn_operand): New predicate for use by the call/call_value
278 (small_call_insn_operand): Update appropriately.
280 2018-11-28 Sam Tebbs <sam.tebbs@arm.com>
282 * config/aarch64/aarch64.c (aarch64_process_target_attr): Replace
283 calls to strtok with strtok_r.
285 2018-11-28 Richard Biener <rguenther@suse.de>
287 PR tree-optimization/88223
288 * tree-ssa-sccvn.c (vn_reference_lookup_3): When skipping
289 over a stored-same value may-alias store make sure to consider
290 partial overlaps which are valid when TBAA reasonings do not
291 apply and byte-granular overlaps are possible at all.
293 2018-11-28 Richard Biener <rguenther@suse.de>
295 PR tree-optimization/88217
296 * vr-values.c (vr_values::extract_range_from_phi_node): Make
297 sure to handle results > +INF and < -INF correctly when
298 trying to drop down to +INF - 1 or -INF + 1.
300 2018-11-28 Alan Modra <amodra@gmail.com>
302 * xcoffout.c (do_block): Signed/unsigned warning fix.
304 2018-11-28 Richard Biener <rguenther@suse.de>
306 PR tree-optimization/79351
307 * tree-ssa-sccvn.c (vn_reference_lookup_3): For assignments from
308 empty CONSTRUCTORs ensure the store is at a constant position.
310 2018-11-28 Richard Biener <rguenther@suse.de>
312 PR tree-optimization/88229
313 * tree-ssa.c (non_rewritable_mem_ref_base): Check DECL_SIZE_UNIT
314 is an INTEGER_CST before accessing it so.
316 2018-11-28 Sam Tebbs <sam.tebbs@arm.com>
318 * doc/invoke.texi (-mtune=): Remove obsolete CPU names.
320 2018-11-28 Jakub Jelinek <jakub@redhat.com>
323 * config/i386/i386.c (ix86_expand_sse_movcc): Handle DFmode and
324 SFmode using sse4_1_blendvs[sd] with TARGET_SSE4_1. Formatting fixes.
325 * config/i386/sse.md (sse4_1_blendv<ssemodesuffix>): New pattern.
327 2018-11-27 Martin Liska <mliska@suse.cz>
329 * gcov.c (generate_results): Append current_working_directory
332 2018-11-27 Mike Gulick <mgulick@mathworks.com>
334 PR preprocessor/83173
335 * input.c (dump_location_info): Dump reason and included_from
336 fields from line_map_ordinary struct. Fix indentation when
338 * diagnostic-show-locus.c (num_digits, num_digits): Move to
339 diagnostic.c to allow it to be utilized by input.c.
340 * diagnostic.c (num_digits, selftest::test_num_digits): Moved
342 (selftest::diagnostic_c_tests): Run selftest::test_num_digits.
343 * diagnostic.h (num_digits): Add extern definition.
345 2018-11-27 Fredrik Noring <noring@nocrew.org>
347 * config/mips/mips.c (mips_reorg_process_insns)
348 (mips_option_override): Handle `-mfix-r5900'.
349 * config/mips/mips.h (ASM_SPEC): Add `mfix-r5900' and
351 * config/mips/mips.opt (mfix-r5900): New option.
352 * doc/invoke.texi: Document the `r5900' processor name, and
353 `-mfix-r5900' and `-mno-fix-r5900' options.
355 2018-11-27 Jakub Jelinek <jakub@redhat.com>
358 * config/rs6000/rs6000.c (print_operand) <case 'D'>: Use
359 output_operand_lossage instead of gcc_assert.
360 <case 't'>: Likewise.
361 <case 'z'>: Likewise.
362 <case 'V'>: Use output_operand_lossage instead of gcc_unreachable.
364 2018-11-27 Jeff Law <law@redhat.com>
366 * config/riscv/riscv (riscv_block_mvoe_straight): Use RETURN_BEGIN
367 in call to move_by_pieces.
369 * config/sh/sh-mem.c (expand_block_move): Use RETURN_BEGIN in call
372 * config/lm32/lm32.c (lm32_block_move_inline): Use RETURN_BEGIN in
373 call to move_by_pieces.
375 * config/mips/mips.c (mips_block_move_straight): Use RETURN_BEGIN
376 in call to move_by_pieces.
378 * config/microblaze/microblaze.c (microblaze_block_move_straight): Use
379 RETURN_BEGIN in call to move_by_pieces.
380 (microblaze_expand_block_move): Likewise.
382 2018-11-27 Tamar Christina <tamar.christina@arm.com>
384 * doc/md.texi (xorsign): Document it.
386 2018-11-27 Martin Liska <mliska@suse.cz>
388 * config/gnu-user.h (TARGET_F951_OPTIONS): New.
389 * gcc.c (find_fortran_preinclude_file): New function
390 to handle Fortran pre-include.
392 2018-11-27 Martin Liska <mliska@suse.cz>
394 * asan.c (asan_emit_stack_protection): Use new enum values
395 instead of int constants.
396 * builtins.c (expand_builtin_memory_copy_args): Replace int
397 type with memop_ret enum type.
398 (expand_builtin_mempcpy_args): Likewise.
399 (expand_builtin_memcpy): Use new enum values
400 instead of int constants. Likewise.
401 (expand_builtin_mempcpy): Likewise.
402 (expand_movstr): Likewise.
403 (expand_builtin_strcpy_args): Likewise.
404 (expand_builtin_stpcpy_1): Likewise.
405 (expand_builtin_strncpy): Likewise.
406 (expand_builtin_memset_args): Likewise.
407 * expr.c (move_by_pieces_d::finish_endp): Rename to ...
408 (move_by_pieces_d::finish_retmode): ... this.
409 (move_by_pieces): Change last argument type to memop_ret.
410 (store_by_pieces): Use new enum values
411 instead of int constants.
412 (emit_block_move_hints): Likewise.
413 (emit_push_insn): Likewise.
414 (store_expr): Likewise.
415 * expr.h (store_by_pieces): Change int to newly added enum
417 * rtl.h (enum memop_ret): Define.
418 (move_by_pieces): Use the enum type.
420 2018-11-27 Alan Modra <amodra@gmail.com>
422 * config/rs6000/aix71.h (ASM_SPEC): Don't select default -maix64
424 (ASM_CPU_SPEC): Do so here. Rewrite using if .. else if .. specs
425 form. Error on missing -mcpu case.
426 * config/rs6000/driver-rs6000.c (asm_names <_AIX>): Update NULL case.
427 (asm_names <!_AIX>): Add missing cpus. Update NULL case. Apply
428 PR63177 fix for -mcpu=power8 and -mcpu=powerpc64le.
429 * config/rs6000/rs6000.h (ASM_CPU_SPEC): Rewrite using if ..
430 else if .. specs form. Error on missing -mcpu case. Don't output
431 duplicate -maltivec. Apply PR63177 fix for -mcpu=powerpc64le.
433 2018-11-26 David Malcolm <dmalcolm@redhat.com>
435 * dump-context.h (dump_context::dump_loc): Convert 1st param from
436 dump_flags_t to const dump_metadata_t &. Convert 2nd param from
437 const dump_location_t & to const dump_user_location_t &.
438 (dump_context::dump_loc_immediate): Convert 2nd param from
439 const dump_location_t & to const dump_user_location_t &.
440 (dump_context::dump_gimple_stmt): Convert 1st param from
441 dump_flags_t to const dump_metadata_t &.
442 (dump_context::void dump_gimple_stmt_loc): Likewise; convert
443 2nd param from const dump_location_t & to
444 const dump_user_location_t &.
445 (dump_context::dump_gimple_expr): Convert 1st param from
446 dump_flags_t to const dump_metadata_t &.
447 (dump_context::dump_gimple_expr_loc): Likewise; convert
448 2nd param from const dump_location_t & to
449 const dump_user_location_t &.
450 (dump_context::dump_generic_expr): Convert 1st param from
451 dump_flags_t to const dump_metadata_t &.
452 (dump_context::dump_generic_expr_loc): Likewise; convert
453 2nd param from const dump_location_t & to
454 const dump_user_location_t &.
455 (dump_context::dump_printf_va): Convert 1st param from
456 dump_flags_t to const dump_metadata_t &.
457 (dump_context::dump_printf_loc_va): Likewise; convert
458 2nd param from const dump_location_t & to
459 const dump_user_location_t &.
460 (dump_context::dump_dec): Convert 1st param from
461 dump_flags_t to const dump_metadata_t &.
462 (dump_context::dump_symtab_node): Likewise.
463 (dump_context::begin_scope): Split out 2nd param into
464 user and impl locations.
465 (dump_context::ensure_pending_optinfo): Add metadata param.
466 (dump_context::begin_next_optinfo): Replace dump_location_t param
467 with metadata and user location.
468 * dumpfile.c (dump_context::dump_loc): Convert 1st param from
469 dump_flags_t to const dump_metadata_t &. Convert 2nd param from
470 const dump_location_t & to const dump_user_location_t &.
471 (dump_context::dump_loc_immediate): Convert 2nd param from
472 const dump_location_t & to const dump_user_location_t &.
473 (dump_context::dump_gimple_stmt): Convert 1st param from
474 dump_flags_t to const dump_metadata_t &.
475 (dump_context::void dump_gimple_stmt_loc): Likewise; convert
476 2nd param from const dump_location_t & to
477 const dump_user_location_t &.
478 (dump_context::dump_gimple_expr): Convert 1st param from
479 dump_flags_t to const dump_metadata_t &.
480 (dump_context::dump_gimple_expr_loc): Likewise; convert
481 2nd param from const dump_location_t & to
482 const dump_user_location_t &.
483 (dump_context::dump_generic_expr): Convert 1st param from
484 dump_flags_t to const dump_metadata_t &.
485 (dump_context::dump_generic_expr_loc): Likewise; convert
486 2nd param from const dump_location_t & to
487 const dump_user_location_t &.
488 (dump_context::dump_printf_va): Convert 1st param from
489 dump_flags_t to const dump_metadata_t &.
490 (dump_context::dump_printf_loc_va): Likewise; convert
491 2nd param from const dump_location_t & to
492 const dump_user_location_t &.
493 (dump_context::dump_dec): Convert 1st param from
494 dump_flags_t to const dump_metadata_t &.
495 (dump_context::dump_symtab_node): Likewise.
496 (dump_context::begin_scope): Split out 2nd param into
497 user and impl locations.
498 (dump_context::ensure_pending_optinfo): Add metadata param.
499 (dump_context::begin_next_optinfo): Replace dump_location_t param
500 with metadata and user location.
501 (dump_gimple_stmt): Convert 1st param from dump_flags_t to
502 const dump_metadata_t &.
503 (dump_gimple_stmt_loc): Likewise; convert 2nd param from
504 const dump_location_t & to const dump_user_location_t &.
505 (dump_gimple_expr): Convert 1st param from dump_flags_t to
506 const dump_metadata_t &.
507 (dump_gimple_expr_loc): Likewise; convert 2nd param from
508 const dump_location_t & to const dump_user_location_t &.
509 (dump_generic_expr): Convert 1st param from dump_flags_t to
510 const dump_metadata_t &.
511 (dump_generic_expr_loc): Likewise; convert 2nd param from
512 const dump_location_t & to const dump_user_location_t &.
513 (dump_printf): Convert 1st param from dump_flags_t to
514 const dump_metadata_t &.
515 (dump_printf_loc): Likewise; convert 2nd param from
516 const dump_location_t & to const dump_user_location_t &.
517 (dump_dec): Convert 1st param from dump_flags_t to
518 const dump_metadata_t &.
519 (dump_symtab_node): Likewise.
520 (dump_begin_scope): Split out 2nd param into user and impl
522 (selftest::assert_impl_location_eq): New function.
523 (ASSERT_IMPL_LOCATION_EQ): New macro.
524 (selftest::test_impl_location): Update to use
525 ASSERT_IMPL_LOCATION_EQ.
526 (selftest::test_capture_of_dump_calls): Convert "loc" to
527 dump_user_location_t. Add ASSERT_IMPL_LOCATION_EQ throughout,
528 verifying line numbers of dump emissions.
529 * dumpfile.h (class dump_metadata_t): New class.
530 (dump_printf): Convert 1st param from dump_flags_t to
531 const dump_metadata_t &.
532 (dump_printf_loc): Likewise; convert 2nd param from
533 const dump_location_t & to const dump_user_location_t &.
534 (dump_generic_expr_loc): Likewise.
535 (dump_generic_expr): Convert 1st param from dump_flags_t to
536 const dump_metadata_t &.
537 (dump_gimple_stmt_loc): Likewise; convert 2nd param from
538 const dump_location_t & to const dump_user_location_t &.
539 (dump_gimple_stmt): Convert 1st param from dump_flags_t to
540 const dump_metadata_t &.
541 (dump_gimple_expr_loc): Likewise; convert 2nd param from
542 const dump_location_t & to const dump_user_location_t &.
543 (dump_gimple_expr): Convert 1st param from dump_flags_t to
544 const dump_metadata_t &.
545 (dump_symtab_node): Likewise.
546 (dump_dec): Likewise.
547 (dump_begin_scope): Split out 2nd param into user and impl
549 (auto_dump_scope::auto_dump_scope): Split "loc" param into a user
550 location and impl_location, and capture the impl_location.
551 (AUTO_DUMP_SCOPE): Rename param from LOC to USER_LOC.
552 * loop-unroll.c (report_unroll): Update for changes to
553 dump_printf_loc and dump_printf.
554 * opt-problem.cc (opt_problem::opt_problem): Update for change to
556 * optinfo-emit-json.cc
557 (selftest::test_building_json_from_dump_calls): Convert "loc" from
558 dump_location_t to dump_user_location_t.
559 * optinfo.cc (optinfo::emit_for_opt_problem): Update for change in
561 * profile.c (compute_branch_probabilities): Update for change to
563 * selftest.h (ASSERT_STR_CONTAINS_AT): New macro.
564 * tree-vect-slp.c (vect_print_slp_tree): Update for change to
567 2018-11-27 Jozef Lawrynowicz <jozef.l@mittosystems.com>
569 * doc/sourcebuild.texi: Document check_effective_target_int_eq_float
570 and check_effective_target_ptr_eq_long.
572 2018-11-27 Alan Modra <amodra@gmail.com>
574 * config.gcc (powerpc*-*-freebsd*, powerpc-*-netbsd*),
575 (powerpc-*-eabisimaltivec*, powerpc-*-eabisim*, powerpc-*-elf*),
576 (powerpc-*-eabialtivec*, powerpc-*-eabi*, powerpc-*-rtems*),
577 (powerpc-wrs-vxworks*, powerpc-*-lynxos*, powerpcle-*-elf*),
578 (powerpcle-*-eabisim*, powerpcle-*-eabi*): Add gnu-user.h to tm_file.
579 * config/rs6000/freebsd.h (CPLUSPLUS_CPP_SPEC),
580 (LINK_GCC_C_SEQUENCE_SPEC): Undef.
581 (ASM_APP_ON, ASM_APP_OFF): Don't define.
582 * config/rs6000/freebsd64.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
583 * config/rs6000/lynx.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
584 * config/rs6000/linux64.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
585 * config/rs6000/netbsd.h (CPLUSPLUS_CPP_SPEC),
586 (LINK_GCC_C_SEQUENCE_SPEC): Undef.
587 * config/rs6000/rtems.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
588 * config/rs6000/sysv4.h (GNU_USER_TARGET_CRTI): Redefine.
589 (GNU_USER_TARGET_CRTN): Redefine.
590 (CC1_SPEC): Use GNU_USER_TARGET_CC1_SPEC.
591 (LIB_LINUX_SPEC): Use GNU_USER_TARGET_LIB_SPEC.
592 (CRTOFFLOADBEGIN, CRTOFFLOADEND): Don't define.
593 (STARTFILE_LINUX_SPEC): Define as GNU_USER_TARGET_STARTFILE_SPEC.
594 (ENDFILE_LINUX_SPEC): Define as GNU_USER_TARGET_ENDFILE_SPEC.
595 (UCLIBC_DYNAMIC_LINKER, CHOOSE_DYNAMIC_LINKER): Don't define.
596 (LINK_EH_SPEC): Don't define.
598 2018-11-26 Martin Sebor <msebor@redhat.com>
601 * expr.c (string_constant): Handle top-level decls of all character
602 types and subobjects of narrow character type.
604 2018-11-27 Alan Modra <amodra@gmail.com>
606 * config.gcc (powerpc*-*-linux*): Add linux.h to tm_file.
607 * config/rs6000/linux.h (TARGET_OS_CPP_BUILTINS): Use
608 GNU_USER_TARGET_OS_CPP_BUILTINS.
609 (RS6000_ABI_NAME): Define.
610 * config/rs6000/linux64.h (TARGET_OS_CPP_BUILTINS): Use
611 GNU_USER_TARGET_OS_CPP_BUILTINS.
612 (MUSL_DYNAMIC_LINKER32): Undef before defining.
613 (UCLIBC_DYNAMIC_LINKER32, UCLIBC_DYNAMIC_LINKER64): Don't define.
614 (CHOOSE_DYNAMIC_LINKER): Don't define.
615 (GNU_USER_DYNAMIC_LINKER32, GNU_USER_DYNAMIC_LINKER64): Don't define.
616 * config/rs6000/sysv4.h (MUSL_DYNAMIC_LINKER): Undef before defining.
617 (CHOOSE_DYNAMIC_LINKER, GNU_USER_DYNAMIC_LINKER): Only define when
619 (CPP_OS_LINUX_SPEC): Remove defines and asserts handled by
620 TARGET_OS_CPP_BUILTINS.
622 2018-11-26 Jozef Lawrynowicz <jozef.l@mittosystems.com>
624 * doc/sourcebuild.texi: Document check_effective_target_newlib_nano_io.
626 2018-11-26 Jakub Jelinek <jakub@redhat.com>
629 * config/i386/i386.c (def_builtin2): If tcode == VOID_FTYPE_UINT64
630 and !TARGET_64BIT, return NULL_TREE.
633 * dwarf2out.c (secname_for_decl): For functions with
634 DECL_SECTION_NAME if in_cold_section_p, try to return
635 current_function_section's name if it is a named section.
637 2018-11-26 Maya Rashish <coypu@sdf.org>
640 * config.host (*-*-netbsd*): Use x-netbsd and host-netbsd.o.
641 * config/x-netbsd: New file.
642 * config/host-netbsd: New file.
644 2018-11-26 Jozef Lawrynowicz <jozef.l@mittosystems.com>
646 * doc/sourcebuild.texi: Document check_effective_target_size20plus.
647 Clarify documentation for check_effective_target_size32plus.
649 2018-11-26 Aaron Sawdey <acsawdey@linux.ibm.com>
651 * config/rs6000/rs6000-string.c (expand_cmp_vec_sequence): Rename
652 and modify expand_strncmp_vec_sequence.
653 (emit_final_compare_vec): Rename and modify emit_final_str_compare_vec.
654 (generate_6432_conversion): New function.
655 (expand_block_compare): Add support for vsx.
656 (expand_block_compare_gpr): New function.
657 * config/rs6000/rs6000.opt (rs6000_block_compare_inline_limit): Increase
658 default limit to 63 because of more compact vsx code.
660 2018-11-26 Uros Bizjak <ubizjak@gmail.com>
663 * config/i386/i386.c (dbx_register_map): Use IGNORED_DWARF_REGNUM
664 for registers for which no debug information can be generated.
665 Use INVALID_REGNUM for non-existent registers.
666 (dbx64_register_map): Ditto.
667 (svr4_dbx_register_map): Ditto.
669 2018-11-26 Sandra Loosemore <sandra@codesourcery.com>
672 * doc/invoke.texi (Optimize Options): Document that
673 -freorder-functions also uses "hot" and "cold" attributes.
675 2018-11-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
677 * config/aarch64/aarch64.c (aarch64_mangle_type): Fix typo in comment.
679 2018-11-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
681 * config/arm/types.md (mul64): Rename to...
682 (widen_mul64): ... This.
683 * config/arm/arm-generic.md: Rename mul64 to widen_mul64.
684 * config/arm/cortex-a15.md: Likewise.
685 * config/arm/cortex-a5.md: Likewise.
686 * config/arm/cortex-a53.md: Likewise.
687 * config/arm/cortex-a57.md: Likewise.
688 * config/arm/cortex-a7.md: Likewise.
689 * config/arm/cortex-m4.md: Likewise.
690 * config/arm/exynos-m1.md: Likewise.
691 * config/arm/marvell-pj4.md: Likewise.
692 * config/arm/xgene1.md: Likewise.
694 2018-11-26 Richard Biener <rguenther@suse.de>
696 PR tree-optimization/88182
697 * tree-vect-loop.c (vectorizable_reduction): Pick up single
698 correct reduc_def_info.
699 * tree-vect-slp.c (vect_analyze_slp_instance): Set
700 STMT_VINFO_REDUC_DEF of the first stmt.
702 2018-11-26 Andreas Krebbel <krebbel@linux.ibm.com>
704 * doc/invoke.texi: Document z14/arch12 -march option.
706 2018-11-25 Sandra Loosemore <sandra@codesourcery.com>
709 * doc/invoke.texi (Option Summary): Add -Wmissing-noreturn.
710 (Warning Options): Likewise.
712 2018-11-25 Sandra Loosemore <sandra@codesourcery.com>
715 * doc/extend.texi (Common Function Attributes): Clarify that
716 functions with "const" attribute can read const global variables.
718 2018-11-26 Alan Modra <amodra@gmail.com>
720 * config/rs6000/rs6000.h (OBJECT_PEF, TARGET_MACOS): Delete.
722 2018-11-25 Sandra Loosemore <sandra@codesourcery.com>
725 * doc/extend.texi (Common Variable Attributes): Use preferred
726 placement of type attributes in examples, plus whitespace fixes.
727 (Type Attributes): Clarify why placement of attributes
728 immediately after struct/union/enum keyword is preferred.
729 (Common Type Attributes): Use preferred placement of type
730 attributes in examples, plus more whitespace fixes.
732 2018-11-25 Paul Koning <ni1d@arrl.net>
734 * config/pdp11/pdp11.h (TARGET_HAS_NO_HW_DIVIDE): Define.
736 2018-11-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
738 * doc/sourcebuild.texi (Effective-Target Keywords, Environment
739 attributes): Document d_runtime.
741 2018-11-25 Vladimir Makarov <vmakarov@redhat.com>
744 * ira-costs.c (record_operand_costs): Initiate register move cost
747 2018-11-23 Jeff Law <law@redhat.com>
749 PR rtl-optimization/87468
750 * tree-ssa-threadupdate.c (create_block_for_threading): Clear
751 EDGE_IGNORE on all outgoing edges of the duplicate block.
753 2018-11-23 Vladimir Makarov <vmakarov@redhat.com>
756 * ira-costs.c (record_operand_costs): Use bigger hard reg class if
757 its mode does not fit to the original class.
759 2018-11-23 Martin Sebor <msebor@redhat.com>
761 PR tree-optimization/87756
762 * expr.c (string_constant): Synthesize a string literal from
763 the address of a constant character.
764 * tree.c (build_string_literal): Add an argument.
765 * tree.h (build_string_literal): Same.
767 2018-11-23 Christoph Muellner <christoph.muellner@theobroma-systems.com>
769 * config/aarch64/aarch64-cores.def: Define emag.
770 * config/aarch64/aarch64-tune.md: Regenerated with emag.
771 * config/aarch64/aarch64.c (emag_tunings): New struct.
772 * doc/invoke.texi: Document mtune value.
774 2018-11-23 Christoph Muellner <christoph.muellner@theobroma-systems.com>
775 Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
777 * config/aarch64/aarch64.c (xgene1_tunings): Optimize Xgene1 tunings
779 * config/aarch64/aarch64.c (xgene1_tunings): Add Xgene1 specific
781 * config/aarch64/aarch64.c (xgene1_addrcost_table): Correct the post-modify
783 * config/arm/aarch-cost-tables.h (xgene1_extra_costs): Update the cost table
786 2018-11-23 Richard Biener <rguenther@suse.de>
788 PR tree-optimization/88149
789 * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
790 the case where there are two different def types for the
791 same operand at different operand position in the same stmt.
793 2018-23-11 Mihail Ionescu <mihail.ionescu@arm.com>
795 * config/arm/arm.c (arm_expand_compare_and_swap): Simplify and call
796 gen_atomic_compare_swap_1.
797 (arm_evpc_neon_vuzp): Likewise gen_neon_vuzp_internal.
798 (arm_evpc_neon_vtrn): Likewise gen_neon_vtrn_internal.
799 (arm_evpc_neon_vext): Likewise gen_neon_vext_internal.
800 (arm_evpc_neon_vzip): Likewise gen_neon_vzip_internal.
801 (arm_evpc_neon_vrev): Replace the function pointer and simplify the mode
803 * config/arm/arm.md (neon_vext<mode>),
804 (neon_vrev64<mode>, neon_vrev32<mode>),
805 (neon_vrev16<mode>, neon_vtrn<mode>_internal),
806 (neon_vzip<mode>_internal, neon_vuzp<mode>_internal): Add an '@'character
807 before the pattern name.
808 * config/arm/sync.md:
809 (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1),
810 (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
812 2018-11-23 Jakub Jelinek <jakub@redhat.com>
814 PR tree-optimization/86614
815 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return early
816 if TREE_NO_WARNING is set on ref.ref.
818 2018-11-22 Sandra Loosemore <sandra@codesourcery.com>
819 Alan Coopersmith <alan.coopersmith@oracle.com>
822 * doc/extend.texi (Designated Inits): Clarify handling of multiple
823 initializers for unions.
825 2018-11-22 Jan Hubicka <jh@suse.cz>
828 * ipa-devirt.c (type_variants_equivalent_p): Drop warn and warned
829 parameters; do not warn here.
830 (odr_subtypes_equivalent_p): Likewise.
831 (warn_odr): Fix typo.
832 (warn_types_mismatch): Do not output confused warnings on integer types.
833 (odr_types_equivalent_p): Update.
835 2018-11-22 Jakub Jelinek <jakub@redhat.com>
837 * config/i386/i386.c (ix86_option_override_internal,
838 ix86_can_inline_p, classify_argument, construct_container,
839 ix86_expand_prologue, ix86_expand_split_stack_prologue,
840 ix86_expand_carry_flag_compare, expand_set_or_movmem_via_loop,
841 expand_setmem_epilogue_via_loop, promote_duplicated_reg,
842 ix86_expand_set_or_movmem, ix86_init_builtins_va_builtins_abi):
845 * config/i386/i386.c (ix86_option_override_internal): For
846 stack_protector_guard related options, use opts_set->x_ instead
847 of global_options_set. and prefix options with opts->x_ . Move
848 defaults for offset and reg into else block.
850 2018-11-22 Uros Bizjak <ubizjak@gmail.com>
852 * config/i386/i386.c (ix86_check_avx_upper_register):
853 Return true for all SSE registers with mode bitsize > 128.
855 2018-11-22 Vladimir Makarov <vmakarov@redhat.com>
857 PR rtl-optimization/87718
858 * ira-costs.c: Remove trailing white-spaces.
859 (record_operand_costs): Add a special treatment for moves
860 involving a hard register.
862 2018-11-22 Uros Bizjak <ubizjak@gmail.com>
864 * config/i386/i386.c (ix86_avx_emit_vzeroupper): Remove.
865 (ix86_emit_mode_set) <case AVX_U128>: Emit vzeroupper here.
867 2018-11-22 Martin Liska <mliska@suse.cz>
869 * common/config/i386/i386-common.c (processor_names): Add
870 static assert and add missing "znver2".
871 (ix86_get_valid_option_values): Add checking assert for null
872 values and add "native" value if feasible.
873 * config/i386/i386.h: Do not declare size of processor_names.
874 * common/config/i386/i386-common.c:
875 * config/i386/i386.c: Add static assert for size
876 of processor_cost_table.
878 2018-11-22 Thomas Preud'homme <thomas.preudhomme@linaro.org>
880 * target-insns.def (stack_protect_combined_set): Define new standard
882 (stack_protect_combined_test): Likewise.
883 * cfgexpand.c (stack_protect_prologue): Try new
884 stack_protect_combined_set pattern first.
885 * function.c (stack_protect_epilogue): Try new
886 stack_protect_combined_test pattern first.
887 * config/arm/arm.c (require_pic_register): Add pic_reg and compute_now
888 parameters to control which register to use as PIC register and force
889 reloading PIC register respectively. Insert in the stream of insns if
891 (legitimize_pic_address): Expose above new parameters in prototype and
892 adapt recursive calls accordingly. Use pic_reg if non null instead of
894 (arm_load_pic_register): Add pic_reg parameter and use it if non null.
895 (arm_legitimize_address): Adapt to new legitimize_pic_address
897 (thumb_legitimize_address): Likewise.
898 (arm_emit_call_insn): Adapt to require_pic_register prototype change.
899 (arm_expand_prologue): Adapt to arm_load_pic_register prototype change.
900 (thumb1_expand_prologue): Likewise.
901 * config/arm/arm-protos.h (legitimize_pic_address): Adapt to prototype
903 (arm_load_pic_register): Likewise.
904 * config/arm/predicated.md (guard_addr_operand): New predicate.
905 (guard_operand): New predicate.
906 * config/arm/arm.md (movsi expander): Adapt to legitimize_pic_address
908 (builtin_setjmp_receiver expander): Adapt to thumb1_expand_prologue
910 (stack_protect_combined_set): New expander..
911 (stack_protect_combined_set_insn): New insn_and_split pattern.
912 (stack_protect_set_insn): New insn pattern.
913 (stack_protect_combined_test): New expander.
914 (stack_protect_combined_test_insn): New insn_and_split pattern.
915 (arm_stack_protect_test_insn): New insn pattern.
916 * config/arm/thumb1.md (thumb1_stack_protect_test_insn): New insn pattern.
917 * config/arm/unspecs.md (UNSPEC_SP_SET): New unspec.
918 (UNSPEC_SP_TEST): Likewise.
919 * doc/md.texi (stack_protect_combined_set): Document new standard
921 (stack_protect_set): Clarify that the operand for guard's address is
923 (stack_protect_combined_test): Document new standard pattern name.
924 (stack_protect_test): Clarify that the operand for guard's address is
927 2018-11-22 Richard Biener <rguenther@suse.de>
929 PR tree-optimization/88148
930 * tree-ssa-loop-niter.c (simplify_replace_tree): Get optional
931 valueization callback parameter and handle it.
932 * tree-ssa-loop-niter.h (simplify_replace_tree): Export.
933 * tree-ssa-sccvn.c (process_bb): Eliminate in loop niter trees.
935 2018-11-22 Richard Biener <rguenther@suse.de>
939 * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs
940 which can appear in size expressions.
941 * tree-streamer-in.c (unpack_ts_base_value_fields): Stream
942 CALL_EXPR_BY_DESCRIPTOR.
943 (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN.
944 * tree-streamer-out.c (pack_ts_base_value_fields): Stream
945 CALL_EXPR_BY_DESCRIPTOR.
946 (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN.
950 * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
953 2018-11-22 Richard Biener <rguenther@suse.de>
955 PR tree-optimization/88069
956 * tree-ssa-sccvn.c (visit_phi): Tweak previous fix to not
957 apply to default defs.
959 2018-11-22 Jakub Jelinek <jakub@redhat.com>
963 * config/i386/i386.c (ix86_option_override_internal): Default
964 ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
966 * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
967 stack_protect_test, stack_protect_test_<mode>): Use empty condition
968 instead of TARGET_SSP_TLS_GUARD.
970 2018-11-22 Martin Liska <mliska@suse.cz>
973 * config/i386/i386.c (ix86_option_override_internal): Error
974 about usage -mabi=ms and -fsanitize=thread.
976 2018-11-21 Uros Bizjak <ubizjak@gmail.com>
979 2013-10-26 Vladimir Makarov <vmakarov@redhat.com>
982 2013-10-25 Vladimir Makarov <vmakarov@redhat.com>
984 * lra-spills.c (lra_final_code_change): Remove useless move insns.
986 2018-11-21 Uros Bizjak <ubizjak@gmail.com>
989 * function.c (expand_function_end): Do not emit extra blockage insn.
991 2018-11-21 Lokesh Janghel <lokeshjanghel91@gmail.com>
994 * config/i386/i386.c (function_value_ms_64): Return AX_REG instead
995 of FIRST_SSE_REG for 4 or 8 byte modes.
997 2018-11-21 Jan Hubicka <jh@suse.cz>
1000 * tree.c (fld_decl_context): Break out from ...
1001 (free_lang_data_in_decl): ... here; free TREE_PUBLIC, TREE_PRIVATE
1002 DECL_ARTIFICIAL of TYPE_DECL; do not free TREE_TYPE of TYPE_DECL.
1003 (fld_incomplete_type_of): Build copy of TYP_DECL.
1004 * ipa-devirt.c (free_enum_values): Rename to ...
1005 (free_odr_warning_data): ... this one; free also duplicated TYPE_DECLs
1006 and TREE_TYPEs of TYPE_DECLs.
1007 (get_odr_type): Initialize odr_vtable_hash if needed.
1009 2018-11-21 Alexandre Oliva <oliva@adacore.com>
1011 * final.c (compute_discriminator): Declare. Renamed from...
1012 (maybe_set_discriminator): ... this. Set and return a local.
1013 (override_discriminator): New.
1014 (final_scan_insn_1): Set it.
1015 (notice_source_line): Adjust. Always set discriminator.
1017 2018-11-21 Jakub Jelinek <jakub@redhat.com>
1020 * config/aarch64/atomics.md (@aarch64_compare_and_swap<mode>): Use
1021 rIJ constraint for aarch64_plus_operand rather than rn.
1023 2018-11-21 Renlin Li <renlin.li@arm.com>
1026 * explow.h (get_dynamic_stack_size): Declare it as external.
1027 * explow.c (record_new_stack_level): Remove function static attribute.
1028 * function.c (assign_stack_local_1): Dynamically align the stack slot
1029 addr for parameter copy on the stack.
1031 2018-11-21 Richard Biener <rguenther@suse.de>
1034 * bitmap.c (bitmap_last_set_bit): Refactor to avoid warning.
1035 * Makefile.in (bitmap.o-warn): Remove again.
1037 2018-11-20 Jeff Law <law@redhat.com>
1039 PR tree-optimization/88069
1040 * tree-ssa-dom.c (record_equivalences_from_phis): Propagate away
1041 degenerate virtual PHIs.
1043 2018-11-21 H.J. Lu <hongjiu.lu@intel.com>
1046 * config/i386/sse.md (sse4_1_<code>v8qiv8hi2<mask_name>): Replace
1047 nonimmediate_operand with register_operand.
1048 (avx2_<code>v8qiv8si2<mask_name>): Likewise.
1049 (sse4_1_<code>v4qiv4si2<mask_name>): Likewise.
1050 (sse4_1_<code>v4hiv4si2<mask_name>): Likewise.
1051 (sse4_1_<code>v2qiv2di2<mask_name>): Likewise.
1052 (avx512f_<code>v8qiv8di2<mask_name>): Likewise.
1053 (avx2_<code>v4qiv4di2<mask_name>): Likewise.
1054 (avx2_<code>v4hiv4di2<mask_name>): Likewise.
1055 (sse4_1_<code>v2hiv2di2<mask_name>): Likewise.
1056 (sse4_1_<code>v2siv2di2<mask_name>): Likewise.
1057 (*sse4_1_<code>v8qiv8hi2<mask_name>_1): New pattern.
1058 (*sse4_1_<code>v8qiv8hi2<mask_name>_2): Likewise.
1059 (*avx2_<code>v8qiv8si2<mask_name>_1): Likewise.
1060 (*avx2_<code>v8qiv8si2<mask_name>_2): Likewise.
1061 (*sse4_1_<code>v4qiv4si2<mask_name>_1): Likewise.
1062 (*sse4_1_<code>v4qiv4si2<mask_name>_2): Likewise.
1063 (*sse4_1_<code>v4hiv4si2<mask_name>_1): Likewise.
1064 (*sse4_1_<code>v4hiv4si2<mask_name>_2): Likewise.
1065 (*avx512f_<code>v8qiv8di2<mask_name>_1): Likewise.
1066 (*avx512f_<code>v8qiv8di2<mask_name>_2): Likewise.
1067 (*avx2_<code>v4qiv4di2<mask_name>_1): Likewise.
1068 (*avx2_<code>v4qiv4di2<mask_name>_2): Likewise.
1069 (*avx2_<code>v4hiv4di2<mask_name>_1): Likewise.
1070 (*avx2_<code>v4hiv4di2<mask_name>_2): Likewise.
1071 (*sse4_1_<code>v2hiv2di2<mask_name>_1): Likewise.
1072 (*sse4_1_<code>v2hiv2di2<mask_name>_2): Likewise.
1073 (*sse4_1_<code>v2siv2di2<mask_name>_1): Likewise.
1074 (*sse4_1_<code>v2siv2di2<mask_name>_2): Likewise.
1076 2018-11-21 H.J. Lu <hongjiu.lu@intel.com>
1078 * read-rtl.c (apply_subst_iterator): Handle define_split and
1079 define_insn_and_split.
1081 2018-11-21 Jakub Jelinek <jakub@redhat.com>
1083 PR rtl-optimization/87817
1084 * config/i386/i386.md (bmi2_bzhi_<mode>3, *bmi2_bzhi_<mode>3,
1085 *bmi2_bzhi_<mode>3_1, *bmi2_bzhi_<mode>3_1_ccz): Use IF_THEN_ELSE
1086 in the pattern to avoid triggering UB when operands[2] is zero.
1087 (tbm_bextri_<mode>): New expander. Renamed the old define_insn to ...
1088 (*tbm_bextri_<mode>): ... this.
1090 2018-11-21 Tom de Vries <tdevries@suse.de>
1093 * params.def (HOT_BB_COUNT_FRACTION): Terminate help message with
1096 2018-11-21 Andreas Krebbel <krebbel@linux.ibm.com>
1098 * configure.ac: Add check for Binutils to determine whether vector
1099 load/store alignments hints are being supported.
1100 * config.in: Regenerate.
1101 * configure: Regenerate.
1102 * config/s390/s390.c (print_operand): Support new output
1104 * config/s390/s390.md ("movti"): Append alignment hint output
1105 using the new output modifier 'A'.
1106 * config/s390/vector.md ("mov<mode>", "*vec_tf_to_v1tf")
1107 ("*vec_ti_to_v1ti"): Likewise.
1109 2018-11-20 Martin Sebor <msebor@redhat.com>
1111 * doc/extend.texi (Other Builtins): Add __builtin_has_attribute.
1113 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
1116 * ipa-devirt.c (odr_types_equivalent_p): Use operand_equal_p to
1117 compare ENUM values.
1119 2018-11-20 Jakub Jelinek <jakub@redhat.com>
1121 PR tree-optimization/87895
1122 * omp-simd-clone.c (ipa_simd_modify_function_body): When removing
1123 or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT.
1124 (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that
1125 redirects edges to EXIT to edges to incr_bb, iterate while EXIT
1126 has any preds and always use EDGE_PRED (, 0).
1128 2018-11-20 Uros Bizjak <ubizjak@gmail.com>
1131 * mode-switching.c (create_pre_exit): After reload, always split the
1132 fallthrough edge to the exit block.
1134 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
1136 * ipa-devirt.c (add_type_duplicate): Do not ICE on incomplete enums.
1137 * tree.c (build_array_type_1): Forward declare.
1138 (fld_type_variant_equal_p): Add INNER_TYPE parameter.
1139 (fld_type_variant): Likewise.
1140 (fld_simplified_types): New hash.
1141 (fld_process_array_type): New function.
1142 (fld_incomplete_type_of): Handle array and enumeration types.
1143 (fld_simplified_type): Handle simplification of arrays.
1144 (free_lang_data): Allocate and free simplified types hash.
1146 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
1149 * ipa-devirt.c (odr_subtypes_equivalent_p): Report ODR violation
1150 when sybtype already violates ODR.
1151 (get_odr_type): Do not ICE when insert is false and type duplicate
1152 is not registered yet.
1153 (register_odr_type): Be sure to register subtypes first.
1155 2018-11-20 Andreas Krebbel <krebbel@linux.ibm.com>
1157 * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the
1158 DImode parts of the target operand.
1160 2018-11-20 Nathan Sidwell <nathan@acm.org>
1163 * Makefile.in (bitmap.o-warn): Use -Wno-error=array-bounds.
1165 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
1168 * tree.c (free_lang_data_in_cgraph): Add argument fld; break out
1170 (free_lang_data) ... here; update call of free_lang_data_in_cgraph.
1172 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
1175 * ipa-fnsummary.c (pass_ipa_fnsummary): Do not remove functions
1176 * ipa.c (possible_inline_candidate_p): Break out from ..
1177 (process_references): ... here ; drop before_inlining_p;
1178 cleanup handling of alises.
1179 (walk_polymorphic_call_targets): Likewise.
1180 (symbol_table::remove_unreachable_nodes): Likewise.
1181 * passes.c (pass_data_ipa_remove_symbols): New structure.
1182 (pass_ipa_remove_symbols): New pass.
1183 (make_pass_ipa_remove_symbols): New function.
1184 * tree-pass.h (make_pass_ipa_remove_symbols): Declare.
1185 * passes.def (pass_ipa_remove_symbols): Schedule after early passes.
1187 2018-11-20 Richard Biener <rguenther@suse.de>
1189 * tree-vect-stmts.c (vectorizable_condition): Do not get
1190 at else_clause vect def for EXTRACT_LAST_REDUCTION. Remove
1191 pointless vect_is_simple_use calls.
1193 2018-11-20 Richard Biener <rguenther@suse.de>
1195 PR tree-optimization/88074
1196 * tree-ssa-dom.c (pass_dominator::execute): Do not walk
1199 2018-11-20 Richard Biener <rguenther@suse.de>
1201 PR tree-optimization/88069
1202 * tree-ssa-sccvn.c (visit_phi): Do not value-number to unvisited
1203 virtual PHI arguments.
1205 2018-11-20 Ilya Leoshkevich <iii@linux.ibm.com>
1208 * config/s390/s390.md: Skip LT(G) peephole when literal pool is
1210 * rtl.h (contains_constant_pool_address_p): New function.
1211 * rtlanal.c (contains_constant_pool_address_p): Likewise.
1213 2018-11-20 Richard Biener <rguenther@suse.de>
1216 * alias.c (component_uses_parent_alias_set_from): Remove
1217 alias-set zero and TYPE_TYPELESS_STORAGE case both already
1218 handled in other ways.
1220 2018-11-20 Richard Biener <rguenther@suse.de>
1222 PR tree-optimization/88087
1223 * tree-ssa-pre.c (create_expression_by_pieces): Re-materialize
1225 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Remember
1228 2018-11-20 Richard Biener <rguenther@suse.de>
1231 * tree-data-ref.c (lambda_matrix_right_hermite): Use abs_hwi.
1233 2018-11-19 Jan Hubicka <hubicka@ucw.cz>
1236 * ipa-devirt.c (free_enum_values): Do not ICE on ODR vilations.
1238 2018-11-19 Sandra Loosemore <sandra@codesourcery.com>
1241 * doc/invoke.texi (Link Options): Mention shared libraries
1242 in documentation for the -l option. Simplify discussion and
1243 point to the system linker documentation for details.
1245 2018-11-19 Peter Bergner <bergner@linux.ibm.com>
1247 PR rtl-optimization/88033
1248 * ira-lives.c (non_conflicting_reg_copy_p): Skip copies from a register
1249 to itself. Use HARD_REGISTER_NUM_P.
1251 2018-11-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1253 * tree-vect-loop.c (vect_transform_loop): Disable further unrolling
1254 of the loop if vf is non-constant.
1256 2018-11-19 David Malcolm <dmalcolm@redhat.com>
1258 PR tree-optimization/87025
1259 * dumpfile.c (dump_context::begin_scope): Call end_any_optinfo
1260 immediately after creating the scope optinfo.
1261 (selftest::test_pr87025): New function.
1262 (selftest::dumpfile_c_tests): Call it.
1263 * optinfo-emit-json.cc (optrecord_json_writer::pop_scope): Assert
1264 that we're not popping the top-level records array.
1265 * optinfo.cc (optinfo::handle_dump_file_kind): Assert that we're
1266 not changing the kind of a "scope" optinfo.
1268 2018-11-19 David Malcolm <dmalcolm@redhat.com>
1270 PR tree-optimization/87025
1271 * dump-context.h: Include "optinfo.h".
1272 (class optrecord_json_writer): New forward decl.
1273 (dump_context::forcibly_enable_optinfo_p): Delete.
1274 (dump_context::optinfo_enabled_p): New member function.
1275 (dump_context::optimization_records_enabled_p): New member
1277 (dump_context::set_json_writer): New member function.
1278 (dump_context::emit_optinfo): New member function.
1279 (dump_context::m_forcibly_enable_optinfo): Delete.
1280 (dump_context::m_json_writer): New member data.
1281 * dumpfile.c (dump_context::set_json_writer): New member function.
1282 (dump_context::finish_any_json_writer): New member function.
1283 (dump_context::end_scope): Replace call to
1284 optimization_records_maybe_pop_dump_scope with call to
1285 m_json_writer->pop_scope.
1286 (dump_context::optinfo_enabled_p): New member function.
1287 (dump_context::end_any_optinfo): Replace call to optinfo::emit with call
1288 to dump_context::emit_optinfo.
1289 (dump_context::emit_optinfo): New member function.
1290 (temp_dump_context::temp_dump_context): Replace
1291 m_forcibly_enable_optinfo with call to set_json_writer.
1292 (temp_dump_context::~temp_dump_context): Clean up any json writer.
1293 * optinfo-emit-json.cc (class optrecord_json_writer): Move to
1295 (the_json_writer): Delete.
1296 (optimization_records_start): Delete.
1297 (optimization_records_finish): Delete.
1298 (optimization_records_enabled_p): Delete, in favor of
1299 dump_context::optimization_records_enabled_p.
1300 (optimization_records_maybe_record_optinfo): Delete.
1301 (optimization_records_maybe_pop_dump_scope): Delete.
1302 * optinfo-emit-json.h: Include "json.h". Delete forward
1304 (optimization_records_start): Delete.
1305 (optimization_records_finish): Delete.
1306 (optimization_records_enabled_p): Delete.
1307 (optimization_records_maybe_record_optinfo): Delete.
1308 (optimization_records_maybe_pop_dump_scope): Delete.
1309 (class optrecord_json_writer): Move here from
1310 optinfo-emit-json.cc.
1311 * optinfo.cc (optinfo::emit_for_opt_problem): Replace call
1312 to optinfo::emit with call to dump_context::emit_optinfo.
1313 (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
1314 (optinfo_enabled_p): Delete, in favor of
1315 dump_context::optinfo_enabled_p.
1316 (optinfo_wants_inlining_info_p): Update for conversion o
1317 optimization_records_enabled_p to a member function of
1319 * optinfo.h (optinfo_enabled_p): Delete, in favor of
1320 dump_context::optinfo_enabled_p.
1321 (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
1322 * toplev.c: Include "dump-context.h".
1323 (compile_file): Replace call to optimization_records_finish with
1324 dump_context::finish_any_json_writer.
1325 (do_compile): Replace call to optimization_records_start with
1326 conditionally creating a optrecord_json_writer for the
1329 2018-11-19 Jakub Jelinek <jakub@redhat.com>
1331 PR tree-optimization/88071
1332 * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead
1333 of false as last argument to gsi_remove.
1334 * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of
1335 false as last argument to gsi_replace.
1338 * omp-expand.c: Don't include debug.h.
1339 (adjust_context_and_scope): Add REGION argument. Find DECL_CONTEXT
1340 from innermost outer parallel, task, teams or target that has a
1341 child_fn set, or, if there is no such outer region, use
1342 current_function_decl. Do the DECL_CONTEXT adjustment regardless of
1343 whether a suitable BLOCK is found or not.
1344 (expand_parallel_call, expand_teams_call): Don't call
1345 adjust_context_and_scope here.
1346 (grid_expand_target_grid_body): Revert 2017-01-25 changes.
1347 (expand_omp_taskreg, expand_omp_target): Likewise. Call
1348 adjust_context_and_scope.
1349 * dwarf2out.c (dwarf2out_early_global_decl): For
1350 decl_function_context recurse instead of calling dwarf2out_decl.
1352 2018-11-20 Eric Botcazou <ebotcazou@adacore.com>
1354 PR rtl-optimization/85925
1355 * rtl.h (word_register_operation_p): New predicate.
1356 * combine.c (record_dead_and_set_regs_1): Only apply specific handling
1357 for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX.
1358 * rtlanal.c (nonzero_bits1): Likewise. Adjust couple of comments.
1359 (num_sign_bit_copies1): Likewise.
1361 2018-11-19 Richard Biener <rguenther@suse.de>
1364 * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
1367 2018-11-19 Eric Botcazou <ebotcazou@adacore.com>
1369 * compare-elim.c (struct comparison): Add not_in_a field.
1370 (is_not): New static function.
1371 (strip_not): Likewise.
1372 (conforming_compare): Handle a NOT in the first operand.
1373 (can_eliminate_compare): Likewise.
1374 (find_comparison_dom_walker::before_dom_children): Likewise.
1375 (try_eliminate_compare): Likewise.
1376 * config/visium/visium.md (negsi2_insn_set_carry): Turn into...
1377 (neg<mode>2_insn_set_carry): ...this and add missing NEG operation.
1379 2018-11-19 Jonathan Wakely <jwakely@redhat.com>
1381 * doc/extend.texi (Common Type Attributes): Fix typos.
1383 2018-11-19 Martin Liska <mliska@suse.cz>
1385 PR gcov-profile/88045
1386 * coverage.c (coverage_begin_function): Add assert.
1388 2018-11-18 Sandra Loosemore <sandra@codesourcery.com>
1391 * doc/extend.texi (Common Function Attributes): Document that
1392 no_instrument_function applies to -p and -pg, too.
1393 * doc/invoke.texi (Instrumentation Options): Add cross-references
1394 to docs for -p, -pg, and -finstrument-functions.
1396 2018-11-18 Alan Modra <amodra@gmail.com>
1398 * config/gnu-user.h (GNU_USER_TARGET_CRTI): Define.
1399 (GNU_USER_TARGET_STARTFILE_SPEC): Use it here.
1400 (GNU_USER_TARGET_CRTN): Define.
1401 (GNU_USER_TARGET_ENDFILE_SPEC): Use it here.
1402 (GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC): Define.
1404 2018-11-18 Alan Modra <amodra@gmail.com>
1406 * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Delete
1407 !HAVE_LD_PIE variant.
1408 (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
1410 2018-11-17 Nick Clifton <nickc@redhat.com>
1411 Sandra Loosemore <sandra@codesourcery.com>
1414 * doc/invoke.texi (Overall Options): Document that --help and
1415 --help=<value> options cannot be combined.
1417 2018-11-17 Jim Wilson <jimw@sifive.com>
1419 * config/riscv/riscv.c (epilogue_cfa_sp_offset): New.
1420 (riscv_restore_reg): If restoring HARD_FRAME_POINTER_REGNUM, and
1421 epilogue_cfa_sp_offset set, then add REG_CFA_DEF_CFA regnote.
1422 (riscv_expand_epilogue): Initialize epilogue_cfa_sp_offset. Set it
1423 to step2 if frame_pointer_needed and step1 is 0.
1425 2018-11-17 Sandra Loosemore <sandra@codesourcery.com>
1428 * doc/cpp.texi (System Headers): Add note about implicit
1429 extern "C" block on targets that define SYSTEM_IMPLICIT_EXTERN_C.
1431 2018-11-17 Sandra Loosemore <sandra@codesourcery.com>
1434 * doc/invoke.texi (C++ Dialect Options): Clarify usage of
1435 -fno-implicit-templates.
1437 2018-11-17 Jakub Jelinek <jakub@redhat.com>
1439 PR tree-optimization/87546
1440 * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
1441 min_precision variable, initially set it to orig_precision, only does
1442 something if op_type's precision is <= min_precision and update
1443 min_precision whenever calling set_op.
1445 2018-11-16 Jan Hubicka <hubicka@ucw.cz>
1448 * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
1450 2018-11-16 Sandra Loosemore <sandra@codesourcery.com>
1452 * doc/invoke.texi (Option Summary): Fix whitespace and line
1453 breaks in @gccoptlist environments.
1454 (Warning Options): Likewise.
1455 (Optimize Options): Likewise.
1456 (PowerPC SPE Options): Likewise.
1457 (RS/6000 and PowerPC Options): Likewise.
1459 2018-11-16 Jeff Law <law@redhat.com>
1461 * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
1464 2018-11-16 Sandra Loosemore <sandra@codesourcery.com>
1467 * doc/invoke.texi (Optimize Options): Update options enabled by
1468 fprofile-generate, -fprofile-use, and -fauto-profile.
1470 2018-11-16 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1473 * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
1474 Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
1475 * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
1477 * target.h: Add psi, pdi, pti to struct asm_int_op definition.
1478 * targhooks.c (default_print_patchable_function_entry): Assert
1479 asm_int_op does not return a NULL string.
1480 * varasm.c (integer_asm_op): Return the op for a partial int type
1481 when the requested size does not correspond to an integer type.
1482 * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
1483 * doc/tm.texi: Regenerate.
1485 2018-11-12 Jason Merrill <jason@redhat.com>
1487 * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
1489 2018-11-16 Michael Meissner <meissner@linux.ibm.com>
1491 * config/rs6000/constraints.md (wF constraint): Remove power9
1492 fusion documentation. Just document wF's use for power8 fusion.
1493 * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
1494 (fusion_gpr_addis): Delete power9 fusion support. Change power8
1495 fusion support to require the upper 12 bits to be all 0's or all
1497 (fusion_gpr_mem_load): Add comment.
1498 (fusion_addis_mem_combo_load): Remove power9 fusion support.
1499 (fusion_addis_mem_combo_store): Delete.
1500 (fusion_offsettable_mem_operand): Delete.
1501 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
1503 (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
1504 (POWERPC_MASKS): Delete power9 fusion.
1505 * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
1506 (fusion_p9_p): Delete.
1507 (expand_fusion_p9_load): Delete.
1508 (expand_fusion_p9_store): Delete.
1509 (emit_fusion_p9_load): Delete.
1510 (emit_fusion_p9_store): Delete.
1511 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
1513 (rs6000_option_override_internal): Set power8 fusion based on
1514 whether we are tuning for power8. Delete power9 fusion support.
1515 (rs6000_opt_masks): Delete -mpower9-fusion switch.
1516 (emit_fusion_load): Rename emit_fusion_load_store to
1517 emit_fusion_load, and drop fusion store support. Update callers.
1518 (emit_fusion_load_store): Likewise.
1519 (emit_fusion_gpr_load): Likewise.
1520 (fusion_p9_p): Delete.
1521 (expand_fusion_p9_load): Delete.
1522 (expand_fusion_p9_store): Delete.
1523 (emit_fusion_p9_load): Delete.
1524 (emit_fusion_p9_store): Delete.
1525 * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
1526 (GPR_FUSION): Delete.
1527 (FPR_FUSION): Delete.
1528 (power9 fusion peephole2s): Delete.
1529 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
1530 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
1531 (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
1532 (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
1533 (fusion_p9_<mode>_constant): Delete.
1534 * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
1535 power9 fusion switch.
1536 * doc/md.texi (PowerPC constraints): Update wF constraint
1537 documentation for power8 fusion only.
1539 2018-11-16 Jakub Jelinek <jakub@redhat.com>
1541 PR rtl-optimization/87475
1542 * cfgrtl.c (patch_jump_insn): Allow redirection failure for
1543 CROSSING_JUMP_P insns.
1544 (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
1546 2018-11-16 Uros Bizjak <ubizjak@gmail.com>
1549 * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
1550 * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
1551 (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
1553 2018-11-16 Jakub Jelinek <jakub@redhat.com>
1556 * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
1557 is VOIDmode, in which case use int_mode. Similarly for op1_mode.
1559 2018-11-16 Eric Botcazou <ebotcazou@adacore.com>
1561 * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
1562 location_t parameter.
1563 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
1564 (switch_decision_tree::do_jump_if_equal): Likewise.
1565 * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
1566 of switch statement to emit_case_nodes.
1567 (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
1568 set it on the newly built GIMPLE comparison statement.
1569 (switch_decision_tree::do_jump_if_equal): Likewise.
1570 (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
1571 in calls to do_jump_if_equal as well as recursive calls.
1573 2018-11-16 Andi Kleen <ak@linux.intel.com>
1575 * config/i386/i386.md: Allow memory operands to ptwrite.
1577 2018-11-16 Michael Matz <matz@suse.de>
1579 * system.h (PRsa): New macro.
1580 (SIZE_AMOUNT): Cast number to uint64_t.
1581 * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
1582 (pool_usage::dump_footer): Likewise and also use PRIu64.
1583 * bitmap.h (bitmap_usage::dump): Likewise.
1584 * ggc-common.c (ggc_usage::dump): Likewise.
1585 * ggc-page.c (ggc_print_statistics): Likewise.
1586 * input.c (dump_line_table_statistics): Likewise.
1587 * mem-stats.h (mem_usage::dump): Likewise.
1588 (mem_usage::dump_footer): Likewise.
1589 * rtl.c (dump_rtx_statistics): Likewise.
1590 * tree-cfg.c (dump_cfg_stats): Likewise.
1591 * tree-dfa.c (dump_dfa_stats): Likewise.
1592 * tree-phinodes.c (phinodes_print_statistics): Likewise.
1593 * tree-ssanames (ssanames_print_statistics): Likewise.
1594 * vec.c (vec_usage::dump): Likewise.
1595 (vec_usage::dump_footer): Likewise.
1597 2018-11-16 Richard Biener <rguenther@suse.de>
1599 PR tree-optimization/88011
1600 * tree-vrp.c (extract_range_from_binary_expr): Fix error in
1601 replacing set_value_range_to_undefined and
1602 set_value_range_to_varying with method calls.
1604 2018-11-16 Ilya Leoshkevich <iii@linux.ibm.com>
1606 * config/s390/s390.md
1607 (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
1609 (*r<noxa>sbg_<mode>_sll): Likewise.
1610 (*r<noxa>sbg_<mode>_srl): Likewise.
1611 (*r<noxa>sbg_sidi_srl): New pattern.
1613 2018-11-16 Jerome Lambourg <lambourg@adacore.com>
1615 * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
1616 Default to TARGET_VXWORKS_RTP.
1617 (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
1618 of TARGET_VXWORKS_RTP.
1619 * config/vxworksae.h: Also define TARGET_VXWORKS_HAVE_CTORS_DTORS.
1620 * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
1621 of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
1623 2018-11-15 Sandra Loosemore <sandra@codesourcery.com>
1626 * doc/extend.texi (Common Type Attributes): Make it explicit
1627 that attribute "packed" can apply to C++ classes.
1629 2018-11-15 Martin Sebor <msebor@redhat.com>
1633 * tree.c (type_argument_type): New function.
1634 * tree.h (type_argument_type): Declare it.
1635 * gcc/doc/extend.texi (alloc_align): Update and clarify.
1636 (alloc_size, nonnull, sentinel): Same.
1638 2018-11-15 Andrew Stubbs <ams@codesourcery.com>
1639 Kwok Cheung Yeung <kcy@codesourcery.com>
1641 * tree-vect-stmts.c (vectorizable_store): Don't ICE when
1642 int_mode_for_size fails.
1643 (vectorizable_load): Likewise.
1645 2018-11-15 David Malcolm <dmalcolm@redhat.com>
1647 * doc/ux.texi (Group logically-related diagnostics): Move
1648 discussion of auto_diagnostic_group into this new subsection.
1649 Give an example of where this grouping is used.
1651 2018-11-15 David Malcolm <dmalcolm@redhat.com>
1654 * Makefile.in (OBJS): Move json.o to...
1655 (OBJS-libcommon): ...here and add diagnostic-format-json.o.
1656 * common.opt (fdiagnostics-format=): New option.
1657 (diagnostics_output_format): New enum.
1658 * diagnostic-format-json.cc: New file.
1659 * diagnostic.c (default_diagnostic_final_cb): New function, taken
1660 from start of diagnostic_finish.
1661 (diagnostic_initialize): Initialize final_cb to
1662 default_diagnostic_final_cb.
1663 (diagnostic_finish): Move "being treated as errors" messages to
1664 default_diagnostic_final_cb. Call any final_cb.
1665 (default_diagnostic_finalizer): Add diagnostic_t param.
1666 (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
1667 diagnostic_finalizer callback.
1668 * diagnostic.h (enum diagnostics_output_format): New enum.
1669 (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
1670 (struct diagnostic_context): Add "final_cb".
1671 (default_diagnostic_finalizer): Add diagnostic_t param.
1672 (diagnostic_output_format_init): New decl.
1673 * doc/invoke.texi (-fdiagnostics-format): New option.
1674 * dwarf2out.c (gen_producer_string): Ignore
1675 OPT_fdiagnostics_format_.
1676 * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
1677 * lto-wrapper.c (append_diag_options): Ignore it.
1678 * opts.c (common_handle_option): Handle it.
1680 2018-11-15 David Malcolm <dmalcolm@redhat.com>
1682 PR tree-optimization/88015
1683 * graphite-isl-ast-to-gimple.c
1684 (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
1686 * graphite-sese-to-poly.c (build_poly_scop): Likewise.
1688 2018-11-15 Richard Biener <rguenther@suse.de>
1691 * gimple.c (gimple_call_flags): Union flags from decl, type
1693 * trans-mem.c (is_tm_pure_call): Simplify.
1695 2018-11-15 Richard Biener <rguenther@suse.de>
1697 PR tree-optimization/88031
1698 * tree-vect-loop.c (vectorizable_reduction): Move check
1699 for multiple types earlier so we get the expected dump.
1700 Simplify calls to vectorizable_condition.
1701 * tree-vect-stmts.h (vectorizable_condition): Update prototype.
1702 * tree-vect-stmts.c (vectorizable_condition): Instead of
1703 reduc_def and reduc_index take just a flag. Simplify
1704 code-generation now that we can rely on the defs being set up.
1705 (vectorizable_comparison): Remove unused argument.
1707 2018-11-15 Richard Biener <rguenther@suse.de>
1709 PR tree-optimization/88030
1710 * tree-complex.c (need_eh_cleanup): New global.
1711 (update_complex_assignment): Mark blocks that need EH update.
1712 (expand_complex_comparison): Likewise.
1713 (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
1714 perform EH cleanup and schedule CFG cleanup if that did anything.
1716 2018-11-15 Jakub Jelinek <jakub@redhat.com>
1718 PR rtl-optimization/88018
1719 * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
1720 on the presence of fallthru edge, rather than if it is a USE or not.
1722 2018-11-15 Richard Biener <rguenther@suse.de>
1725 * tree-data-ref.c (analyze_miv_subscript): Guard calls to
1726 analyze_subscript_affine_affine properly.
1728 2018-11-15 Xianmiao Qu <xianmiao_qu@c-sky.com>
1730 * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
1731 (GLIBC_DYNAMIC_LINKER): Define.
1732 (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
1734 2018-11-15 Bin Cheng <bin.cheng@linux.alibaba.com>
1736 PR tree-optimization/84648
1737 * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
1738 (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
1739 by calling adjust_cond_for_loop_until_wrap.
1741 2018-11-15 Sandra Loosemore <sandra@codesourcery.com>
1744 * doc/extend.texi (Common Function Attributes): Clarify linker
1745 restrictions on "aligned" attribute.
1746 (Common Variable Attributes): Likewise. Mention that linker
1747 restrictions don't apply to stack-allocated variables.
1749 2018-11-15 Joern Rennecke <joern.rennecke@riscy-ip.com>
1751 * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
1754 2018-11-14 Sandra Loosemore <sandra@codesourcery.com>
1758 * doc/invoke.texi (Optimize Options): Remove bad example about
1759 interaction between -flto and -O. Replace it with a note that
1760 you need to compile with -O and not just link. Copy-edit -flto
1761 discussion to reduce verbiage and improve flow.
1763 2018-11-14 Sandra Loosemore <sandra@codesourcery.com>
1766 * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
1767 also suppress many optimizations. Alphabetize option lists for
1768 -O1, -O2, and -Os. Add list of options disabled with -Og, and
1769 correct documentation for those options to say that.
1770 * opts.c (default_options_table): Sort table by level and option
1771 name, to make it easier to correlate to the manual.
1773 2018-11-14 Uros Bizjak <ubizjak@gmail.com>
1775 * config/i386/i386.c (ix86_print_operand_address_as): Simplify
1776 printing of the names of segment registers.
1778 2018-11-14 Nathan Sidwell <nathan@acm.org>
1782 * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
1785 2018-11-14 David Malcolm <dmalcolm@redhat.com>
1787 * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
1789 2018-11-14 Jakub Jelinek <jakub@redhat.com>
1791 P1236R1 - Signed integers are two's complement
1792 * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
1795 * hash-map.h (hash_map::iterator::reference_pair): New class.
1796 (hash_map::iterator::operator*): Return it rather than std::pair.
1798 2018-11-14 Jeff Law <law@redhat.com>
1800 * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
1801 iff the operand is a constant.
1803 2018-11-14 Aldy Hernandez <aldyh@redhat.com>
1805 * gimple-ssa-evrp-analyze.c
1806 (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
1807 ignore_equivs_equal_p to equal_p.
1808 * ipa-cp.c (meet_with_1): Use equal_p instead of
1809 ignore_equivs_equal_p.
1810 * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
1811 * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
1812 (value_range::operator==): Remove.
1813 (value_range::operator!=): Remove.
1814 (vrp_prop::visit_stmt): Use equal_p.
1815 * tree-vrp.h (value_range): Remove operator==, operator!=,
1816 ignore_equivs_equal_p.
1817 * vr-values.c (update_value_range): Use equal_p.
1819 2018-11-14 Michael Matz <matz@suse.de>
1822 * gimplify.c (collect_fallthrough_labels): Add new argument,
1823 return location via that, don't modify statements.
1824 (warn_implicit_fallthrough_r): Adjust call, don't use
1825 statement location directly.
1827 2018-11-14 Richard Biener <rguenther@suse.de>
1830 * tree-data-ref.c (split_constant_offset): Add wrapper
1831 allocating a cache hash-map.
1832 (split_constant_offset_1): Cache results of expanding
1833 expressions from SSA def stmts.
1835 2018-11-14 Richard Biener <rguenther@suse.de>
1838 * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
1840 (lambda_vector_mult_const): Likewise.
1841 (lambda_matrix_right_hermite): Use lambda_int temporaries.
1843 2018-11-14 Wilco Dijkstra <wdijkstr@arm.com>
1844 Jackson Woodruff <jackson.woodruff@arm.com>
1846 PR 71026/tree-optimization
1847 * match.pd: Simplify floating point comparisons.
1849 2018-11-14 Jakub Jelinek <jakub@redhat.com>
1851 PR rtl-optimization/87817
1852 * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
1853 last argument has low 8 bits clear, fold to 0.
1855 2018-11-14 Iain Sandoe <iain@sandoe.co.uk>
1857 * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
1858 MAX_OFILE_ALIGNMENT as needed.
1860 2018-11-14 Richard Biener <rguenther@suse.de>
1862 PR tree-optimization/88019
1863 * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
1864 COND_EXPR arguments to non-trapping overflow.
1866 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
1868 * config/arc/arc.h (reg_class): Reorder registers classes, remove
1869 unused register classes.
1870 (REG_CLASS_NAMES): Likewise.
1871 (REG_CLASS_CONTENTS): Likewise.
1872 (FIXED_REGISTERS): Make lp_count fixed.
1873 (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
1874 (PROGRAM_COUNTER_REGNO): Remove.
1875 * config/arc/arc.c (arc_conditional_register_usage): Remove unused
1876 register classes, use constants for register numbers, remove
1877 reg_class_contents references.
1878 (arc_process_double_reg_moves): Add asserts.
1879 (arc_secondary_reload): Remove LPCOUNT_REG reference, use
1880 lra_in_progress predicate.
1881 (arc_init_reg_tables): Remove unused register classes.
1882 (arc_register_move_cost): Likewise.
1883 (arc_preferred_reload_class): Likewise.
1884 (hwloop_optimize): Update rtx patterns involving lp_count register.
1885 (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
1886 * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
1887 CHEAP_CORE_REGS. Former one will be used for LRA.
1888 ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS. Former
1889 one will be used for LRA.
1890 ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS. Former
1891 one will be used for LRA.
1892 ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
1893 Former one will be used for LRA.
1894 ("f"): Delete constraint.
1897 ("l"): Change it from register constraint to constraint.
1898 * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
1899 (movhi_insn): Likewise.
1900 (movsi_insn): Update pattern.
1903 (stack_tie): Remove 'b' constraint letter.
1905 (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
1906 (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
1907 (R57_REG, R59_REG, PCL_REG): Likewise.
1908 (ILINK1_REGNUM): Renamed to ILINK1_REG.
1909 (ILINK2_REGNUM): Renamed to ILINK2_REG.
1911 (SP_REGS): Likewise.
1912 (Rcw): Remove unused reg classes.
1913 * config/arc/predicates.md (dest_reg_operand): Just default on
1914 register_operand predicate.
1915 (mpy_dest_reg_operand): Likewise.
1916 (move_dest_operand): Use macros instead of constants.
1917 * config/arc/arc.opt (mlra): Switch to lra as default.
1919 2018-11-14 Richard Biener <rguenther@suse.de>
1921 PR tree-optimization/87974
1922 * tree-vect-loop.c (vectorizable_reduction): When computing
1923 the vectorized reduction PHI vector type ignore constant
1926 2018-11-14 Jakub Jelinek <jakub@redhat.com>
1928 PR tree-optimization/87977
1929 * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
1930 stmt, build a new one and replace the old one with it. Formatting fix.
1931 Call release_ssa_name (x) if !has_other_use and !delete_div.
1932 (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
1933 verify lhs of stmt is still def.
1935 2018-11-13 Peter Bergner <bergner@linux.ibm.com>
1937 PR rtl-optimization/87507
1938 * lower-subreg.c (operand_for_swap_move_operator): New function.
1939 (simple_move): Strip simple operators.
1940 (find_pseudo_copy): Likewise.
1941 (resolve_operand_for_swap_move_operator): New function.
1942 (resolve_simple_move): Strip simple operators and swap operands.
1944 2018-11-13 Jakub Jelinek <jakub@redhat.com>
1946 PR tree-optimization/87898
1947 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
1948 (ipa_simd_modify_function_body): Remove debug stmts where the first
1949 argument was changed into a non-decl.
1951 2018-11-13 Peter Bergner <bergner@linux.ibm.com>
1953 PR rtl-optimization/87899
1954 * lra-lives.c (start_living): Update white space in comment.
1955 (enum point_type): New.
1956 (sparseset_contains_pseudos_p): New function.
1957 (update_pseudo_point): Likewise.
1958 (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
1959 (make_hard_regno_dead): Likewise. Remove ignore_reg_for_conflicts
1960 handling. Move early exit after adding conflicts.
1961 (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro. Add early exit
1962 if regno is already live. Remove all handling of program points.
1963 (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro. Add early exit
1964 after adding conflicts. Remove all handling of program points and
1965 ignore_reg_for_conflicts.
1966 (mark_regno_live): Use HARD_REGISTER_NUM_P macro. Remove return value
1967 and do not guard call to mark_pseudo_live.
1968 (mark_regno_dead): Use HARD_REGISTER_NUM_P macro. Remove return value
1969 and do not guard call to mark_pseudo_dead.
1970 (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
1971 (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
1972 Use new function update_pseudo_point. Handle register copies by
1973 removing the source register from the live set. Handle INOUT operands.
1974 Update to the next program point using the unused_set, dead_set and
1976 (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
1978 2018-11-13 David Malcolm <dmalcolm@redhat.com>
1980 * builtins.c: Replace "source_location" with "location_t".
1981 * diagnostic-show-locus.c: Likewise.
1982 * diagnostic.c: Likewise.
1983 * dumpfile.c: Likewise.
1984 * gcc-rich-location.h: Likewise.
1985 * genmatch.c: Likewise.
1986 * gimple.h: Likewise.
1987 * gimplify.c: Likewise.
1988 * input.c: Likewise.
1989 * input.h: Likewise. Eliminate the typedef.
1990 * omp-expand.c: Likewise.
1991 * selftest.h: Likewise.
1992 * substring-locations.h (get_source_location_for_substring):
1994 (get_location_within_string): ...this.
1995 * tree-cfg.c: Replace "source_location" with "location_t".
1996 * tree-cfgcleanup.c: Likewise.
1997 * tree-diagnostic.c: Likewise.
1998 * tree-into-ssa.c: Likewise.
1999 * tree-outof-ssa.c: Likewise.
2000 * tree-parloops.c: Likewise.
2001 * tree-phinodes.c: Likewise.
2002 * tree-phinodes.h: Likewise.
2003 * tree-ssa-loop-ivopts.c: Likewise.
2004 * tree-ssa-loop-manip.c: Likewise.
2005 * tree-ssa-phiopt.c: Likewise.
2006 * tree-ssa-phiprop.c: Likewise.
2007 * tree-ssa-threadupdate.c: Likewise.
2008 * tree-ssa.c: Likewise.
2009 * tree-ssa.h: Likewise.
2010 * tree-vect-loop-manip.c: Likewise.
2012 2018-11-13 Michael Ploujnikov <michael.ploujnikov@oracle.com>
2014 * doc/extend.texi: Fix typo in the weakref description.
2016 2018-11-13 Richard Biener <rguenther@suse.de>
2018 PR tree-optimization/86991
2019 * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
2020 group building until we have successfully detected the SLP
2022 (vect_is_simple_reduction): Remove fixup code here.
2024 2018-11-13 David Malcolm <dmalcolm@redhat.com>
2026 * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
2027 (dump_gimple_stmt): Use it.
2028 (dump_gimple_stmt_loc): Likewise.
2029 (dump_gimple_expr): Likewise.
2030 (dump_gimple_expr_loc): Likewise.
2031 (dump_generic_expr): Likewise.
2032 (dump_generic_expr_loc): Likewise.
2033 (dump_printf): Likewise.
2034 (dump_printf_loc): Likewise.
2035 (dump_dec): Likewise.
2036 (dump_dec): Likewise.
2037 (dump_hex): Likewise.
2038 (dump_symtab_node): Likewise.
2039 * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
2040 Guard dump call with dump_enabled_p.
2041 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
2042 * graphite-optimize-isl.c (optimize_isl): Likewise.
2043 * graphite.c (graphite_transform_loops): Likewise.
2044 * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
2045 * tree-parloops.c (parallelize_loops): Likewise.
2046 * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
2047 * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
2048 (vect_prune_runtime_alias_test_list): Likewise.
2049 * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
2050 (vect_estimate_min_profitable_iters): Likewise.
2051 * tree-vect-slp.c (vect_record_max_nunits): Likewise.
2052 (vect_build_slp_tree_2): Likewise.
2053 (vect_supported_load_permutation_p): Likewise.
2054 (vect_slp_analyze_operations): Likewise.
2055 (vect_slp_analyze_bb_1): Likewise.
2056 (vect_slp_bb): Likewise.
2057 * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
2058 * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
2059 (pass_slp_vectorize::execute): Likewise.
2060 (increase_alignment): Likewise.
2062 2018-11-13 David Malcolm <dmalcolm@redhat.com>
2065 * ipa-inline.c (report_inline_failed_reason): Guard calls to
2066 cl_target_option_print_diff and cl_optimization_print_diff with
2069 2018-11-13 David Malcolm <dmalcolm@redhat.com>
2071 * doc/invoke.texi (-fsave-optimization-record): Note that the
2072 output is compressed.
2073 * optinfo-emit-json.cc: Include <zlib.h>.
2074 (optrecord_json_writer::write): Compress the output.
2076 2018-11-13 Aldy Hernandez <aldyh@redhat.com>
2078 * tree-vrp.c (value_range_base::dump): Dump type.
2079 Do not use INF nomenclature for 1-bit types.
2080 (dump_value_range): Group all variants to common dumping code.
2081 (debug): New overloaded functions for value_ranges.
2082 (value_range_base::dump): Remove no argument version.
2083 (value_range::dump): Same.
2085 2018-11-13 Richard Biener <rguenther@suse.de>
2087 PR tree-optimization/87931
2088 * tree-vect-loop.c (vect_is_simple_reduction): Restrict
2089 nested cycles we support to latch computations vectorizable_reduction
2092 2018-11-13 Martin Liska <mliska@suse.cz>
2094 PR tree-optimization/87885
2095 * cfghooks.c (account_profile_record): Rename
2097 (profile_record_check_consistency): ... this.
2098 Calculate missing num_mismatched_freq_in.
2099 (profile_record_account_profile): New function
2100 that calculates time and size of a function.
2101 * cfghooks.h (struct profile_record): Remove
2103 (struct cfg_hooks): Remove after_pass flag.
2104 (account_profile_record): Rename to ...
2105 (profile_record_check_consistency): ... this.
2106 (profile_record_account_profile): New.
2107 * cfgrtl.c (rtl_account_profile_record): Remove
2109 * passes.c (check_profile_consistency): Do only
2111 (account_profile): Calculate size and time of
2113 (pass_manager::dump_profile_report): Reformat
2115 (execute_one_ipa_transform_pass): Call
2116 consistency check before clean upand call account_profile
2117 after a clean up is done.
2118 (execute_one_pass): Call check_profile_consistency and
2119 account_profile instead of using after_pass flag..
2120 * tree-cfg.c (gimple_account_profile_record): Likewise.
2122 2018-11-13 Martin Liska <mliska@suse.cz>
2125 * config/i386/i386.c (ix86_option_override_internal): Error
2126 about usage -mabi=ms and -fsanitize={,kernel-}address.
2128 2018-11-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
2130 * config/aarch64/aarch64.c
2131 (aarch64_vectorize_preferred_vector_alignment): Change return type to
2133 (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
2134 alignment being a poly int.
2135 * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
2136 return type to poly_uint64.
2137 * target.def (default_preferred_vector_alignment): Likewise.
2138 * targhooks.c (default_preferred_vector_alignment): Likewise.
2139 * targhooks.h (default_preferred_vector_alignment): Likewise.
2140 * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
2141 (vect_compute_data_ref_alignment): Adapt to vector alignment being a
2143 (vect_update_misalignment_for_peel): Likewise.
2144 (vect_enhance_data_refs_alignment): Likewise.
2145 (vect_find_same_alignment_drs): Likewise.
2146 (vect_duplicate_ssa_name_ptr_info): Likewise.
2147 (vect_setup_realignment): Likewise.
2148 (vect_can_force_dr_alignment_p): Change alignment parameter type to
2150 * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
2151 mask with a compile time variable vector alignment.
2152 (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
2154 (vect_do_peeling): Exit early if vector alignment is not constant.
2155 * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
2157 (vectorizable_store): Likewise.
2158 (vectorizable_load): Likweise.
2159 * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
2161 (vect_known_alignment_in_bytes): Adapt to vector alignment being a
2163 (vect_can_force_dr_alignment_p): Change alignment parameter type to
2166 2018-11-13 Richard Biener <rguenther@suse.de>
2168 PR tree-optimization/87962
2169 * tree-vect-loop.c (vect_is_simple_reduction): More reliably
2170 detect outer reduction for disqualifying in-loop uses.
2172 2018-11-13 Richard Biener <rguenther@suse.de>
2174 PR tree-optimization/87967
2175 * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
2176 for constants for the scalar loop.
2178 2018-11-13 Alan Modra <amodra@gmail.com>
2180 * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
2181 integer constants when -mcmodel=medium.
2183 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
2185 * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
2186 * config/arc/elf.h (PROFILE_HOOK): Define.
2187 * config/arc/linux.h (PROFILE_HOOK): Likewise.
2189 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
2191 * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
2192 jump table data in the text section.
2194 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
2196 * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
2197 to fit the eh_return pattern.
2198 * config/arc/arc.md (eh_return): Define.
2199 (VUNSPEC_ARC_EH_RETURN): Likewise.
2200 * config/arc/arc-protos.h (arc_eh_return_address_location): Match
2202 * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
2205 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
2207 * common/config/arc/arc-common.c (arc_option_optimization_table):
2208 Millicode optimization is default on for size optimizations.
2209 * config/arc/arc-protos.h (arc_check_multi): New function.
2210 * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
2211 (ENTER_LEAVE_START_REG): Define.
2212 (ENTER_LEAVE_END_REG): Likewise.
2213 (arc_override_options): Disable millicode when long calls option
2215 (arc_frame_info): Change it from int to bool.
2216 (arc_compute_frame_size): Clean up.
2217 (arc_save_restore): Remove.
2218 (frame_save_reg): New function.
2219 (frame_restore_reg): Likewise.
2220 (arc_enter_leave_p): Likewise.
2221 (arc_save_callee_saves): Likewise.
2222 (arc_restore_callee_saves): Likewise.
2223 (arc_save_callee_enter): Likewise.
2224 (arc_restore_callee_leave): Likewise.
2225 (arc_save_callee_milli): Likewise.
2226 (arc_restore_callee_milli): Likewise.
2227 (arc_expand_prologue): Reimplement to emit enter/leave
2229 (arc_expand_epilogue): Likewise.
2230 (arc_check_multi): New function.
2231 * config/arc/arc.md (push_multi_fp): New pattern.
2232 (push_multi_fp_blink): Likewise.
2233 (pop_multi_fp): Likewise.
2234 (pop_multi_fp_blink): Likewise.
2235 (pop_multi_fp_ret): Likewise.
2236 (pop_multi_fp_blink_ret): Likewise.
2237 * config/arc/arc.opt (mmillicode): Update option.
2238 (mcode-density-frame): New option.
2239 * config/arc/predicates.md (push_multi_operand): New predicate.
2240 (pop_multi_operand): Likewise.
2241 * doc/invoke.texi (ARC): Update ARC options information.
2244 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
2246 * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
2247 * config/arc/arc.c (operands_ok_ldd_std): New function.
2248 (mem_ok_for_ldd_std): Likewise.
2249 (gen_operands_ldd_std): Likewise.
2250 * config/arc/arc.md: Add peephole2 rules for std/ldd.
2252 2018-11-13 Eric Botcazou <ebotcazou@adacore.com>
2254 * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
2255 test on stack_usage_file.
2256 (lang_dependent_init): Do not open the .su file if generating LTO.
2258 2018-11-13 Jakub Jelinek <jakub@redhat.com>
2260 PR rtl-optimization/87918
2261 * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
2262 simplify_gen_relational rather than simplify_gen_binary.
2264 2018-11-13 Richard Biener <rguenther@suse.de>
2266 * tree-ssanames.h (set_range_info): Use value_range_base.
2267 (get_range_info): Likewise.
2268 * tree-ssanames.c (set_range_info): Likewise.
2269 (get_range_info): Likewise.
2270 * tree-vrp.c (value_range_base::union_helper): Split
2271 out common parts of value_range[_base]::union_.
2272 (value_range_base::union_): Update.
2273 (value_range::union_): Likewise.
2274 (determine_value_range_1): Use value_range_base.
2275 (determine_value_range): Likewise.
2276 * tree-vrp.h (value_range_base::union_helper): Move ...
2277 (value_range::union_helper): ... from here.
2279 2018-11-13 Alan Modra <amodra@gmail.com>
2281 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
2283 (rs6000_secondary_reload_gpr): Don't call find_replacement.
2285 2018-11-13 Sandra Loosemore <sandra@codesourcery.com>
2288 * doc/invoke.texi (Optimize Options): Clarify that the
2289 l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
2290 apply to data cache size.
2292 2018-11-13 Alan Modra <amodra@gmail.com>
2294 * config/rs6000/predicates.md (logical_const_operand),
2295 (logical_operand): Correct comment.
2296 * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
2298 2018-11-13 Alan Modra <amodra@gmail.com>
2300 * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
2302 2018-11-13 Alan Modra <amodra@gmail.com>
2304 * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
2306 * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
2307 covered by alternative.
2308 (movcc_internal1): Ignore h for register preference.
2309 (mov<mode>_hardfloat64): Likewise.
2310 (mov<mode>_softfloat): Ignore c, l, h for register preference.
2312 2018-11-12 Sandra Loosemore <sandra@codesourcery.com>
2314 PR preprocessor/47823
2315 * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
2318 * doc/extend.texi (Pragmas): Note additional pragmas documented
2321 2018-11-13 Kugan Vivekanandarajah <kuganv@linaro.org>
2325 * tree-scalar-evolution.c (expression_expensive_p): Make
2326 BUILTIN POPCOUNT as expensive when backend does not define it.
2328 2018-11-12 Fredrik Noring <noring@nocrew.org>
2330 * config.gcc: Update with-llsc defaults for MIPS r5900.
2332 2018-11-12 Martin Liska <mliska@suse.cz>
2334 PR gcov-profile/87442
2335 * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
2337 * doc/invoke.texi: Document them.
2338 * tree-profile.c (parse_profile_filter): New.
2339 (parse_profile_file_filtering): Likewise.
2340 (release_profile_file_filtering): Likewise.
2341 (include_source_file_for_profile): Likewise.
2342 (tree_profiling): Filter source files based on the
2343 newly added options.
2345 2018-11-12 Martin Liska <mliska@suse.cz>
2348 * doc/extend.texi: Add missing values for __builtin_cpu_is and
2349 __builtin_cpu_supports for x86 target.
2351 2018-11-12 Sandra Loosemore <sandra@codesourcery.com>
2354 * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
2356 (Arithmetic): Likewise for "lo_sum".
2358 2018-11-12 Renlin Li <renlin.li@arm.com>
2361 * dse.c (get_stored_val): Add check for compile-time constantness
2364 2018-11-12 Sudakshina Das <sudi.das@arm.com>
2366 * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
2367 (ARMv8_5a): New fgroup.
2368 (armv8.5-a): New arch.
2369 (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
2371 * config/arm/arm-tables.opt: Regenerate.
2372 * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
2373 * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
2374 * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
2375 Add matching rules for -march=armv8.5-a and extensions.
2376 * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
2377 Add sb and predres to all armv8-a except armv8.5-a.
2379 2018-11-12 Richard Biener <rguenther@suse.de>
2381 * tree-vrp.h (value_range[_base]::set): Make public. Provide
2382 overload for single value.
2383 (value_range[_base]::set_nonnull): New.
2384 (value_range[_base]::set_null): Likewise.
2385 (value_range): Document bitmap copying behavior, mark
2386 copy constructor and assignment operator deleted.
2387 (value_range::move): New.
2388 (value_range::set_and_canonicalize): Default bitmap to zero.
2389 (set_value_range_to_nonnull): Remove.
2390 (set_value_range_to_null): Likewise.
2391 (set_value_range): Likewise.
2392 (set_value_range_to_value): Likewise.
2393 (extract_range_from_unary_expr): Work on value_range_base.
2394 (extract_range_from_binary_expr_1): Likewise. Rename to...
2395 (extract_range_from_binary_expr): ... this.
2396 * tree-vrp.c (value_range::update): Clear equiv bitmap
2398 (value_range::move): New, move equiv bitmap.
2399 (value_range_base::set_undefined): Avoid assignment.
2400 (value_range::set_undefined): Likewise.
2401 (value_range_base::set_varying): Likewise.
2402 (value_range::set_varying): Likewise.
2403 (set_value_range): Remove.
2404 (value_range_base::set): New overload for value.
2405 (value_range::set): Likewise.
2406 (set_value_range_to_nonnull): Remove.
2407 (value_range_base::set_nonnull): New.
2408 (value_range::set_nonnull): Likewise.
2409 (set_value_range_to_null): Remove.
2410 (value_range_base::set_null): New.
2411 (value_range::set_null): Likewise.
2412 (range_is_null): Work on value_range_base.
2413 (range_is_nonnull): Likewise.
2414 (ranges_from_anti_range): Likewise.
2415 (extract_range_into_wide_ints): Likewise.
2416 (extract_range_from_multiplicative_op): Likewise.
2417 (extract_range_from_binary_expr): Likewise. Update for API changes.
2418 (extract_range_from_unary_expr): Likewise. Remove OBJ_TYPE_REF
2420 (value_range::intersect_helper): Avoid copy and assignment.
2421 (value_range::union_helper): Likewise.
2422 (determine_value_range_1): Adjust.
2423 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
2424 Avoid assignment by using move.
2425 (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
2426 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
2428 * tree-ssanames.c (get_range_info): Likewise.
2429 * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
2430 * vr-values.c (vr_values::get_value_range): Adjust.
2431 (vr_values::update_value_range): Likewise.
2432 (symbolic_range_based_on_p): Work on value_range_base.
2433 (vr_values::extract_range_from_binary_expr): Use value_range_base.
2434 (vr_values::extract_range_from_unary_expr): Likewise.
2435 (vr_values::extract_range_from_cond_expr): Avoid assignment.
2436 (vr_values::extract_range_from_comparison): Adjust.
2437 (vr_values::check_for_binary_op_overflow): Use value_range_base.
2438 (vr_values::extract_range_basic): Adjust.
2439 (vr_values::adjust_range_with_scev): Likewise.
2440 (vr_values::vrp_visit_assignment_or_call): Likewise.
2441 (vr_values::get_vr_for_comparison): Change API to avoid
2442 assignment and copy construction.
2443 (vr_values::compare_name_with_value): Adjust accordingly.
2444 (vr_values::compare_names): Likewise.
2445 (vr_values::extract_range_from_phi_node): Avoid assignment and
2446 bogus in-place modify of equiv bitmap.
2447 (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
2448 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
2449 for extract_range_from_unary_expr API change.
2450 * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
2452 2018-11-12 Eric Botcazou <ebotcazou@adacore.com>
2454 * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
2456 2018-11-12 Richard Biener <rguenther@suse.de>
2458 * tree-vrp.h (value_range_base::symbolic_p,
2459 value_range_base::constant_p, value_range_base::zero_p,
2460 value_range_base::singleton_p): Move from value_range.
2461 (value_range::dump): Add.
2462 * gimple-ssa-evrp-analyze.c
2463 (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
2464 * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
2465 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
2467 * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
2468 (value_range::constant_p): Likewise.
2469 (value_range::singleton_p): Likewise.
2470 (value_range_base::dump): Add.
2471 (set_value_range_to_undefined): Remove.
2472 (set_value_range_to_varying): Likewise.
2473 (range_int_cst_p): Take value_range_base argument.
2474 (range_int_cst_singleton_p): Likewise.
2475 (value_range_constant_singleton): Likewise.
2476 (vrp_set_zero_nonzero_bits): Likewise.
2477 (extract_range_from_multiplicative_op): Use set_varying.
2478 (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
2479 (extract_range_from_unary_expr): Likewise.
2480 (dump_value_range_base): Change to overload of dump_value_range.
2481 (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
2482 (vrp_prop::visit_stmt): Likewise.
2483 (value_range::intersect_helper): Likewise.
2484 (value_range::union_helper): Likewise.
2485 (determine_value_range_1): Likewise.
2487 2018-11-12 Richard Biener <rguenther@suse.de>
2489 * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
2490 (set_value_range_to_null): Likewise.
2491 * vr-values.c (vr_values::extract_range_from_comparison):
2492 Clear equiv for constant singleton ranges.
2494 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
2496 * config/i386/sse.md: Combine VFIXUPIMM* patterns
2497 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2498 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2499 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
2500 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2501 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2502 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
2504 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
2507 * doc/extend.texi (Common Type Attributes): For the align type
2508 attribute, copy language about decreasing alignment from the
2509 corresponding variable attribute.
2511 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
2513 * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
2516 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
2519 * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
2520 -frtti and -fno-rtti code.
2522 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
2525 * doc/extend.texi (Other Builtins): Document probability associated
2526 with __builtin_expect.
2528 2018-11-11 Uros Bizjak <ubizjak@gmail.com>
2531 * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
2532 instead of (TARGET_64BIT && ix86_abi == MS_ABI).
2533 * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
2534 * config/i386/cygming.h (STACK_BOUNDARY): Remove.
2536 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
2538 * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
2540 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
2542 * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
2544 2018-11-11 Richard Biener <rguenther@suse.de>
2546 * tree-vrp.h (class value_range_base): New base class for
2547 value_range containing all but the m_equiv member.
2548 (dump_value_range_base): Add.
2549 (range_includes_zero_p): Work on value_range_base.
2550 * tree-vrp.c (value_range_base::set): Split out base handling
2552 (value_range::set): this.
2553 (value_range::set_equiv): New.
2554 (value_range_base::value_range_base): New constructors.
2555 (value_range_base::check): Split out base handling from...
2556 (value_range::check): this.
2557 (value_range::equal_p): Refactor in terms of
2558 ignore_equivs_equal_p which is now member of the base.
2559 (value_range_base::set_undefined): New.
2560 (value_range_base::set_varying): Likewise.
2561 (value_range_base::dump):Split out base handling from...
2562 (value_range::dump): this.
2563 (value_range_base::set_and_canonicalize): Split out base handling
2565 (value_range::set_and_canonicalize): this.
2566 (value_range_base::union_): New.
2567 * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
2569 * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
2570 instead of value_range everywhere.
2571 (ipcp_vr_lattice::print): Use dump_value_range_base.
2572 (ipcp_vr_lattice::meet_with): Adjust.
2573 (ipcp_vr_lattice::meet_with_1): Likewise.
2574 (ipa_vr_operation_and_type_effects): Likewise.
2575 (propagate_vr_across_jump_function): Likewise.
2576 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
2577 (ipa_get_value_range): Likewise.
2578 (ipa_set_jfunc_vr): Likewise.
2579 (ipa_compute_jump_functions_for_edge): Likewise.
2581 2018-11-10 Sandra Loosemore <sandra@codesourcery.com>
2584 * doc/invoke.texi (Optimize Options): Add @opindex entries
2585 for the positive forms of -fno-xxx and -mno-xxx options
2586 that were lacking them.
2588 2018-11-10 Segher Boessenkool <segher@kernel.crashing.org>
2590 * combine.c (make_more_copies): Only make an intermediate copy if the
2591 dest of a move is a pseudo.
2593 2018-11-09 Maya Rashish <coypu@sdf.org>
2596 * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
2597 (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
2599 2018-11-09 Sandra Loosemore <sandra@codesourcery.com>
2603 * doc/invoke.texi (Optimize Options): Clarify default behavior
2604 for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
2606 2018-11-09 Bernd Edlinger <bernd.edlinger@hotmail.de>
2608 PR tree-optimization/87940
2609 * expr.c (string_constant): Don't strip NOPS in subexpressions.
2610 Fold PLUS_EXPR correctly.
2612 2018-11-09 Ilya Leoshkevich <iii@linux.ibm.com>
2615 * config/s390/s390.c (s390_safe_relative_long_p): New function.
2616 (annotate_constant_pool_refs): Skip insns which support
2617 relative addressing.
2618 (annotate_constant_pool_refs_1): New helper function.
2619 (find_constant_pool_ref): Skip insns which support relative
2621 (find_constant_pool_ref_1): New helper function.
2622 (replace_constant_pool_ref): Skip insns which support
2623 relative addressing.
2624 (replace_constant_pool_ref_1): New helper function.
2625 (s390_mainpool_start): Adapt to the new signature.
2626 (s390_mainpool_finish): Likewise.
2627 (s390_chunkify_start): Likewise.
2628 (s390_chunkify_finish): Likewise.
2629 (pass_s390_early_mach::execute): Likewise.
2630 (s390_prologue_plus_offset): Likewise.
2631 (s390_emit_prologue): Likewise.
2632 (s390_emit_epilogue): Likewise.
2634 2018-11-09 Jakub Jelinek <jakub@redhat.com>
2636 * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
2637 but unsupported lastprivate with conditional modifier.
2639 2018-11-09 Jeff Law <law@redhat.com>
2641 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
2642 unused argument better. Add gcc_unreachable to silence warning.
2644 2018-11-09 Martin Sebor <msebor@redhat.com>
2647 * attribs.c (has_attribute): New helper function.
2648 (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
2649 * attribs.h (decls_mismatched_attributes): Declare.
2650 * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
2651 (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
2652 * common.opt (-Wattribute-alias): Take an argument.
2653 (-Wno-attribute-alias): New option.
2654 * doc/extend.texi (Common Function Attributes): Document copy.
2655 (Common Variable Attributes): Same.
2656 * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
2657 (-Wattribute-alias): Document new option argument.
2659 2018-11-09 Richard Earnshaw <rearnsha@arm.com>
2661 * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
2662 matching criteria. Remove unused array initializer.
2664 2018-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
2665 Jinsong Ji <jji@us.ibm.com>
2667 * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
2668 constraints by introducing a new temporary.
2669 (_mm_cvtss_si64): Likewise.
2671 2018-11-09 Martin Liska <mliska@suse.cz>
2673 * common.opt: Add -fipa-stack-alignment flag.
2674 * doc/invoke.texi: Document it.
2675 * final.c (rest_of_clean_state): Guard stack
2676 shrinking with flag.
2678 2018-11-09 Martin Liska <mliska@suse.cz>
2680 * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
2682 (ipa_discover_variable_flags): ... this.
2683 * common.opt: Come up with new flag -fipa-reference-addressable.
2684 * doc/invoke.texi: Document it.
2685 * ipa-reference.c (propagate): Call the renamed fn.
2686 * ipa-visibility.c (whole_program_function_and_variable_visibility):
2688 * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
2690 (ipa_discover_variable_flags): ... this. Discover
2691 non-addressable variables only with the newly added flag.
2692 * opts.c: Enable the newly added flag with -O1 and higher
2695 2018-11-09 David Malcolm <dmalcolm@redhat.com>
2697 * json.cc (selftest::test_writing_literals): Fix comment.
2699 2018-11-09 Martin Liska <mliska@suse.cz>
2701 * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
2702 string to a stack buffer.
2703 (aarch64_parse_cpu): Likewise.
2704 (aarch64_parse_tune): Likewise.
2706 2018-11-09 Richard Biener <rguenther@suse.de>
2708 PR tree-optimization/87953
2709 * tree-vect-loop.c (vectorizable_reduction): For analysis
2710 always pass ops[0] to vectorizable_condition.
2712 2018-11-09 Stafford Horne <shorne@gmail.com>
2713 Richard Henderson <rth@twiddle.net>
2714 Joel Sherrill <joel@rtems.org>
2716 * common/config/or1k/or1k-common.c: New file.
2717 * config/or1k/*: New.
2718 * config.gcc (or1k*-*-*): New.
2719 * configure.ac (or1k*-*-*): New test for openrisc tls.
2720 * configure: Regenerated.
2721 * doc/install.texi: Document OpenRISC triplets.
2722 * doc/invoke.texi: Document OpenRISC arguments.
2723 * doc/md.texi: Document OpenRISC.
2725 2018-11-09 Richard Earnshaw <rearnsha@arm.com>
2727 * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
2728 (arm7tdmi-s): Delete CPU.
2729 (arm710t): Add aliases for arm720t and arm740t.
2730 (arm720t, arm740t): Delete CPUs.
2731 (arm920t): Add aliases for arm920, arm922t and arm940t.
2732 (arm920, arm922t, arm940t): Delete CPUs.
2733 (arm10tdmi): Add alias for arm1020t.
2734 (arm1020t): Delete CPU.
2735 (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
2736 (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
2737 (arm10e): Add aliases for arm1020e and arm1022e.
2738 (arm1020e, arm1022e): Delete CPU.
2739 * config/arm/arm.md (generic_sched): Remove entries that are now
2741 (generic_vfp): Likewise.
2742 * config/arm/arm1020e.md: Simplify tuning selection based on alias
2744 * config/arm/arm-tune.md: Regenerated.
2745 * config/arm/arm-tables.opt: Regenerated.
2747 2018-11-09 Richard Biener <rguenther@suse.de>
2749 PR tree-optimization/87621
2750 * tree-vect-loop.c (vectorizable_reduction): Handle reduction
2751 op with only phi inputs.
2752 * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
2753 (ch_base::copy_headers): Run CSE on copied loop headers.
2754 (pass_ch_vect::process_loop_p): Simplify.
2756 2018-11-09 Alexandre Oliva <oliva@adacore.com>
2758 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
2759 for non-w64 x86_64 biarch.
2761 2018-11-09 Alexandre Oliva <aoliva@redhat.com>
2763 PR rtl-optimization/86438
2764 * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
2765 of in_b for the compare if in_b is SET_DEST.
2768 * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
2769 non-toplevel UNSPEC.
2771 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
2773 * tree-vrp.c (value_range::check): Do not access internals
2775 (value_range::singleton_p): Same.
2776 (value_range::type): Same.
2777 (vrp_finalize): Use value_range API.
2779 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
2781 * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
2783 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
2785 * vr-values.c (vr_values::get_value_range): Use value_range API
2786 instead of piecing together ranges.
2787 (vr_values::update_value_range): Same.
2789 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
2791 * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
2792 of performing ad-hoc calculations.
2793 * tree-ssanames.c (set_range_info): New overloaded function
2794 accepting value_range &.
2795 (get_range_info): Same.
2796 * tree-ssanames.h (set_range_info_raw): Remove.
2797 (set_range_info): New prototype.
2798 (get_range_info): Same.
2799 * tree-vrp.h (value_range::null_p): Rename to zero_p.
2800 * tree-vrp.c (value_range::null_p): Same.
2802 2018-11-09 Jan Hubicka <jh@suse.cz>
2804 * tree.c (fld_type_variant_equal_p): Test user align flag.
2805 (flt_type_variant): Copy user align flag.
2806 (fld_incomplete_type_of): Clear it.
2808 2018-11-09 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2810 * config/arm/neon.md (div<mode>3): New pattern.
2812 2018-11-08 Andi Kleen <ak@linux.intel.com>
2814 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
2815 (OPTION_MASK_ISA_PTWRITE_UNSET): New.
2816 (ix86_handle_option): Handle OPT_mptwrite.
2817 * config/i386/cpuid.h (bit_PTWRITE): Add.
2818 * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
2819 * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
2820 * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
2821 * config/i386/i386.c (ix86_target_string): Handle ptwrite.
2822 (ix86_option_override_internal): Handle PTA_PTWRITE.
2823 (ix86_valid_target_attribute_inner_p): Define ptwrite.
2824 (def_builtin2): Force UINT64 to be 64bit only.
2825 * config/i386/i386.h (TARGET_PTWRITE): Add.
2826 (TARGET_PTWRITE_P): Add.
2828 * config/i386/i386.md: Define ptwrite.
2829 * config/i386/i386.opt: Add -mptwrite.
2830 * config/i386/immintrin.h (_ptwrite64): Add.
2832 * doc/extend.texi: Document __builtin_ia32_ptwrite*.
2833 * doc/invoke.texi: Document -mptwrite.
2835 2018-11-08 Peter Bergner <bergner@linux.ibm.com>
2837 PR rtl-optimization/87600
2838 * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
2839 * lra-constraints.c (process_alt_operands): Skip illegal hard
2840 register usage. Prefer reloading non hard register operands.
2842 2018-11-08 Sandra Loosemore <sandra@codesourcery.com>
2845 * doc/invoke.texi (Optimize Options): Clarify default behavior
2846 for -fno-sched-interblock and -fno-sched-spec.
2848 2018-11-08 Roman Geissler <roman.geissler@amadeus.com>
2850 * collect2.c (linker_select): Add USE_LLD_LD.
2851 (ld_suffixes): Add ld.lld.
2852 (main): Handle -fuse-ld=lld.
2853 * common.opt (-fuse-ld=lld): New option.
2854 * doc/invoke.texi (-fuse-ld=lld): Document.
2855 * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
2857 2018-11-08 Paul Koning <ni1d@arrl.net>
2859 * config/pdp11/constraints.md: Add "Z" series constraints for use
2860 with pre-dec and post-inc addressing.
2861 * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
2862 (pdp11_expand_operands): Add int argument (word count).
2863 (pdp11_sp_frame_offset): Delete.
2864 (pdp11_cmp_length): New function.
2865 (pushpop_regeq): New function.
2866 * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
2868 (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
2870 (pdp11_initial_elimination_offset): Ditto.
2871 (pdp11_expand_operands): Add word count argument. Bugfixes.
2872 (output_move_multiple): Change how pointer adjustment is done.
2873 (pdp11_gen_int_label): Correct format.
2874 (output_ascii): Ditto.
2875 (pdp11_asm_output_var): Add code for DEC assembler case.
2876 (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
2878 (legitimate_const_double_p): Ditto.
2879 (pdp11_register_move_cost): Adjust for new register classes.
2880 (pdp11_regno_reg_class): Ditto.
2881 (expand_block_move): Delete.
2882 (pushpop_regeq): New function.
2883 (pdp11_legitimate_address_p): Bugfix in check for constant
2885 (pdp11_sp_frame_offset): Delete.
2886 (pdp11_reg_save_size): New helper function for new frame layout.
2887 (output_addr_const_pdp11): Remove CONST_DOUBLE case.
2888 (pdp11_expand_shift): Bugfix in check for constant shift count.
2889 (pdp11_shift_length): Ditto.
2890 (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
2891 (pdp11_cmp_length): New function.
2892 * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
2893 some compile options.
2894 (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
2895 (CALL_USED_REGISTERS): Ditto.
2896 (ELIMINABLE_REGS): Ditto.
2897 (REGISTER_NAMES): Ditto.
2898 (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
2900 (REG_CLASS_NAMES): Ditto.
2901 (REG_CLASS_CONTENTS): Ditto. Also remove
2902 HARD_FRAME_POINTER_REGNUM.
2903 (CPU_REG_CLASS): New macro.
2904 (CLASS_MAX_NREGS): Adjust for new register classes.
2905 (FUNCTION_PROFILER): Make no-op.
2906 (may_call_alloca): Remove unused declaration.
2907 (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
2908 (ASM_OUTPUT_SKIP): Fix format.
2909 * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
2910 (HARD_FRAME_POINTER_REGNUM): Remove.
2912 (*rts): Rename. Remove epilogue related checks.
2913 (cmpsi, cmpdi): New insn.
2914 (cbranch<mode>4): Change to apply to SI and DI modes as well.
2915 (mov<mode>): Change constraints to enforce that push/pop
2916 destination cannot use the same register as source.
2917 (*mov<mode><cc_cc>): Ditto.
2918 (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
2919 at assembly output rather than as RTL expander.
2920 (zero_extendqihi2): Bugfix in check for same registers.
2921 (adddi3_nocc): Bugfix in check for constant operand.
2922 (addsi3_nocc): Ditto.
2923 (subdi3_nocc): Ditto.
2924 (subsi3_nocc): Ditto.
2925 (negdi2_nocc): Copy input to pdp11_expand_operands.
2926 (negsi2_nocc): Ditto.
2927 (bswap2_nocc): Ditto.
2928 * config/pdp11/pdp11.opt (mlra): Fix documentation.
2929 * config/pdp11/t-pdp11: Use -Os.
2931 2018-11-08 Richard Earnshaw <rearnsha@arm.com>
2933 * config/arm/parsecpu.awk (/alias/): New parsing rule.
2934 (/begin cpu/): Check that the cpu name hasn't been previously defined.
2935 (gen_comm_data): Print out CPU alias tables.
2936 (check_cpu): Match aliases when checking the CPU name.
2937 * config/arm/arm-protos.h (cpu_alias): New structure.
2938 (cpu_option): Add entry for aliases.
2939 * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
2940 strongarm1100 and strongarm1110.
2941 (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
2942 (config/arm/arm-generic.md): Remove redundant references to
2943 strongarm110, strongarm1100 and strongarm1110.
2944 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
2945 Scan aliases for additional hints.
2946 (arm_parse_cpu_option_name): Also match a cpu name against the list
2948 * config/arm/arm-tables.opt: Regenerated.
2949 * config/arm/arm-tune.md: Regenerated.
2951 2018-11-08 Jakub Jelinek <jakub@redhat.com>
2953 * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
2954 BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
2955 BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
2956 BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
2957 BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
2958 BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
2959 * gengtype.c (open_base_files): Add omp-general.h.
2960 * gimple.c (gimple_build_omp_critical):
2961 (gimple_build_omp_taskgroup): Add CLAUSES argument. Call
2962 gimple_omp_taskgroup_set_clauses.
2963 (gimple_build_omp_atomic_load): Add mo argument, call
2964 gimple_omp_atomic_set_memory_order.
2965 (gimple_build_omp_atomic_store): Likewise.
2966 (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
2967 * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
2969 (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
2970 of GSS_OMP_SINGLE_LAYOUT, adjust comments.
2971 * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
2972 and GF_OMP_ATOMIC_MEMORY_ORDER. Remove GF_OMP_ATOMIC_SEQ_CST, use
2973 different value for GF_OMP_ATOMIC_NEED_VALUE.
2974 (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
2976 (struct gimple_statement_omp_single_layout): And remove here.
2977 (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
2978 than gimple_statement_omp_single_layout.
2979 (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
2981 (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
2982 (gimple_omp_subcode): Formatting fix.
2983 (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
2984 gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
2985 gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
2986 gimple_omp_teams_host, gimple_omp_teams_set_host,
2987 gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
2988 gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
2989 gimple_omp_taskgroup_set_clauses): New inline functions.
2990 (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
2991 (gimple_build_omp_atomic_store): Likewise.
2992 (gimple_omp_atomic_seq_cst_p): Remove.
2993 (gimple_omp_atomic_memory_order): New function.
2994 (gimple_omp_atomic_set_seq_cst): Remove.
2995 (gimple_omp_atomic_set_memory_order): New function.
2996 (gimple_build_omp_taskgroup): Add clauses argument.
2997 * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
2998 (dump_gimple_omp_task): Print taskwait with depend clauses.
2999 (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
3000 dump_omp_atomic_memory_order.
3001 (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
3002 * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
3003 GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
3004 (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
3005 renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
3006 ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
3007 ORT_UNTIED_TASKLOOP enumerators.
3008 (enum gimplify_defaultmap_kind): New.
3009 (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
3010 target_map_pointers_as_0len_arrays members, add defaultmap.
3011 (new_omp_context): Initialize defaultmap member.
3012 (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
3013 (maybe_fold_stmt): Don't fold even in host teams regions.
3014 (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
3015 ORT_WORKSHARE. Test ctx->defaultmap[GDMK_SCALAR] instead of
3016 ctx->omp_firstprivatize_variable.
3017 (omp_add_variable): Don't add private/firstprivate for VLAs in
3019 (omp_default_clause): Print "taskloop" rather than "task" if
3021 (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
3022 Handle new defaultmap clause kinds.
3023 (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE. Allow simd
3024 iterator to be lastprivate or private. Fix up diagnostics if linear
3025 is used on collapse>1 simd iterator.
3026 (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
3027 (gimplify_omp_depend): New function.
3028 (gimplify_scan_omp_clauses): Add shared clause on parallel for
3029 combined parallel master taskloop{, simd} if taskloop has
3030 firstprivate, lastprivate or reduction clause. Handle
3031 OMP_CLAUSE_REDUCTION_TASK diagnostics. Adjust tests for
3032 ORT_COMBINED_TEAMS. Gimplify depend clauses with iterators. Handle
3033 cancel and simd OMP_CLAUSE_IF_MODIFIERs. Handle
3034 OMP_CLAUSE_NONTEMPORAL. Handle new defaultmap clause kinds. Handle
3035 OMP_CLAUSE_{TASK,IN}_REDUCTION. Diagnose invalid conditional
3037 (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL. Handle
3038 GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
3039 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL. Handle
3040 OMP_CLAUSE_{TASK,IN}_REDUCTION.
3041 (gimplify_omp_task): Handle taskwait with depend clauses.
3042 (gimplify_omp_for): Add shared clause on parallel for combined
3043 parallel master taskloop{, simd} if taskloop has firstprivate,
3044 lastprivate or reduction clause. Use ORT_TASKLOOP or
3045 ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK. Adjust
3046 tests for ORT_COMBINED_TEAMS. Handle C++ range for loops with
3047 NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS. Firstprivatize
3048 __for_end and __for_range temporaries on OMP_PARALLEL for
3049 distribute parallel for{, simd}. Move OMP_CLAUSE_REDUCTION
3050 and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
3051 sandwiched in between two taskloops.
3052 (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
3053 instead of ctx->omp_firstprivatize_variable.
3054 (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
3055 ORT_COMBINED_HOST_TEAMS if not inside of target construct. If
3056 host teams, use gimplify_and_return_first etc. for body like
3057 for target or target data constructs, and at the end call
3058 gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
3059 (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
3060 of OMP_ATOMIC_SEQ_CST, pass it as new argument to
3061 gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
3062 gimple_omp_atomic_set_seq_cst calls.
3063 (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
3064 case, handle taskgroup clauses.
3065 * lto-streamer-out.c (hash_tree): Handle
3066 OMP_CLAUSE_{TASK,IN}_REDUCTION.
3067 * Makefile.in (GTFILES): Add omp-general.h.
3068 * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
3069 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
3070 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
3071 BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
3072 BUILT_IN_GOMP_LOOP_DOACROSS_START,
3073 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
3074 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
3075 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
3076 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
3077 BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
3078 BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
3079 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
3080 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
3081 BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
3082 BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
3083 BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
3084 BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
3085 BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
3086 BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
3087 BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
3088 * omp-expand.c (workshare_safe_to_combine_p): Return false for
3089 non-worksharing loops.
3090 (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
3091 (determine_parallel_type): Don't combine parallel with worksharing
3092 which has _reductemp_ clause.
3093 (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
3094 GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
3095 if there is nonmonotonic modifier or if there is no modifier and no
3096 ordered clause. For dynamic and guided schedule without monotonic
3097 and nonmonotonic modifier, default to nonmonotonic.
3098 (expand_omp_for): Likewise. Adjust expand_omp_for_generic caller, use
3099 GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
3101 (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
3102 there are any reduction clauses.
3103 (expand_taskwait_call): New function.
3104 (expand_teams_call): New function.
3105 (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
3106 expand_teams_call for it. Formatting fix. Handle taskwait with
3108 (expand_omp_for_generic): Add SCHED_ARG argument. Handle expansion
3109 of worksharing loops with task reductions.
3110 (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
3111 expansion of worksharing loops with task reductions.
3112 (expand_omp_sections): Handle expansion of sections with task
3114 (expand_omp_synch): For host teams call expand_omp_taskreg.
3115 (omp_memory_order_to_memmodel): New function.
3116 (expand_omp_atomic_load, expand_omp_atomic_store,
3117 expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
3118 instead of gimple_omp_atomic_seq_cst_p.
3119 (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
3120 depend clauses as a standalone directive.
3121 * omp-general.c (enum omp_requires): New variable.
3122 (omp_extract_for_data): Initialize have_reductemp member. Allow
3123 NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
3124 GT_EXPR loops depending on incr sign. Formatting fixes.
3125 * omp-general.h (struct omp_for_data): Add have_reductemp member.
3126 (enum omp_requires): New enum.
3127 (omp_requires_mask): Declare.
3128 * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
3130 * omp-low.c (struct omp_context): Add task_reductions and
3131 task_reduction_map fields.
3132 (is_host_teams_ctx): New function.
3133 (is_taskreg_ctx): Return true also if is_host_teams_ctx.
3134 (use_pointer_for_field): Use is_global_var instead of
3135 TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
3137 (build_outer_var_ref): Ignore taskgroup outer contexts.
3138 (delete_omp_context): Release task_reductions and task_reduction_map.
3139 (scan_sharing_clauses): Don't add any fields for reduction clause on
3140 taskloop. Handle OMP_CLAUSE__REDUCTEMP_. Handle
3141 OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
3142 modifier. Don't ignore shared clauses in is_host_teams_ctx contexts.
3143 Handle OMP_CLAUSE_NONTEMPORAL.
3144 (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
3146 (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
3147 clauses with task modifier.
3148 (scan_omp_task): Handle taskwait with depend clauses.
3149 (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
3150 first. Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
3151 Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
3152 (scan_omp_for): Fix comment formatting.
3153 (scan_omp_teams): Handle host teams constructs.
3154 (check_omp_nesting_restrictions): Allow teams with no outer
3155 OpenMP context. Adjust diagnostics for teams strictly nested into
3156 some explicit OpenMP construct other than target. Allow OpenMP atomics
3157 inside of simd regions.
3158 (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
3159 (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
3160 taskreg_nesting_level while scanning host teams construct.
3161 (task_reduction_read): New function.
3162 (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
3163 construct. Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
3164 clauses. Handle OMP_CLAUSE_REDUCTION with task modifier. Remove
3165 second argument create_tmp_var if it is NULL. Don't ignore shared
3166 clauses in is_host_teams_ctx contexts. Handle
3167 OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
3169 (lower_reduction_clauses): Ignore reduction clauses with task
3170 modifier. Remove second argument create_tmp_var if it is NULL.
3171 Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
3172 (lower_send_clauses): Ignore reduction clauses with task modifier.
3173 Handle OMP_CLAUSE__REDUCTEMP_. Don't send anything for
3174 OMP_CLAUSE_REDUCTION on taskloop. Handle OMP_CLAUSE_IN_REDUCTION.
3175 (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
3176 rely that it is the last stmt in body so far. Ignore outer taskgroup
3178 (omp_task_reductions_find_first, omp_task_reduction_iterate,
3179 lower_omp_task_reductions): New functions.
3180 (lower_omp_sections): Handle reduction clauses with taskgroup
3181 modifiers. Adjust maybe_add_implicit_barrier_cancel caller.
3182 (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
3183 (lower_omp_for): Likewise. Handle reduction clauses with taskgroup
3185 (lower_omp_taskgroup): Handle taskgroup reductions.
3186 (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
3187 Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
3188 (lower_depend_clauses): If there are any
3189 OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
3190 depend clauses, use a new array format. If OMP_CLAUSE_DEPEND_LAST is
3191 seen, assume lowering is done already and return early. Set kind
3192 on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
3193 (lower_omp_taskreg): Handle reduction clauses with task modifier on
3194 parallel construct. Handle reduction clause on taskloop construct.
3195 Handle taskwait with depend clauses.
3196 (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
3197 for host teams constructs.
3198 * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
3199 nontemporal and _reductemp_ clause entries.
3200 (omp_clause_code_name): Likewise.
3201 (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
3202 OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
3203 * tree-core.h (enum omp_clause_code): Add
3204 OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
3205 (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
3206 (struct tree_base): Add omp_atomic_memory_order field into union.
3207 Remove OMP_ATOMIC_SEQ_CST comment.
3208 (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
3209 and OMP_CLAUSE_DEPEND_DEPOBJ.
3210 (struct tree_omp_clause): Add subcode.defaultmap_kind.
3211 * tree.def (OMP_TASKGROUP): Add another operand, move next to other
3212 OpenMP constructs with body and clauses operands.
3213 * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
3214 (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
3215 (OMP_TASKGROUP_CLAUSES): Define.
3216 (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
3217 OMP_CLAUSE__LOOPTEMP_.
3218 (OMP_ATOMIC_SEQ_CST): Remove.
3219 (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
3220 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
3221 (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
3222 OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
3223 OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
3224 OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
3225 OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
3226 (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
3227 OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
3228 OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
3230 * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
3231 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
3232 OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
3233 (convert_local_omp_clauses): Likewise. Remove useless test.
3234 * tree-parloops.c (create_call_for_reduction_1): Pass
3235 OMP_MEMORY_ORDER_RELAXED as new argument to
3236 dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
3237 * tree-pretty-print.c (dump_omp_iterators): New function.
3238 (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
3239 OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION. Print
3240 reduction modifiers. Handle OMP_CLAUSE_DEPEND_DEPOBJ and
3241 OMP_CLAUSE_DEPEND_MUTEXINOUTSET. Print iterators in depend clauses.
3242 Print __internal__ for OMP_CLAUSE_DEPEND_LAST. Handle cancel and
3243 simd OMP_CLAUSE_IF_MODIFIERs. Handle new kinds of
3244 OMP_CLAUSE_DEFAULTMAP. Print conditional: for
3245 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
3246 (dump_omp_atomic_memory_order): New function.
3247 (dump_generic_node): Use it. Print taskgroup clauses. Print
3248 taskwait with depend clauses.
3249 * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
3250 * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
3251 Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
3252 * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
3253 write_ts_omp_clause_tree_pointers): Likewise.
3255 2018-11-08 David Malcolm <dmalcolm@redhat.com>
3258 * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
3259 "-missed", "-note", and "-all" sub-options.
3260 * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
3262 (can_early_inline_edge_p): Likewise.
3263 (want_early_inline_function_p): Likewise.
3264 (want_inline_self_recursive_call_p): Likewise.
3265 (recursive_inlining): Likewise.
3266 (inline_small_functions): Likewise.
3267 (flatten_function): Likewise.
3268 (ipa_inline): Likewise.
3269 (inline_always_inline_functions): Likewise.
3270 (early_inline_small_functions): Likewise.
3271 (early_inliner): Likewise.
3272 * tree-inline.c (expand_call_inline): Likewise.
3274 2018-11-08 David Malcolm <dmalcolm@redhat.com>
3276 * pretty-print.c (pp_format): Handle %f.
3277 (selftest::test_pp_format): Add test of %f.
3278 * pretty-print.h (pp_double): New macro.
3280 2018-11-08 David Malcolm <dmalcolm@redhat.com>
3282 * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
3283 * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
3284 (dump_pretty_printer::decode_format): Implement "%C" for
3286 (selftest::test_capture_of_dump_calls): Rename "where" to
3287 "stmt_loc". Convert test_decl to a function decl and set its
3288 location. Add a symbol_table_test RAII instance and a
3289 cgraph_node, using it to test "%C" and dump_symtab_node.
3291 2018-11-08 Eric Botcazou <ebotcazou@adacore.com>
3294 * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
3296 2018-11-08 David Malcolm <dmalcolm@redhat.com>
3298 * cgraph.c: Include "selftest.h".
3299 (saved_symtab): New variable.
3300 (selftest::symbol_table_test::symbol_table_test): New ctor.
3301 (selftest::symbol_table_test::~symbol_table_test): New dtor.
3302 (selftest::test_symbol_table_test): New test.
3303 (selftest::cgraph_c_tests): New.
3304 * cgraph.h (saved_symtab): New decl.
3305 (selftest::symbol_table_test): New class.
3306 * selftest-run-tests.c (selftest::run_tests): Call
3307 selftest::cgraph_c_tests.
3308 * selftest.h (selftest::cgraph_c_tests): New decl.
3310 2018-11-08 Richard Biener <rguenther@suse.de>
3312 * tree-data-ref.h (lambda_int): New typedef.
3313 (lambda_vector_gcd): Adjust.
3314 (lambda_vector_new): Likewise.
3315 (lambda_matrix_new): Likewise.
3316 * tree-data-ref.c (print_lambda_vector): Adjust.
3318 2018-11-08 Richard Biener <rguenther@suse.de>
3320 PR tree-optimization/87929
3321 * tree-complex.c (expand_complex_comparison): Clean EH.
3323 2018-11-08 Martin Liska <mliska@suse.cz>
3325 * doc/extend.texi: Reword.
3326 * predict.c (expr_expected_value_1): Likewise.
3328 2018-11-08 Richard Biener <rguenther@suse.de>
3330 PR tree-optimization/87913
3331 * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
3332 of extreme values to ordered comparisons.
3334 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
3337 * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
3339 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
3341 * doc/invoke.texi: Remove leading dash from @opindex entries
3342 throughout the file.
3344 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
3347 * doc/invoke.texi (Option Summary): Add -e and --entry.
3348 (Link Options): Likewise.
3350 2018-11-07 Nathan Sidwell <nathan@acm.org>
3353 * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
3354 --disable-checking bootstrap.
3356 2018-11-07 Hafiz Abid Qadeer <abidh@codesourcery.com>
3358 * configure: Regenerated.
3360 2018-11-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3363 * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
3364 to the mode of the widest field iff the widest field has mode class
3365 MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
3368 2018-11-07 Nikolai Merinov <n.merinov@inango-systems.com>
3370 * common.opt: Add -Wattribute-warning.
3371 * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
3372 * expr.c (expand_expr_real_1): Add new attribute to warning_at
3373 call to allow user configure behavior of "warning" attribute.
3375 2018-11-07 Segher Boessenkool <segher@kernel.crashing.org>
3377 * target.def: Put @: after every vs., e.g., and i.e. where it is
3378 followed by whitespace.
3379 * doc/extend.texi: Ditto.
3380 * doc/fragments.texi: Ditto.
3381 * doc/gimple.texi: Ditto.
3382 * doc/implement-c.texi: Ditto.
3383 * doc/install.texi: Ditto.
3384 * doc/invoke.texi: Ditto.
3385 * doc/md.texi: Ditto.
3386 * doc/plugins.texi: Ditto.
3387 * doc/rtl.texi: Ditto.
3388 * doc/sourcebuild.texi: Ditto.
3389 * doc/tm.texi.in: Ditto.
3390 * doc/ux.texi: Ditto.
3391 * doc/tm.texi: Regenerate.
3393 2018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3395 * config/arm/arm-cpus.in (ares): New entry.
3396 * config/arm/arm-tables.opt: Regenerate.
3397 * config/arm/arm-tune.md: Likewise.
3398 * doc/invoke.texi (ARM Options): Document ares.
3400 2018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3402 * config/aarch64/aarch64-cores.def (ares): Define.
3403 * config/aarch64/aarch64-tune.md: Regenerate.
3404 * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
3406 2018-11-07 Jan Hubicka <jh@suse.cz>
3408 * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
3409 (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
3410 functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
3413 2018-11-07 Richard Biener <rguenther@suse.de>
3415 PR tree-optimization/87914
3416 * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
3418 (vectorizable_reduction): Handle shifts and rotates by dispatching
3419 to vectorizable_shift.
3420 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
3421 in-loop uses of vect_nested_cycle defs. Merge cycle and internal
3423 (vectorizable_shift): Export and handle being called as
3425 (vect_analyze_stmt): Call vectorizable_shift after
3426 vectorizable_reduction.
3427 * tree-vectorizer.h (vectorizable_shift): Declare.
3429 2018-11-07 Jan Hubicka <jh@suse.cz>
3431 * ipa-devirt.c (odr_types_equivalent_p): Expect constants
3432 than const decls in TREE_VALUE of enum.
3433 (dump_type_inheritance_graph): Improve duplicate dumping.
3434 (free_enum_values): New.
3435 (build_type_inheritance_graph): Use it.
3436 * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
3437 which are not main variants or not ODR types.
3438 (verify_type_variant): Expect variants to have no TYPE_VALUES.
3440 2018-11-07 Richard Biener <rguenther@suse.de>
3442 * ipa-inline.c (want_inline_small_function_p): Compute
3443 big_speedup_p lazily and last.
3445 2018-11-07 Jan Hubicka <jh@suse.cz>
3447 * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
3448 building incomplete variant of complete type.
3449 (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
3450 variant of complete type.
3452 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
3454 * config/mips/mips.c: Fix typo in documentation of
3455 mips_loongson_ext2_prefetch_cookie.
3456 (mips_option_override): fix brain twister logical.
3457 * config/mips/mips.h: Fix typo in documentation of
3458 ISA_HAS_CTZ_CTO and define pattern.
3459 * config/mips/mips.md (prefetch): Hoist EXT2 above
3461 (prefetch_indexed): Hoist EXT2 above the EXT block.
3463 2018-11-07 Jan Hubicka <jh@suse.cz>
3465 * tree.c (free_lang_data_in_type): Add fld parameter; simplify
3466 return and parameter types of function and method types.
3467 (free_lang_data_in_cgraph): Update.
3469 2018-11-07 Martin Liska <mliska@suse.cz>
3471 PR rtl-optimization/87868
3472 * postreload-gcse.c (eliminate_partially_redundant_load): Set
3473 threshold to max_count if we would overflow.
3474 * profile-count.h: Make max_count a public constant.
3476 2018-11-07 Martin Liska <mliska@suse.cz>
3478 * mem-stats.h: Fix GNU coding style.
3480 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
3482 * config/mips/gs264e.md: New.
3483 * config/mips/mips-cpus.def: Define gs264e.
3484 * config/mips/mips-tables.opt: Regenerate.
3485 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
3487 (mips_issue_rate): Add support for gs264e.
3488 (mips_multipass_dfa_lookahead): Likewise.
3489 * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
3490 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
3491 (MIPS_ASE_MSA_SPEC): New.
3492 (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
3493 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
3494 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
3495 * config/mips/mips.md: Include gs264e.md.
3496 (processor): Add gs264e.
3497 * config/mips/mips.opt (MSA): Use Mask instead of Var.
3498 * doc/invoke.texi: Add gs264e to supported architectures.
3500 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
3502 * config/mips/gs464e.md: New.
3503 * config/mips/mips-cpus.def: Define gs464e.
3504 * config/mips/mips-tables.opt: Regenerate.
3505 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
3507 (mips_issue_rate): Add support for gs464e.
3508 (mips_multipass_dfa_lookahead): Likewise.
3509 (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
3510 * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
3511 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
3512 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
3513 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
3514 * config/mips/mips.md: Include gs464e.md.
3515 (processor): Add gs464e.
3516 * doc/invoke.texi: Add gs464e to supported architectures.
3518 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
3520 * config/mips/loongson3a.md: Rename to ...
3521 * config/mips/gs464.md: ... here.
3522 * config/mips/mips-cpus.def: Define gs464; Add loongson3a
3523 as an alias of gs464 processor.
3524 * config/mips/mips-tables.opt: Regenerate.
3525 * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
3526 instead of PROCESSOR_LOONGSON_3A.
3527 (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
3529 (mips_option_override): Enable MMI and EXT for gs464.
3530 * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
3531 Rename TUNE_LOONGSON_3A to TUNE_GS464.
3532 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
3533 (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
3534 ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
3536 * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
3537 (processor): Add gs464;
3538 * doc/invoke.texi: Add gs464 to supported architectures.
3540 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
3542 * config/mips/mips-protos.h
3543 (mips_loongson_ext2_prefetch_cookie): New prototype.
3544 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
3545 (mips_option_override): Enable TARGET_LOONGSON_EXT when
3546 TARGET_LOONGSON_EXT2 is true.
3547 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
3548 __mips_loongson_ext2, __mips_loongson_ext_rev=2.
3549 (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
3550 (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
3551 TARGET_LOONGSON_EXT2.
3552 (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
3553 (define_insn "ctz<mode>2"): New insn pattern.
3554 (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
3555 (define_insn "prefetch_indexed_<mode>"): Include
3556 TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
3557 * config/mips/mips.opt (-mloongson-ext2): Add option.
3558 * gcc/doc/invoke.texi (-mloongson-ext2): Document.
3560 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
3562 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
3563 __mips_loongson_ext.
3564 (MIPS_ASE_LOONGSON_EXT_SPEC): New.
3565 (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
3567 (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
3568 * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
3569 <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
3570 instead of TARGET_LOONGSON_3A.
3571 * config/mips/mips.opt (-mloongson-ext): Add option.
3572 * gcc/doc/invoke.texi (-mloongson-ext): Document.
3574 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
3576 * config.gcc (extra_headers): Add loongson-mmiintrin.h.
3577 * config/mips/loongson.md: Move to ...
3578 * config/mips/loongson-mmi.md: here; Adjustment.
3579 * config/mips/loongson.h: Move to ...
3580 State as deprecated. Include loongson-mmiintrin.h for back
3581 compatibility and warning.
3582 * config/mips/loongson-mmiintrin.h: ... here.
3583 * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
3584 mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
3585 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
3586 (mips_option_override): Make sure MMI use hard float;
3587 (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
3588 mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
3589 mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
3590 TARGET_LOONGSON_VECTORS.
3591 * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
3592 (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
3593 (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
3594 (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
3596 (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
3597 TARGET_LOONGSON_VECTORS.
3598 * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
3599 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
3600 (Loongson MMI patterns): Include loongson-mmi.md instead of
3602 * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
3603 * gcc/doc/invoke.texi (-mloongson-mmi): Document.
3605 2018-11-07 Richard Biener <rguenther@suse.de>
3608 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
3609 BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
3611 2018-11-07 Alexandre Oliva <aoliva@redhat.com>
3613 PR rtl-optimization/87874
3614 * lra.c (lra_substitute_pseudo): Do not create a subreg for
3617 2018-11-06 Aaron Sawdey <acsawdey@linux.ibm.com>
3619 * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
3620 if not in indexed or indirect form.
3621 (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
3622 (bswap<mode>2_store): Ditto.
3624 2018-11-06 Richard Earnshaw <rearnsha@arm.com>
3626 * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
3629 2018-11-06 Richard Biener <rguenther@suse.de>
3631 PR tree-optimization/86850
3632 * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
3633 instead of src.m_vec.
3635 2018-11-06 Jan Hubicka <jh@suse.cz>
3637 * tree.c (fld_simplified_type_name): Break out form ...
3638 (free_lang_data_in_type): ... here.
3639 (fld_type_variant_equal_p): Use it.
3641 2018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3643 * config/default-d.c: Include memmodel.h.
3645 * config/sol2-d.c: New file.
3646 * config/t-sol2 (sol2-d.o): New rule.
3647 * config.gcc <*-*-solaris2*>: Set d_target_objs,
3648 target_has_targetdm.
3650 2018-11-06 Jan Hubicka <jh@suse.cz>
3652 * tree.c (fld_type_variant): Also copy alignment; be sure that
3653 new variant is equal.
3655 2018-11-06 Ilya Leoshkevich <iii@linux.ibm.com>
3658 * config/s390/s390.md: Add relative_long attribute.
3660 2018-11-06 Jan Hubicka <jh@suse.cz>
3662 * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
3663 * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
3664 * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
3665 stream TYPE_NEEDS_CONSTRUCTING.
3666 * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
3667 * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
3669 2018-11-06 Richard Biener <rguenther@suse.de>
3671 * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
3673 (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
3675 2018-11-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3677 * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
3678 REG_CLASS_CONTENTS[GEN_REGS].
3679 (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
3681 2018-11-06 Jan Hubicka <jh@suse.cz>
3683 * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
3685 2018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3688 * config/sol2.h (ASAN_CC1_SPEC): Define.
3689 (LD_WHOLE_ARCHIVE_OPTION): Define.
3690 (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
3691 (ASAN_REJECT_SPEC): Provide default.
3692 (LIBASAN_EARLY_SPEC): Define.
3693 (LIBTSAN_EARLY_SPEC): Define.
3694 (LIBLSAN_EARLY_SPEC): Define.
3695 * config/i386/sol2.h (CC1_SPEC): Redefine.
3696 (ASAN_REJECT_SPEC): Define.
3698 * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
3699 (TARGET_ASAN_SHADOW_OFFSET): Define.
3700 (sparc_asan_shadow_offset): New function.
3701 * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
3702 (ASAN_REJECT_SPEC): Define.
3704 2018-11-06 Jan Hubicka <jh@suse.cz>
3706 * tree.c (fld_type_variant): Copy canonical type.
3707 (fld_incomplete_type_of): Check that canonical types looks sane;
3708 copy canonical type.
3709 (verify_type): Accept when incomplete type has complete canonical type.
3711 2018-11-06 Jan Hubicka <jh@suse.cz>
3713 * tree.c (free_lang_data): Reset overwite_assembler_name,
3714 print_xnode, print_decl, print_type and print_identifier of
3717 2018-11-06 Richard Biener <rguenther@suse.de>
3719 PR tree-optimization/87889
3720 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
3721 Do nothing if old and new arg are the same
3723 2018-11-06 Andreas Krebbel <krebbel@linux.ibm.com>
3726 * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
3727 attributes for operands 3 and 4.
3729 2018-11-06 Richard Biener <rguenther@suse.de>
3732 * simplify-rtx.c (simplify_binary_operation_1): Add pattern
3733 matching bitfield insertion.
3735 2018-11-06 Alexandre Oliva <aoliva@redhat.com>
3737 * auto-inc-dec.c: Include valtrack.h. Improve comments.
3738 (reg_next_debug_use): New.
3739 (attempt_change): Propagate adjusted expression into affected
3741 (merge_in_block): Track uses in debug insns.
3742 (pass_inc_dec::execute): Allocate and release
3745 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
3747 * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
3748 (_mm512_fixupimm_round_pd): Update parameters and builtin.
3749 (_mm512_maskz_fixupimm_round_pd): Ditto.
3750 (_mm512_fixupimm_round_ps): Ditto.
3751 (_mm512_maskz_fixupimm_round_ps): Ditto.
3752 (_mm_fixupimm_round_sd): Ditto.
3753 (_mm_maskz_fixupimm_round_sd): Ditto.
3754 (_mm_fixupimm_round_ss): Ditto.
3755 (_mm_maskz_fixupimm_round_ss): Ditto.
3756 (_mm512_fixupimm_pd): Ditto.
3757 (_mm512_maskz_fixupimm_pd): Ditto.
3758 (_mm512_fixupimm_ps): Ditto.
3759 (_mm512_maskz_fixupimm_ps): Ditto.
3760 (_mm_fixupimm_sd): Ditto.
3761 (_mm_maskz_fixupimm_sd): Ditto.
3762 (_mm_fixupimm_ss): Ditto.
3763 (_mm_maskz_fixupimm_ss): Ditto.
3764 (_mm512_mask_fixupimm_round_pd): Update builtin.
3765 (_mm512_mask_fixupimm_round_ps): Ditto.
3766 (_mm_mask_fixupimm_round_sd): Ditto.
3767 (_mm_mask_fixupimm_round_ss): Ditto.
3768 (_mm512_mask_fixupimm_pd): Ditto.
3769 (_mm512_mask_fixupimm_ps): Ditto.
3770 (_mm_mask_fixupimm_sd): Ditto.
3771 (_mm_mask_fixupimm_ss): Ditto.
3772 * config/i386/avx512vlintrin.h:
3773 (_mm256_fixupimm_pd): Update parameters and builtin.
3774 (_mm256_maskz_fixupimm_pd): Ditto.
3775 (_mm256_fixupimm_ps): Ditto.
3776 (_mm256_maskz_fixupimm_ps): Ditto.
3777 (_mm_fixupimm_pd): Ditto.
3778 (_mm_maskz_fixupimm_pd): Ditto.
3779 (_mm_fixupimm_ps): Ditto.
3780 (_mm_maskz_fixupimm_ps): Ditto.
3781 (_mm256_mask_fixupimm_pd): Update builtin.
3782 (_mm256_mask_fixupimm_ps): Ditto.
3783 (_mm_mask_fixupimm_pd): Ditto.
3784 (_mm_mask_fixupimm_ps): Ditto.
3785 * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
3786 * config/i386/i386-builtin.def: Update builtin definitions.
3787 * config/i386/i386.c: Handle new builtin types and remove useless ones.
3788 * config/i386/sse.md: Update VFIXUPIMM* patterns.
3789 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
3790 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
3791 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
3792 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
3793 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
3794 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
3795 * config/i386/subst.md:
3796 (round_saeonly_sd_mask_operand4): Add new subst_attr.
3797 (round_saeonly_sd_mask_op4): Ditto.
3798 (round_saeonly_expand_operand5): Ditto.
3799 (round_saeonly_expand): Update.
3801 2018-11-05 Max Filippov <jcmvbkbc@gmail.com>
3803 * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
3805 2018-11-05 Segher Boessenkool <segher@kernel.crashing.org>
3807 PR rtl-optimization/87871
3808 * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
3810 2018-11-05 Paul Koning <ni1d@arrl.net>
3812 * doc/sourcebuild.texi (target attributes): Document new "inf"
3813 effective target keyword.
3815 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
3817 * config/s390/s390.c (s390_register_move_cost): Increase costs for
3818 moves involving the CC reg.
3820 2018-11-05 Richard Biener <rguenther@suse.de>
3822 PR tree-optimization/87873
3823 * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
3825 * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
3826 * tree-vect-loop.c (vect_transform_loop): When splitting the
3827 loop exit also create forwarder PHIs for constants.
3828 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
3829 Handle constant to_arg, add extra checking we match up the correct
3832 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
3834 * config/s390/s390.md: QImode and HImode for load on condition.
3836 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
3838 * config/s390/predicates.md: Fix typo.
3839 * config/s390/s390.md: Allow immediates for load on condition.
3841 2018-11-05 Martin Liska <mliska@suse.cz>
3843 * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
3844 * bitmap.h (struct bitmap_usage): Likewise.
3845 * ggc-common.c (SCALE): Remove.
3847 (struct ggc_usage): Use SIZE_AMOUNT. And update
3849 * ggc-page.c (SCALE): Remove.
3850 (STAT_LABEL): Remove.
3851 (ggc_print_statistics): Use SIZE_AMOUNT.
3852 * gimple.h (SCALE): Remove.
3854 * input.c (ONE_K): Remove.
3857 (STAT_LABEL): Likewise.
3858 (FORMAT_AMOUNT): Likewise.
3859 (dump_line_table_statistics): Use SIZE_AMOUNT.
3860 * mem-stats.h (struct mem_usage): Likewise.
3861 * rtl.c (dump_rtx_statistics): Likewise.
3862 (rtx_alloc_counts): Change type to size_t.
3863 (rtx_alloc_sizes): Likewise.
3864 (rtx_count_cmp): New.
3865 (dump_rtx_statistics): Sort first based on counts.
3866 * tree.c (tree_nodes_cmp): New.
3867 (tree_codes_cmp): New.
3868 (dump_tree_statistics): Sort first based on counts.
3869 * system.h (ONE_K): New.
3871 (SIZE_SCALE): Likewise.
3872 (SIZE_LABEL): Likewise.
3873 (SIZE_AMOUNT): Likewise.
3874 * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
3875 * tree-dfa.c (dump_dfa_stats): Likewise.
3876 * tree-phinodes.c (phinodes_print_statistics): Likewise.
3877 * tree-ssanames.c (ssanames_print_statistics): Likewise.
3878 * tree.c (dump_tree_statistics): Likewise.
3879 * vec.c (struct vec_usage): Likewise.
3880 * trans-mem.c (tm_mangle): Enlarge buffer in order to not
3881 trigger a -Werror=format-overflow with
3882 --enable-gather-detailed-stats.
3884 2018-11-05 Martin Liska <mliska@suse.cz>
3886 * mem-stats.h (mem_alloc_description::release_instance_overhead):
3888 * vec.c (struct vec_usage): Register m_element_size.
3889 (vec_prefix::register_overhead): New arguments: elements and
3891 (vec_prefix::release_overhead): Subtract elements.
3892 * vec.h (struct vec_prefix): Change signature.
3893 (va_heap::reserve): Pass proper arguments.
3894 (va_heap::release): Likewise.
3896 2018-11-05 Martin Liska <mliska@suse.cz>
3898 * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
3902 2018-11-05 Richard Biener <rguenther@suse.de>
3904 * tree-scalar-evolution.h (final_value_replacement_loop): Update
3906 * tree-scalar-evolution.c (final_value_replacement_loop): Return
3907 whether anything was done.
3908 (scev_const_prop): Remove constant propagation part, fold
3910 * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
3911 (pass_data_scev_cprop): TODO_cleanup_cfg is now done
3914 2018-11-05 Jakub Jelinek <jakub@redhat.com>
3916 PR tree-optimization/87859
3917 * gimple-ssa-store-merging.c (struct merged_store_group): Add
3918 only_constants and first_nonmergeable_order members.
3919 (merged_store_group::merged_store_group): Initialize them.
3920 (merged_store_group::do_merge): Clear only_constants member if
3921 adding something other than INTEGER_CST store.
3922 (imm_store_chain_info::coalesce_immediate_stores): Don't merge
3923 stores with order >= first_nonmergeable_order. Use
3924 merged_store->only_constants instead of always recomputing it.
3925 Set merged_store->first_nonmergeable_order if we've skipped any
3926 stores. Attempt to merge overlapping INTEGER_CST stores that
3927 we would otherwise skip.
3930 * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
3932 2018-11-05 Xuepeng Guo <xuepeng.guo@intel.com>
3935 * config/i386/emmintrin.h (__v16qs): New to cope with option
3937 (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
3938 (_mm_cmplt_epi8): Likewise.
3939 (_mm_cmpgt_epi8): Likewise.
3941 2018-11-05 Richard Biener <rguenther@suse.de>
3943 PR rtl-optimization/87852
3944 * fwprop.c (use_killed_between): Only consider single-defs of the
3945 use whose definition statement dominates the use.
3947 2018-11-05 Martin Liska <mliska@suse.cz>
3950 * doc/invoke.texi: Remove options that are
3951 not disabled with -Os.
3953 2018-11-05 Martin Liska <mliska@suse.cz>
3956 * doc/extend.texi: Update constrain about the last argument
3957 of __builtin_expect_with_probability.
3959 2018-11-05 Martin Liska <mliska@suse.cz>
3962 * predict.c (expr_expected_value_1): Verify
3963 that last argument is a real constants and emit
3966 2018-11-05 Martin Liska <mliska@suse.cz>
3968 PR gcov-profile/77698
3969 * ipa-profile.c (ipa_profile): Adjust hotness threshold
3972 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
3974 PR tree-optimization/86572
3975 * builtins.c (c_strlen): Handle negative offsets in a safe way.
3977 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
3979 PR tree-optimization/87672
3980 * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
3981 * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
3983 2018-11-04 Uros Bizjak <ubizjak@gmail.com>
3986 * cfgexpand.c (pass_expand::execute): Move the call to
3987 finish_eh_generation in front of the call to expand_stack_alignment.
3989 2018-11-04 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
3991 * common/config/i386/i386-common.c (processor_alias_table): Add
3993 * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
3994 (case ${target}): Add znver2.
3995 * config/i386/driver-i386.c: (host_detect_local_cpu): Let
3996 -march=native recognize znver2 processors.
3997 * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
3998 * config/i386/i386.c (m_znver2): New definition.
3999 (m_ZNVER): New definition.
4000 (m_AMD_MULTIPLE): Includes m_znver2.
4001 (processor_cost_table): Add znver2 entry.
4002 (processor_target_table): Add znver2 entry.
4003 (get_builtin_code_for_version): Set priority for
4005 (processor_model): Add M_AMDFAM17H_ZNVER2.
4006 (arch_names_table): Ditto.
4007 (ix86_reassociation_width): Include znver2.
4008 * config/i386/i386.h (TARGET_znver2): New definition.
4009 (struct ix86_size_cost): Add TARGET_ZNVER2.
4010 (enum processor_type): Add PROCESSOR_ZNVER2.
4011 * config/i386/i386.md (define_attr "cpu"): Add znver2.
4012 * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
4013 * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
4014 (ix86_adjust_cost): Add znver2.
4015 * config/i386/x86-tune.def: Replace m_ZNVER1 by m_ZNVER.
4016 * gcc/doc/extend.texi: Add details about znver2.
4017 * gcc/doc/invoke.texi: Add details about znver2.
4019 2018-11-03 Sandra Loosemore <sandra@codesourcery.com>
4023 * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
4026 2018-11-02 Aaron Sawdey <acsawdey@linux.ibm.com>
4028 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
4029 attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
4031 2018-11-02 Richard Earnshaw <rearnsha@arm.com>
4033 * config/aarch64/aarch64.c ((aarch64_override_options): Disable
4034 shrink-wrapping when -mtrack-speculation.
4036 2018-11-02 Richard Biener <rguenther@suse.de>
4038 * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
4039 (pop_cost_one_pair): Do not free pair.
4040 (pop_best_coalesce): Likewise.
4041 (create_coalesce_list): Initialize obstack.
4042 (delete_coalesce_list): Free obstack.
4043 (find_coalesce_pair): Obstack-allocate coalesce pairs.
4044 (add_cost_one_coalesce): Likewise.
4045 (struct live_track): Remove bitmap pointer indirections.
4046 (new_live_track): Adjust.
4047 (delete_live_track): Likewise.
4048 (live_track_remove_partition): Likewise.
4049 (live_track_add_partition): Likewise.
4050 (live_track_live_p): Likewise.
4051 (live_track_process_def): Likewise.
4052 (live_track_clear_base_vars): Likewise.
4054 2018-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4056 * configure.ac (gcc_cv_as_sparc_register_op): Remove.
4057 * configure: Regenerate.
4058 * config.in: Regenerate.
4059 * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
4060 HAVE_AS_REGISTER_PSEUDO_OP guard.
4061 * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
4062 !HAVE_AS_REGISTER_PSEUDO_OP support.
4063 (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
4066 2018-11-02 Richard Biener <rguenther@suse.de>
4068 * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
4071 2018-11-02 Richard Biener <rguenther@suse.de>
4073 PR tree-optimization/87776
4074 * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
4075 executable when iterating but running into rpo-vn-max-loop-depth
4076 and not eliding the iteration.
4078 2018-11-30 Jan Hubicka <jh@suse.cz>
4080 * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
4083 2018-11-01 Aaron Sawdey <acsawdey@linux.ibm.com>
4085 * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
4087 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
4088 Combine with rs6000_address_for_fpconvert.
4089 (rs6000_address_for_fpconvert): Combine with
4090 rs6000_force_indexed_or_indirect_mem.
4091 (rs6000_expand_vector_init): Change function call from
4092 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
4093 * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
4094 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
4095 (floatsi<mode>2_lfiwax_mem): Ditto.
4096 (floatunssi<mode>2_lfiwzx): Ditto.
4097 (floatunssi<mode>2_lfiwzx_mem): Ditto.
4098 (float<QHI:mode><FP_ISA3:mode>2): Ditto.
4099 (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
4100 (fix_trunc<mode>si2_stfiwx): Ditto.
4101 (fixuns_trunc<mode>si2_stfiwx): Ditto.
4102 (float_<mode>si2_hw): Ditto.
4103 (floatuns_<mode>si2_hw): Ditto.
4104 * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
4105 (vsx_splat_<mode>): Ditto.
4107 2018-11-01 Joseph Myers <joseph@codesourcery.com>
4109 * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
4110 at end of assembler input text.
4111 * configure: Regenerate.
4113 2018-11-01 Jakub Jelinek <jakub@redhat.com>
4115 PR tree-optimization/87826
4116 * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
4117 negative or larger or equal to type's precision.
4119 2018-10-31 Alexandre Oliva <aoliva@redhat.com>
4121 * opts.c (default_options_table): Do not enable
4122 OPT_fdelayed_branch at -Og.
4123 * doc/invoke.texi (-fdelayed-branch): Document it.
4125 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
4127 * optabs-libfuncs.c (build_libfunc_function_visibility):
4128 New, split out from...
4129 (build_libfunc_function): ... here.
4130 (init_one_libfunc_visibility): New, split out from ...
4131 (init_one_libfunc): ... here.
4133 * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
4134 scratch register need not be early-clobber. Document the reason
4135 why we cannot use ST<OP>.
4137 2018-10-31 Joseph Myers <joseph@codesourcery.com>
4140 * configure.ac: Remove AC_PREREQ. Use AC_LANG_SOURCE. Use single
4141 line for second argument of AC_DEFINE_UNQUOTED.
4142 * doc/install.texi (Tools/packages necessary for modifying GCC):
4143 Update to autoconf 2.69 and automake 1.15.1.
4144 * aclocal.m4, config.in, configure: Regenerate.
4146 2018-10-31 Pat Haugen <pthaugen@us.ibm.com>
4148 * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
4151 2018-10-31 Martin Liska <mliska@suse.cz>
4154 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
4155 Add new argument invalid_extension.
4156 (aarch64_get_all_extension_candidates): New function.
4157 (aarch64_rewrite_selected_cpu): Add NULL to function call.
4158 * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
4160 (aarch64_get_all_extension_candidates): New function.
4161 * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
4162 argument invalid_extension.
4163 (aarch64_parse_cpu): Likewise.
4164 (aarch64_print_hint_for_extensions): New function.
4165 (aarch64_validate_mcpu): Provide hint about invalid extension.
4166 (aarch64_validate_march): Likewise.
4167 (aarch64_handle_attr_arch): Pass new argument.
4168 (aarch64_handle_attr_cpu): Provide hint about invalid extension.
4169 (aarch64_handle_attr_isa_flags): Likewise.
4171 2018-10-31 Richard Biener <rguenther@suse.de>
4175 * tree-outof-ssa.c (insert_backedge_copies): Restrict
4176 copy generation to useful cases. Place the copy before
4177 the definition of the backedge value when possible.
4179 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
4181 * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
4182 * config/arc/arc.c (arc_active_insn): New function.
4183 (check_store_cacheline_hazard): Likewise.
4184 (workaround_arc_anomaly): Use check_store_cacheline_hazard.
4185 (arc_override_options): Disable delay slot scheduler for older
4187 (arc_store_addr_hazard_p): New implementation, old one renamed to
4189 (arc_store_addr_hazard_internal_p): Renamed.
4190 (arc_reorg): Don't combine into brcc instructions which are part
4191 of hardware hazard solution.
4192 * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
4193 (tune_arc700): Likewise.
4194 * config/arc/arc.opt (arc7xx): New tune value.
4195 * config/arc/arc700.md: Improve A7 scheduler.
4197 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
4199 * config/arc/arc.c (arc_override_options): Remove
4200 TARGET_COMPACT_CASESI.
4201 * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
4202 (CASE_VECTOR_MODE): Likewise.
4203 (CASE_VECTOR_PC_RELATIVE): Likewise.
4204 (CASE_VECTOR_SHORTEN_MODE): Likewise.
4205 (CASE_VECTOR_SHORTEN_MODE1): Delete.
4206 (ADDR_VEC_ALIGN): Update.
4207 (ASM_OUTPUT_CASE_LABEL): Undefine.
4208 (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
4209 (TARGET_BI_BIH): Define.
4210 (DEFAULT_BRANCH_INDEX): Likewise.
4211 * config/arc/arc.md (casesi): Rework to accept BI/BIH
4212 instructions, remove compact_casesi use case.
4213 (casesi_compact_jump): Remove.
4214 (casesi_dispatch): New pattern.
4215 * config/arc/arc.opt: Add mbranch-index option. Deprecate
4216 compact_casesi option.
4217 * doc/invoke.texi: Document mbranch-index option.
4219 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
4221 * config/arc/arc.c (arc_get_tp): Remove function.
4222 (arc_emit_call_tls_get_addr): Likewise.
4223 (arc_call_tls_get_addr): New function.
4224 (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
4225 * config/arc/arc.md (tls_load_tp_soft): Remove.
4226 (tls_gd_get_addr): Likewise.
4228 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
4230 * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
4231 (mulsi3_600_lib): Remove pattern.
4232 (umulsi3_highpart_600_lib_le): Likewise.
4233 (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
4234 (umulsidi3): Remove call to umulsidi3_600_lib.
4235 (umulsidi3_600_lib): Remove pattern.
4236 (peephole2): Remove peephole using the above deprecated patterns.
4238 2018-10-31 Thomas Preud'homme <thomas.preudhomme@linaro.org>
4241 * config/arm/arm.c (arm_option_check_internal): Disable the combined
4242 use of -mslow-flash-data and -mword-relocations.
4243 (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
4244 * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
4246 * doc/invoke.texi (-mword-relocations): Mention conflict with
4248 (-mslow-flash-data): Reciprocally.
4250 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
4252 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
4253 16-byte modes held in GP registers to use an even regno.
4255 * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
4256 (aarch64_atomic_ldop_supported_p): Remove.
4257 (aarch64_gen_atomic_ldop): Remove.
4258 * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
4259 Fully expand LSE operations here.
4260 (atomic_fetch_<atomic_optab><ALLI>): Likewise.
4261 (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
4262 (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
4263 and use ATOMIC_LDOP instead; use register_operand for the input;
4264 drop the split and emit insns directly.
4265 (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
4266 (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
4267 (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
4269 * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
4270 (aarch64_gen_atomic_ldop): Don't call it.
4271 * config/aarch64/atomics.md (atomic_exchange<ALLI>):
4272 Use aarch64_reg_or_zero.
4273 (aarch64_atomic_exchange<ALLI>): Likewise.
4274 (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
4275 operand 0; use aarch64_reg_or_zero for input; merge ...
4276 (@aarch64_atomic_swp<ALLI>): ... this and remove.
4278 * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
4279 (aarch64_split_compare_and_swap): Use it.
4280 (aarch64_expand_compare_and_swap): Likewise. Remove convert_modes;
4281 test oldval against the proper predicate.
4282 * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
4283 Use nonmemory_operand for expected.
4284 (cas_short_expected_pred): New.
4285 (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
4286 (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
4287 * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
4288 (aarch64_plushi_operand): New.
4290 * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
4291 Force oldval into the rval register for TARGET_LSE; emit the compare
4292 during initial expansion so that it may be deleted if unused.
4293 (aarch64_gen_atomic_cas): Remove.
4294 * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
4295 Change =&r to +r for operand 0; use match_dup for operand 2;
4296 remove is_weak and mod_f operands as unused. Drop the split
4298 (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
4299 (@aarch64_compare_and_swap<GPI>_lse): Similarly.
4300 (@aarch64_atomic_cas<GPI>): Similarly.
4302 2018-10-31 Richard Biener <rguenther@suse.de>
4304 * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
4307 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
4309 * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
4310 * config/aarch64/aarch64.md: Include saphira.md
4311 * config/aarch64/saphira.md: New file for pipeline description.
4313 2018-10-30 Martin Sebor <msebor@redhat.com>
4316 * gimple-ssa-sprintf.c (format_directive): Use %G to include
4318 (sprintf_dom_walker::compute_format_length):
4319 Avoid setting POSUNDER4K here.
4320 (get_destination_size): Handle null argument values.
4321 (get_user_idx_format): New function.
4322 (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
4323 functions, including user-defined with attribute format printf.
4324 Use %G to include inlining context.
4325 Set POSUNDER4K here.
4327 2018-10-30 Jan Hubicka <jh@suse.cz>
4329 * params.def (lto-partitions): Bump from 32 to 128.
4331 2018-10-30 Jan Hubicka <jh@suse.cz>
4334 (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
4336 (free_lang_data_in_type): Forward declare.
4337 (fld_type_variant_equal_p): New function.
4338 (fld_type_variant): New function
4339 (fld_incomplete_types): New hash.
4340 (fld_incomplete_type_of): New function
4341 (fld_simplfied-type): New function.
4342 (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
4343 (free_lang_data): Allocate and free fld_incomplete_type; update call
4344 of free_lang_data_in_decl.
4346 2018-10-30 Eric Botcazou <ebotcazou@adacore.com>
4348 * gcov.c (output_lines): Remove duplicate line.
4350 2018-10-30 Aaron Sawdey <acsawdey@linux.ibm.com>
4352 * config/rs6000/rs6000.md (bswapdi2): Force address into register
4353 if not in indexed or indirect form.
4354 (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
4355 (bswapdi2_store): Ditto.
4356 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
4358 * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
4359 Prototype for helper function.
4361 2018-10-30 Martin Sebor <msebor@redhat.com>
4363 * doc/extend.texi (optimize): Clarify/expand attribute documentation.
4364 (target, pragma GCC optimize, pragma GCC target): Ditto.
4366 2018-10-30 Jonathan Wakely <jwakely@redhat.com>
4368 * doc/extend.texi: Fix prototype and description of
4369 __builtin_expect_with_probability.
4371 2018-10-30 Michael Ploujnikov <michael.ploujnikov@oracle.com>
4373 * cgraph.h (clone_function_name_1): Replaced by new
4374 clone_function_name_numbered that takes name as string; for
4375 privatize_symbol_name_1 use only.
4376 (clone_function_name): Renamed to
4377 clone_function_name_numbered to be explicit about numbering.
4378 (clone_function_name): New two-argument function that does
4379 not number its output.
4380 (clone_function_name): New three-argument function that
4381 takes a number to append to its output.
4382 * cgraphclones.c (duplicate_thunk_for_node):
4383 (clone_function_name_1): Renamed.
4384 (clone_function_name_numbered): Two new functions.
4385 (clone_function_name): Improved documentation.
4386 (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
4387 * config/rs6000/rs6000.c (make_resolver_func): Ditto.
4388 * final.c (final_scan_insn_1): Use the new clone_function_name
4390 * multiple_target.c (create_dispatcher_calls): Ditto.
4391 (create_target_clone): Ditto.
4392 * omp-expand.c (grid_expand_target_grid_body): Ditto.
4393 * omp-low.c (create_omp_child_function_name): Ditto.
4394 * omp-simd-clone.c (simd_clone_create): Ditto.
4395 * symtab.c (simd_symtab_node::noninterposable_alias): Use the
4396 new clone_function_name without numbering.
4398 2018-10-30 Richard Earnshaw <rearnsha@arm.com>
4400 * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
4401 Assert that the allocation size is not zero.
4403 2018-10-30 Richard Biener <rguenther@suse.de>
4405 PR tree-optimization/87800
4406 * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
4407 non-induction or reduction PHIs.
4409 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
4411 * config/aarch64/falkor-tag-collision-avoidance.c
4412 (execute_tag_collision_avoidance): Call df_note_add_problem.
4414 2018-10-30 Martin Liska <mliska@suse.cz>
4416 * doc/extend.texi: Fix typo in documentation
4417 of __builtin_expect_with_probability.
4419 2018-10-29 David Malcolm <dmalcolm@redhat.com>
4422 * input.c (get_substring_ranges_for_loc): Detect if
4423 linemap_resolve_location gives us a NULL map, and reject
4426 2018-10-29 Iain Buclaw <ibuclaw@gdcproject.org>
4428 * config.gcc (xstormy16-*-elf): Set tm_d_file.
4430 2018-10-29 David Malcolm <dmalcolm@redhat.com>
4431 Martin Sebor <msebor@redhat.com>
4432 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
4434 * doc/ux.texi (Quoting): New subsection, adapted from material at
4435 https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
4436 MartinSebor and ManuelLopezIbanez.
4437 (Fix-it hints): Note that fix-it hints shouldn't be marked for
4440 2018-10-29 Kugan Vivekanandarajah <kuganv@linaro.org>
4443 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
4446 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
4448 * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
4450 2018-10-29 Paul Koning <ni1d@arrl.net>
4452 * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
4454 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
4456 * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
4457 _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
4458 _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
4459 _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
4460 _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
4461 _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
4462 _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
4463 _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
4464 Change 'vector' to '__vector'.
4465 * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
4466 _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
4467 _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
4468 _mm_avg_pu16): Likewise. And, whitespace corrections.
4470 2018-10-29 Richard Biener <rguenther@suse.de>
4472 PR tree-optimization/87785
4473 * tree-vect-slp.c (vect_gather_slp_loads): Only gather
4476 2018-10-29 Olivier Hainque <hainque@adacore.com>
4478 * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
4480 2018-10-29 Olivier Hainque <hainque@adacore.com>
4482 * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
4483 ports configuration macro, defaults to "gnu".
4484 (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
4485 TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
4487 2018-10-29 Olivier Hainque <hainque@adacore.com>
4489 * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
4491 2018-10-29 Richard Biener <rguenther@suse.de>
4493 PR tree-optimization/87790
4494 * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
4495 (vect_make_slp_decision): Adjust.
4496 (vect_slp_analyze_bb_1): Likewise.
4497 (vect_detect_hybrid_slp_stmts): Properly union SLP type over
4500 2018-10-29 Richard Biener <rguenther@suse.de>
4502 PR tree-optimization/87785
4503 * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
4505 (vect_build_slp_tree): Likewise.
4506 (vect_gather_slp_loads): New function.
4507 (vect_analyze_slp_instance): Gather loads separately from the
4510 2018-10-29 Martin Liska <mliska@suse.cz>
4512 * Makefile.in: Make dependency to json.o.
4513 * doc/gcov.texi: Document new JSON format, remove
4514 old intermediate format documentation.
4515 * gcov.c (struct function_info): Come up with m_name and
4517 (function_info::function_info): Initialize it.
4518 (function_info::~function_info): Release it.
4519 (main): Rename flag_intermediate_format to flag_json_format.
4520 (print_usage): Describe --json-format.
4521 (process_args): Set flag_json_format.
4522 (output_intermediate_line): Remove.
4523 (output_intermediate_json_line): Likewise.
4524 (get_gcov_intermediate_filename): Return new extension
4526 (output_intermediate_file): Implement JSON emission.
4527 (output_json_intermediate_file): Implement JSON emission.
4528 (generate_results): Use ::get_name for function name.
4529 Handle JSON output file.
4530 (read_graph_file): Use ::get_name instead of cplus_demangle.
4531 (read_count_file): Likewise.
4532 (solve_flow_graph): Likewise.
4533 (add_line_counts): Likewise.
4534 (accumulate_line_counts): Use new flag_json_format.
4535 (output_function_details): Use ::get_name instead of cplus_demangle.
4536 (output_lines): Likewise.
4537 * json.cc (test_writing_literals): Add new tests.
4538 * json.h (class literal): Add new boolean constructor.
4540 2018-10-29 Segher Boessenkool <segher@kernel.crashing.org>
4542 PR rtl-optimization/87701
4543 PR rtl-optimization/87780
4544 * combine.c (make_more_copies): Rewrite.
4546 2018-10-28 Kugan Vivekanandarajah <kuganv@linaro.org>
4548 * doc/generic.texi (ABSU_EXPR): Document.
4549 * match.pd (absu(x)*absu(x) -> x*x): Handle.
4550 (absu(absu(X)) -> absu(X)): Likewise.
4551 (absu(-X) -> absu(X)): Likewise.
4552 (absu(X) where X is nonnegative -> X): Likewise.
4554 2018-10-28 Iain Buclaw <ibuclaw@gdcproject.org>
4556 * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
4557 (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
4558 (tm_d.h, cs-tm_d.h, default-d.o): New rules.
4559 (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
4560 (s-tm-texi): Also check timestamp on d-target.def.
4561 (generated_files): Add TM_D_H and d-target-hooks-def.h.
4562 (build/genhooks.o): Also depend on D_TARGET_DEF.
4563 * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
4565 * config/aarch64/aarch64-d.c: New file.
4566 * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
4568 * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
4570 * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
4571 * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
4572 * config/arm/arm-d.c: New file.
4573 * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
4574 * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
4575 * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
4576 * config/arm/t-arm (arm-d.o): New rule.
4577 * config/default-d.c: New file.
4578 * config/glibc-d.c: New file.
4579 * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4580 * config/i386/i386-d.c: New file.
4581 * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
4582 * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
4583 * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
4584 (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
4585 * config/i386/t-i386 (i386-d.o): New rule.
4586 * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4587 * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4588 * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
4589 * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4590 * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
4591 * config/mips/mips-d.c: New file.
4592 * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
4593 * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
4594 * config/mips/t-mips (mips-d.o): New rule.
4595 * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4596 * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4597 * config/powerpcspe/powerpcspe-d.c: New file.
4598 * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
4600 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
4601 Support GNU D by using 0 as the language type.
4602 * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
4603 * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
4604 * config/riscv/riscv-d.c: New file.
4605 * config/riscv/riscv-protos.h (riscv_d_target_versions): New
4607 * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
4608 * config/riscv/t-riscv (riscv-d.o): New rule.
4609 * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4610 * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
4611 * config/rs6000/rs6000-d.c: New file.
4612 * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
4614 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
4615 Support GNU D by using 0 as the language type.
4616 * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
4617 * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
4618 * config/s390/s390-d.c: New file.
4619 * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
4620 * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
4621 * config/s390/t-s390 (s390-d.o): New rule.
4622 * config/sparc/sparc-d.c: New file.
4623 * config/sparc/sparc-protos.h (sparc_d_target_versions): New
4625 * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
4626 * config/sparc/t-sparc (sparc-d.o): New rule.
4627 * config/t-glibc (glibc-d.o): New rule.
4628 * configure: Regenerated.
4629 * configure.ac (tm_d_file): New variable.
4630 (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
4631 * doc/contrib.texi (Contributors): Add self for the D frontend.
4632 * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
4633 * doc/install.texi (Configuration): Mention libphobos as an option for
4634 --enable-shared. Mention d as an option for --enable-languages.
4635 (Testing): Mention check-d as a target.
4636 * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
4637 name suffixes. Mention d as a -x option.
4638 * doc/sourcebuild.texi (Top Level): Mention libphobos.
4639 * doc/standards.texi (Standards): Add section on D language.
4640 * doc/tm.texi: Regenerated.
4641 * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
4642 TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
4643 * dwarf2out.c (is_dlang): New function.
4644 (gen_compile_unit_die): Use DW_LANG_D for D.
4645 (declare_in_namespace): Return module die for D, instead of adding
4646 extra declarations into the namespace.
4647 (gen_namespace_die): Generate DW_TAG_module for D.
4648 (gen_decl_die): Handle CONST_DECLSs for D.
4649 (dwarf2out_decl): Likewise.
4650 (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
4651 (prune_unused_types_walk): Handle DW_tag_interface_type same as other
4652 kinds of aggregates.
4653 * gcc.c (default_compilers): Add entries for .d, .dd and .di.
4654 * genhooks.c: Include d/d-target.def.
4656 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
4659 * config/rs6000/darwin.h (STACK_BOUNDARY): New.
4660 (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
4661 (STACK_DYNAMIC_OFFSET): Likewise.
4663 2018-10-27 Sandra Loosemore <sandra@codesourcery.com>
4666 * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
4669 2018-10-26 Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
4671 * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
4672 return type and other typos.
4674 2018-10-26 Aaron Sawdey <acsawdey@linux.ibm.com>
4676 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
4677 a shorter sequence with fewer branches.
4678 (emit_final_str_compare_gpr): Ditto.
4680 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
4682 * config/rs6000/tmmintrin.h: New file.
4683 * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
4685 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
4687 * config/rs6000/mmintrin.h: Enable 32bit compilation.
4688 * config/rs6000/xmmintrin.h: Likewise.
4690 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
4692 * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
4694 2018-10-26 Richard Biener <rguenther@suse.de>
4696 * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
4698 (vect_mark_slp_stmts_relevant): Likewise.
4699 (vect_detect_hybrid_slp_stmts): Likewise.
4700 (vect_bb_slp_scalar_cost): Likewise.
4701 (vect_remove_slp_scalar_calls): Likewise.
4703 2018-10-26 Jan Hubicka <jh@suse.cz>
4705 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
4706 (warn_types_mismatch): Fix walk of DECL_NAME.
4707 (odr_types_equivalent_p): Fix overactive assert.
4709 2018-10-26 Richard Biener <rguenther@suse.de>
4711 PR tree-optimization/87105
4712 * tree-vectorizer.h (_slp_tree::refcnt): New member.
4713 * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
4715 (vect_create_new_slp_node): Initialize refcnt to one.
4717 (scalar_stmts_set_t, bst_fail): Remove.
4718 (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
4719 (vect_build_slp_tree): Add bst_map argument and lookup
4720 already created SLP nodes.
4721 (vect_print_slp_tree): Handle a SLP graph, print SLP node
4723 (vect_slp_rearrange_stmts): Handle a SLP graph.
4724 (vect_analyze_slp_instance): Adjust and free SLP nodes from
4725 the CSE map. Fix indenting.
4726 (vect_schedule_slp_instance): Add short-cut.
4728 2018-10-26 Martin Liska <mliska@suse.cz>
4731 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
4732 addr_expr and not with pointers.
4734 2018-10-26 Jan Hubicka <jh@suse.cz>
4736 * tree.c (free_lang_data_in_type): Only check main variants.
4737 * ipa-devirt.c (warn_odr): Make static.
4738 (types_same_for_odr): Drop strict variant.
4739 (types_odr_comparable): Likewise.
4740 (odr_or_derived_type_p): Look for main variants.
4741 (odr_name_hasher::equal): Cleanup comment.
4742 (odr_subtypes_equivalent): Add warn and warned arguments; check main
4744 (type_variants_equivalent_p): break out from ...
4745 (odr_types_equivalent): ... here; go for main variants where needed.
4746 (warn_odr): ... here; turn static.
4747 (warn_types_mismatch): Compare mangled names of main variants.
4748 * ipa-utils.h (types_odr_comparable): Drop strict parameter.
4749 (type_with_linkage_p): Sanity check that we look at main variant.
4750 * lto.c (lto_read_decls): Only consider main variant to be ODR type.
4751 * tree.h (types_same_for_odr): Drop strict argument.
4753 2018-10-26 Richard Biener <rguenther@suse.de>
4755 PR tree-optimization/87746
4756 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
4757 Simplify and fix WRT strided store groups with size not
4758 equal to step in element count.
4759 (vect_analyze_group_access_1): Dump the whole group.
4761 2018-10-25 Carl Love <cel@us.ibm.com>
4763 * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
4764 P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
4765 P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name. Add quad
4766 precicion entry for each overloaded builtin.
4767 * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
4768 VSCEDPUO): Rename overloaded name.
4769 (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
4770 VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
4771 * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
4772 define_expand for xscmpexqp instruction.
4773 (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
4775 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
4776 Jinsong Ji <jji@us.ibm.com>
4778 * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
4779 function with vec_sl.
4780 (_mm_slli_epi32): Likewise.
4781 (_mm_slli_epi64): Likewise.
4782 (_mm_srai_epi16): Replace deprecated function with vec_sra.
4783 (_mm_srai_epi32): Likewise.
4784 (_mm_srli_epi16): Replace deprecated function with vec_sr.
4785 (_mm_srli_epi32): Likewise.
4786 (_mm_srli_epi64): Likewise.
4787 (_mm_sll_epi16): Replace deprecated function with vec_sl.
4788 (_mm_sll_epi32): Likewise.
4789 (_mm_sll_epi64): Likewise.
4790 (_mm_sra_epi16): Replace deprecated function with vec_sra.
4791 (_mm_sra_epi32): Likewise.
4792 (_mm_srl_epi16): Replace deprecated function with vec_sr.
4793 (_mm_srl_epi32): Likewise.
4794 (_mm_srl_epi64): Likewise.
4796 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
4797 Jinsong Ji <jji@us.ibm.com>
4799 * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
4800 comparison operators with vec_cmp* for compatibility due to
4801 unfortunate history; clean up formatting and use types more
4803 (_mm_sll_epi32): Likewise.
4804 (_mm_sll_epi64): Likewise.
4805 (_mm_srl_epi16): Likewise.
4806 (_mm_srl_epi32): Likewise.
4807 (_mm_srl_epi64): Likewise.
4809 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
4810 Jinsong Ji <jji@us.ibm.com>
4812 * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
4813 * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
4814 __vector __bool int. Use vec_cmpgt in preference to deprecated
4815 function vec_vcmpgtfp.
4816 (_mm_max_ps): Likewise.
4818 2018-10-25 Jeff Law <law@redhat.com>
4820 * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
4823 2018-10-25 Martin Sebor <msebor@redhat.com>
4825 * doc/extend.texi (aligned): Expand attribute description.
4826 (Alignment): Rename section. Discuss function arguments.
4828 2018-10-25 Jan Hubicka <jh@suse.cz>
4830 * ipa-devirt.c (main_odr_variant): Remove.
4831 (hash_odr_name, types_same_for_odr, types_odr_comparable,
4832 odr_name_hasher::equal, odr_subtypes_equivalent_p):
4833 Drop use of main_odr_variant.
4834 (add_type_duplicate): Silence confused warnings on integer types.
4835 (get_odr_type): Always look for main variant.
4836 (register_odr_type): Simplify.
4838 2018-10-25 Richard Biener <rguenther@suse.de>
4840 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
4841 Initialize ng to silence error with release checking bootstrap.
4843 2018-10-25 Richard Biener <rguenther@suse.de>
4845 * tree-if-conv.c: Include tree-ssa-sccvn.h.
4846 (tree_if_conversion): Run CSE on the if-converted loop body.
4848 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
4850 * config/s390/constraints.md (ZL): New constraint.
4851 * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
4853 * config/s390/s390.md (movdi_larl): Remove.
4854 (movdi_64): Add the LARL alternative.
4856 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
4859 * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
4860 (RTX_CODE_HWINT_P): New macro.
4861 (rtx_code_size): Use RTX_CODE_HWINT_P ().
4863 2018-10-25 Jan Hubicka <jh@suse.cz>
4865 * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
4868 2018-10-25 Richard Biener <rguenther@suse.de>
4870 PR tree-optimization/87665
4871 PR tree-optimization/87745
4872 * tree-vectorizer.h (get_earlier_stmt): Remove.
4873 (get_later_stmt): Pick up UID from the original non-pattern stmt.
4875 2018-10-25 Sam Tebbs <sam.tebbs@arm.com>
4877 * options.texi (Deprecated): Move list to Var section.
4879 2018-10-24 Bill Schmidt <wschmidt@linux.ibm.com>
4880 Jinsong Ji <jji@us.ibm.com>
4882 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
4883 __vector long to __vector long long.
4884 (_mm_cvtpd_ps): Likewise.
4885 (_mm_cvttpd_epi32): Likewise.
4886 (_mm_cvtpi32_pd): Likewise.
4887 (_mm_unpackhi_epi64): Likewise.
4888 (_mm_unpacklo_epi64): Likewise.
4890 2018-10-24 Segher Boessenkool <segher@kernel.crashing.org>
4892 PR rtl-optimization/87720
4893 * combine.c (make_more_copies): Skip if the dest is pc_rtx.
4895 2018-10-24 Alexandre Oliva <aoliva@redhat.com>
4897 * gimple-ssa-isolate-paths.c
4898 (find_implicit_erroneous_behavior): Do not change code if the
4899 pass is running for warnings only.
4900 (find_explicit_erroneous_behavior): Likewise.
4902 2018-10-24 Michael Meissner <meissner@linux.ibm.com>
4904 * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
4905 Define as rs6000_mangle_decl_assembler_name.
4906 (rs6000_mangle_decl_assembler_name): If the user switched from IBM
4907 long double to IEEE long double, switch the names of the long
4908 double built-in functions to be <func>f128 instead of <func>l.
4910 2018-10-24 Martin Sebor <msebor@redhat.com>
4912 * doc/extend.texi (nonnull): List no-argument form. Reference
4913 -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
4915 2018-10-24 Richard Biener <rguenther@suse.de>
4917 * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
4919 2018-10-24 Martin Liska <mliska@suse.cz>
4921 PR tree-optimization/84436
4922 * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
4924 (switch_conversion::contains_linear_function_p): New.
4925 (switch_conversion::build_one_array): Support linear
4926 transformation on input.
4927 * tree-switch-conversion.h (struct switch_conversion): Add
4928 contains_linear_function_p declaration.
4930 2018-10-24 Richard Biener <rguenther@suse.de>
4932 * varasm.c (const_hash_1): Return hash of ADDR_EXPR
4933 if its argument is CONSTANT_CLASS_P.
4935 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
4937 * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
4938 it is wrong for forward declarations.
4940 2018-10-24 Ilya Leoshkevich <iii@linux.ibm.com>
4942 * config/s390/s390.c (s390_check_qrst_address): Add the missing
4943 SYMBOL_REF_P () check.
4945 2018-10-24 Richard Biener <rguenther@suse.de>
4947 PR tree-optimization/87105
4948 * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
4949 dump classification.
4950 (vect_analyze_data_ref_accesses): Handle duplicate loads and
4951 stores by splitting the affected group after the fact.
4952 * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
4953 fail the SLP build because of size constraints.
4955 2018-10-24 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4957 * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
4958 * configure: Regenerate.
4959 * config.in: Regenerate.
4960 * varasm.c (mergeable_string_section): Use readonly_data_section
4961 if linker doesn't support SHF_MERGE with alignment > 8.
4962 (mergeable_constant_section): Likewise.
4964 2018-10-24 Richard Biener <rguenther@suse.de>
4966 PR tree-optimization/84013
4967 * tree-ssa-structalias.c (struct msdi_data): New struct for
4968 marshalling data to walk_stmt_load_store_ops.
4969 (maybe_set_dependence_info): Refactor as callback for
4970 walk_stmt_load_store_ops.
4971 (compute_dependence_clique): Set restrict info on all stmt kinds.
4973 2018-10-24 Martin Liska <mliska@suse.cz>
4975 * cgraph.c (cgraph_node::dump):
4976 Remove reduntant dumps and make tp_first_run dump more compact.
4978 2018-10-24 Richard Biener <rguenther@suse.de>
4980 PR tree-optimization/87665
4981 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
4984 2018-10-12 Jeff Law <law@redhat.com>
4986 * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
4989 2018-10-23 Richard Biener <rguenther@suse.de>
4991 * tree-vrp.c (add_assert_info): Guard dump_printf with
4993 * gimple-ssa-evrp-analyze.c
4994 (evrp_range_analyzer::record_ranges_from_incoming_edge):
4995 Use value_range::ignore_equivs_equal_p.
4997 2018-10-23 Richard Biener <rguenther@suse.de>
4999 PR tree-optimization/87105
5000 PR tree-optimization/87608
5001 * passes.def (pass_all_early_optimizations): Add early phi-opt
5003 * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
5004 addition to debug stmts.
5005 (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
5006 and abs replacement early.
5007 * tree-cfg.c (gimple_empty_block_p): Likewise.
5009 2018-10-23 Richard Earnshaw <rearnsha@arm.com>
5012 * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
5013 specified to configure.
5014 (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
5016 2018-10-23 Richard Biener <rguenther@suse.de>
5018 PR tree-optimization/87700
5019 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
5021 2018-10-23 Jakub Jelinek <jakub@redhat.com>
5024 * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
5025 second argument from __mmask16 to __mmask8.
5026 * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
5027 _mm_mask_packs_epi32): Likewise.
5028 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
5030 (_mm512_mask_prefetch_i64scatter_pd): Likewise. Formatting fix.
5032 2018-10-23 Richard Biener <rguenther@suse.de>
5034 * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
5036 2018-10-23 Richard Biener <rguenther@suse.de>
5038 PR tree-optimization/86144
5039 * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
5040 over simd attribute.
5042 2018-10-23 Richard Biener <rguenther@suse.de>
5044 PR tree-optimization/87693
5045 * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
5046 the case we do not find the taken edge.
5048 2018-10-22 Bill Schmidt <wschmidt@linux.ibm.com>
5049 Jinsong Ji <jji@us.ibm.com>
5051 * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
5052 (_mm_store_pd): Use unaligned vector type for pointer cast.
5053 (_mm_maskmoveu_si128): Likewise.
5054 * config/rs6000/xmmintrin.h (__m128_u): New typedef.
5055 (_mm_store_ps): Use unaligned vector type for pointer cast.
5057 2018-10-22 Paul Koning <ni1d@arrl.net>
5059 * symtab.c (symtab_node::increase_alignment): Correct max
5062 2018-10-22 Yury Gribov <tetra2005@gmail.com>
5064 PR tree-optimization/87633
5065 * match.pd: Do not generate unordered integer comparisons.
5067 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
5069 PR rtl-optimization/87600
5070 * combine.c: Add include of expr.h.
5071 (cant_combine_insn_p): Do not combine moves from any hard non-fixed
5072 register to a pseudo.
5073 (make_more_copies): New function, add a copy to a new pseudo after
5074 the moves from hard registers into pseudos.
5075 (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
5076 later. Call make_more_copies.
5078 2018-10-22 Andrew Stubbs <ams@codesourcery.com>
5080 * lra-constraints.c (process_alt_operands): New local array,
5081 matching_early_clobber. Check matching_early_clobber before
5082 decrementing reject, and set matching_early_clobber after.
5084 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
5087 * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
5088 call output_addr_const and hope for the best.
5090 2018-10-22 Richard Biener <rguenther@suse.de>
5092 * gimple-ssa-evrp-analyze.c
5093 (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
5094 smarter about what ranges to use.
5095 * tree-vrp.c (add_assert_info): Dump here.
5096 (register_edge_assert_for_2): Instead of here at multiple but
5099 * gcc.dg/tree-ssa/evrp12.c: New testcase.
5100 * gcc.dg/predict-6.c: Adjust.
5101 * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
5102 * gcc.dg/tree-ssa/vrp02.c: Likewise.
5103 * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
5105 2018-10-22 Steven Bosscher <steven@gcc.gnu.org>
5106 Richard Biener <rguenther@suse.de>
5108 * bitmap.h: Update data structure documentation, including a
5109 description of bitmap views as either linked-lists or splay trees.
5110 (struct bitmap_element_def): Update comments for splay tree bitmaps.
5111 (struct bitmap_head_def): Likewise.
5112 (bitmap_list_view, bitmap_tree_view): New prototypes.
5113 (bitmap_initialize_stat): Initialize a bitmap_head's indx and
5115 (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
5116 (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
5117 * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
5118 released bitmap element here.
5119 (bitmap_element_free): Remove.
5120 (bitmap_elt_clear_from): Work on splay tree bitmaps.
5121 (bitmap_list_link_element): Renamed from bitmap_element_link. Move
5122 this function similar ones such that linked-list bitmap implementation
5123 functions are grouped.
5124 (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
5125 and moved for grouping.
5126 (bitmap_list_insert_element_after): Renamed from
5127 bitmap_elt_insert_after, and moved for grouping.
5128 (bitmap_list_find_element): New function spliced from bitmap_find_bit.
5129 (bitmap_tree_link_left, bitmap_tree_link_right,
5130 bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
5131 bitmap_tree_link_element, bitmap_tree_unlink_element,
5132 bitmap_tree_find_element): New functions for splay-tree bitmap
5134 (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
5135 Renamed and moved, see above entries.
5136 (bitmap_tree_listify_from): New function to convert part of a splay
5137 tree bitmap to a linked-list bitmap.
5138 (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
5139 (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
5140 (bitmap_find_bit): Remove.
5141 (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
5142 bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
5143 Handle splay tree bitmaps.
5144 (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
5145 bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
5146 bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
5147 bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
5148 bitmap_intersect_compl_p, bitmap_ior_and_compl,
5149 bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
5150 bitmap_hash): Reject trying to act on splay tree bitmaps. Make
5151 corresponding changes to use linked-list specific bitmap_element
5152 manipulation functions as applicable for efficiency.
5153 (bitmap_tree_to_vec): New function.
5154 (debug_bitmap_elt_file): New function split out from ...
5155 (debug_bitmap_file): ... here. Handle splay tree bitmaps.
5156 (bitmap_print): Likewise.
5158 PR tree-optimization/63155
5159 * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
5161 * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
5164 2018-10-22 Martin Liska <mliska@suse.cz>
5166 PR tree-optimization/87686
5168 2018-08-29 Martin Liska <mliska@suse.cz>
5170 * tree-switch-conversion.c (switch_conversion::expand):
5171 Strenghten assumption about gswitch statements.
5173 2018-10-22 Martin Liska <mliska@suse.cz>
5175 * ipa-icf.c (sem_item::compare_attributes): Remove.
5176 (sem_item::compare_referenced_symbol_properties): Use
5177 attribute_list_equal instead.
5178 (sem_function::equals_wpa): Likewise.
5179 * ipa-icf.h: Remove compare_attributes.
5181 2018-10-22 Richard Biener <rguenther@suse.de>
5184 * mem-stats.h (mem_usage::operator==): Fix pasto.
5186 2018-10-22 Richard Biener <rguenther@suse.de>
5188 PR tree-optimization/87640
5189 * tree-vrp.c (set_value_range_with_overflow): Decompose
5191 (extract_range_from_binary_expr_1): Adjust.
5193 2018-10-22 Martin Jambor <mjambor@suse.cz>
5195 * tree-eh.h (stmt_could_throw_p): Add function parameter.
5196 (stmt_can_throw_external): Likewise.
5197 (stmt_can_throw_internal): Likewise.
5198 * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
5199 (lower_eh_constructs_2): Likewise.
5200 (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
5201 (stmt_can_throw_external): Likewise.
5202 (stmt_can_throw_internal): Likewise.
5203 (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
5204 (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
5205 (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
5206 (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
5207 (pass_lower_eh_dispatch::execute): Pass cfun to
5208 stmt_can_throw_external.
5209 (cleanup_empty_eh): Likewise.
5210 (verify_eh_edges): Pass cfun to stmt_could_throw_p.
5211 * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
5212 stmt_can_throw_external instead of pushing it to cfun.
5213 (symbol_table::create_edge): Likewise.
5214 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
5215 stmt_can_throw_internal.
5216 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
5217 to stmt_could_throw_p.
5218 * gimple-ssa-store-merging.c (handled_load): Pass cfun to
5219 stmt_can_throw_internal.
5220 (pass_store_merging::execute): Likewise.
5221 * gimple-ssa-strength-reduction.c
5222 (find_candidates_dom_walker::before_dom_children): Pass cfun to
5224 * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
5225 stmt_can_throw_internal.
5226 * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
5227 to stmt_can_throw_external.
5228 (check_stmt): Pass cfun to stmt_could_throw_p.
5229 (check_stmt): Pass cfun to stmt_can_throw_external.
5230 (pass_nothrow::execute): Likewise.
5231 * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
5232 * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
5233 stmt_can_throw_internal.
5234 (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
5235 (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
5236 (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
5237 * tree-complex.c (expand_complex_libcall): Pass cfun to
5238 stmt_could_throw_p and to stmt_can_throw_internal.
5239 (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
5240 * tree-inline.c (copy_edges_for_bb): Likewise.
5241 (maybe_move_debug_stmts_to_successors): Likewise.
5242 * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
5244 * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
5245 * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
5246 * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
5247 stmt_can_throw_internal.
5248 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
5249 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
5251 (mark_aliased_reaching_defs_necessary_1): Pass cfun to
5252 stmt_can_throw_internal.
5253 * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
5254 * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
5256 * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
5257 (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
5258 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
5259 (convert_mult_to_fma_1): Likewise.
5260 (convert_to_divmod): Likewise.
5261 * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
5262 * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
5263 * tree-ssa-propagate.c
5264 (substitute_and_fold_dom_walker::before_dom_children): Likewise.
5265 * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
5266 (maybe_optimize_range_tests): Likewise.
5267 (linearize_expr_tree): Likewise.
5268 (reassociate_bb): Likewise.
5269 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
5270 * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
5271 * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
5272 (handle_char_store): Likewise.
5273 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
5274 stmt_can_throw_internal.
5275 * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
5277 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
5278 (vectorizable_call): Pass cfun to stmt_can_throw_internal.
5279 (vectorizable_simd_clone_call): Likewise.
5280 * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
5281 (gimple_stringop_fixed_value): Likewise.
5283 2018-10-22 Ilya Leoshkevich <iii@linux.ibm.com>
5285 * config/s390/s390.c (s390_loadrelative_operand_p): Accept
5286 literal pool references.
5287 (s390_check_qrst_address): Adapt to the new behavior of
5288 s390_loadrelative_operand_p ().
5290 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
5293 * config/i386/sse.md (*andnot<mode>3_bcst): New.
5295 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
5298 * config/i386/sse.md (*<code><mode>3_bcst): New.
5300 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
5303 * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
5304 V4DI, V16SI and V8DI.
5305 (*sub<mode>3<mask_name>_bcst): New.
5306 (*add<mode>3<mask_name>_bcst): Likewise.
5308 2018-10-21 Bill Schmidt <wschmidt@linux.ibm.com>
5309 Jinsong Ji <jji@us.ibm.com>
5311 * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
5312 __m64 with __vector unsigned long long for compatibility.
5313 (_mm_movemask_epi8): Likewise.
5314 * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
5315 (_mm_cvttps_pi32): Likewise.
5316 (_mm_cvtpi32_ps): Likewise.
5317 (_mm_cvtps_pi16): Likewise.
5318 (_mm_loadh_pi): Likewise.
5319 (_mm_storeh_pi): Likewise.
5320 (_mm_movehl_ps): Likewise.
5321 (_mm_movelh_ps): Likewise.
5322 (_mm_loadl_pi): Likewise.
5323 (_mm_storel_pi): Likewise.
5324 (_mm_movemask_ps): Likewise.
5325 (_mm_shuffle_pi16): Likewise.
5327 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
5330 * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
5331 __builtin_ia32_vfnmsubpd512_mask.
5332 (_mm512_mask_fnmsub_round_pd): Likewise.
5333 (_mm512_fnmsub_pd): Likewise.
5334 (_mm512_mask_fnmsub_pd): Likewise.
5335 (_mm512_maskz_fnmsub_round_pd): Use
5336 __builtin_ia32_vfnmsubpd512_maskz.
5337 (_mm512_maskz_fnmsub_pd): Likewise.
5338 (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
5339 (_mm512_mask_fnmsub_round_ps): Likewise.
5340 (_mm512_fnmsub_ps): Likewise.
5341 (_mm512_mask_fnmsub_ps): Likewise.
5342 (_mm512_maskz_fnmsub_round_ps): Use
5343 __builtin_ia32_vfnmsubps512_maskz.
5344 (_mm512_maskz_fnmsub_ps): Likewise.
5345 * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
5346 __builtin_ia32_vfnmsubpd256_mask.
5347 (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
5348 (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
5349 (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
5350 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
5351 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
5352 (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
5353 (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
5354 (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
5355 * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
5356 __builtin_ia32_vfnmsubpd.
5357 (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
5358 (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
5359 (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
5360 (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
5361 (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
5362 * config/i386/i386-builtin.def: Add
5363 __builtin_ia32_vfnmsubpd256_mask,
5364 __builtin_ia32_vfnmsubpd256_maskz,
5365 __builtin_ia32_vfnmsubpd128_mask,
5366 __builtin_ia32_vfnmsubpd128_maskz,
5367 __builtin_ia32_vfnmsubps256_mask,
5368 __builtin_ia32_vfnmsubps256_maskz,
5369 __builtin_ia32_vfnmsubps128_mask,
5370 __builtin_ia32_vfnmsubps128_maskz,
5371 __builtin_ia32_vfnmsubpd512_mask,
5372 __builtin_ia32_vfnmsubpd512_maskz,
5373 __builtin_ia32_vfnmsubps512_mask,
5374 __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
5375 __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
5376 __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
5377 __builtin_ia32_vfnmsubpd256.
5378 * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
5379 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
5380 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
5382 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
5384 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
5386 (fmai_vmfnmsub_<mode><round_name>): Likewise.
5388 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
5391 * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
5392 __builtin_ia32_vfnmaddpd512_mask.
5393 (_mm512_mask_fnmadd_round_pd): Likewise.
5394 (_mm512_fnmadd_pd): Likewise.
5395 (_mm512_mask_fnmadd_pd): Likewise.
5396 (_mm512_maskz_fnmadd_round_pd): Use
5397 __builtin_ia32_vfnmaddpd512_maskz.
5398 (_mm512_maskz_fnmadd_pd): Likewise.
5399 (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
5400 (_mm512_mask_fnmadd_round_ps): Likewise.
5401 (_mm512_fnmadd_ps): Likewise.
5402 (_mm512_mask_fnmadd_ps): Likewise.
5403 (_mm512_maskz_fnmadd_round_ps): Use
5404 __builtin_ia32_vfnmaddps512_maskz.
5405 (_mm512_maskz_fnmadd_ps): Likewise.
5406 * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
5407 __builtin_ia32_vfnmaddpd256_mask.
5408 (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
5409 (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
5410 (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
5411 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
5412 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
5413 (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
5414 (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
5415 (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
5416 * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
5417 __builtin_ia32_vfnmaddpd.
5418 (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
5419 (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
5420 (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
5421 (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
5422 (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
5423 * config/i386/i386-builtin.def: Add
5424 __builtin_ia32_vfnmaddpd256_mask,
5425 __builtin_ia32_vfnmaddpd256_maskz,
5426 __builtin_ia32_vfnmaddpd128_mask,
5427 __builtin_ia32_vfnmaddpd128_maskz,
5428 __builtin_ia32_vfnmaddps256_mask,
5429 __builtin_ia32_vfnmaddps256_maskz,
5430 __builtin_ia32_vfnmaddps128_mask,
5431 __builtin_ia32_vfnmaddps128_maskz,
5432 __builtin_ia32_vfnmaddpd512_mask,
5433 __builtin_ia32_vfnmaddpd512_maskz,
5434 __builtin_ia32_vfnmaddps512_mask,
5435 __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
5436 __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
5437 __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
5438 __builtin_ia32_vfnmaddpd256.
5439 * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
5440 (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
5441 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
5443 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
5445 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
5447 (fmai_vmfnmadd_<mode><round_name>): Likewise.
5449 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
5452 * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
5453 __builtin_ia32_vfmsubpd512_mask.
5454 (_mm512_mask_fmsub_round_pd): Likewise.
5455 (_mm512_fmsub_pd): Likewise.
5456 (_mm512_mask_fmsub_pd): Likewise.
5457 (_mm512_maskz_fmsub_round_pd): Use
5458 __builtin_ia32_vfmsubpd512_maskz.
5459 (_mm512_maskz_fmsub_pd): Likewise.
5460 (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
5461 (_mm512_mask_fmsub_round_ps): Likewise.
5462 (_mm512_fmsub_ps): Likewise.
5463 (_mm512_mask_fmsub_ps): Likewise.
5464 (_mm512_maskz_fmsub_round_ps): Use
5465 __builtin_ia32_vfmsubps512_maskz.
5466 (_mm512_maskz_fmsub_ps): Likewise.
5467 * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
5468 __builtin_ia32_vfmsubpd256_mask.
5469 (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
5470 (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
5471 (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
5472 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
5473 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
5474 (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
5475 (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
5476 (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
5477 * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
5478 __builtin_ia32_vfmsubpd.
5479 (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
5480 (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
5481 (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
5482 (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
5483 (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
5484 * config/i386/i386-builtin.def: Add
5485 __builtin_ia32_vfmsubpd256_mask,
5486 __builtin_ia32_vfmsubpd256_maskz,
5487 __builtin_ia32_vfmsubpd128_mask,
5488 __builtin_ia32_vfmsubpd128_maskz,
5489 __builtin_ia32_vfmsubps256_mask,
5490 __builtin_ia32_vfmsubps256_maskz,
5491 __builtin_ia32_vfmsubps128_mask,
5492 __builtin_ia32_vfmsubps128_maskz,
5493 __builtin_ia32_vfmsubpd512_mask,
5494 __builtin_ia32_vfmsubpd512_maskz,
5495 __builtin_ia32_vfmsubps512_mask,
5496 __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
5497 __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
5498 __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
5499 __builtin_ia32_vfmsubpd256.
5500 * config/i386/sse.md (fma4i_fmsub_<mode>): New.
5501 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
5502 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
5504 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
5506 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
5508 (fmai_vmfmsub_<mode><round_name>): Likewise.
5510 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
5512 * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
5513 Remove plus. Renamed to ...
5514 (*sub<mode>3<mask_name>_bcst): This.
5515 (*add<mode>3<mask_name>_bcst_2): Renamede to ...
5516 (*add<mode>3<mask_name>_bcst): This.
5518 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
5521 * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
5523 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
5526 * i386/avx512vlintrin.h (_mm256_or_epi32): New.
5527 (_mm_or_epi32): Likewise.
5528 (_mm256_xor_epi32): Likewise.
5529 (_mm_xor_epi32): Likewise.
5530 (_mm256_or_epi64): Likewise.
5531 (_mm_or_epi64): Likewise.
5532 (_mm256_xor_epi64): Likewise.
5533 (_mm_xor_epi64): Likewise.
5535 2018-10-20 H.J. Lu <hongjiu.lu@intel.com>
5538 * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
5540 2018-10-20 Jakub Jelinek <jakub@redhat.com>
5543 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
5545 2018-10-20 Andreas Schwab <schwab@linux-m68k.org>
5547 * doc/ux.texi: Move @section directly after @node.
5549 2018-10-19 Jakub Jelinek <jakub@redhat.com>
5553 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
5554 depend closely nested inside of loop with ordered clause with
5557 2018-10-19 David Malcolm <dmalcolm@redhat.com>
5559 * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
5560 * doc/gccint.texi: Include ux.texi and use it in top-level menu.
5561 * doc/ux.texi: New file.
5563 2018-10-19 Segher Boessenkool <segher@kernel.crashing.org>
5565 * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
5566 be the first CR field allocated.
5568 2018-10-19 Richard Biener <rguenther@suse.de>
5571 * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
5572 TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
5574 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
5577 * config/i386/sse.md
5578 (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
5579 (*add<mode>3<mask_name>_bcst_2): Likewise.
5581 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
5583 * config/i386/sse.md
5584 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
5585 Replace nonimmediate_operand with register_operand.
5586 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
5588 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
5591 2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
5593 PR rtl-optimization/87596
5594 * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
5595 lra_get_insn_recog_data () instead of lra_insn_recog_data[]
5596 for instructions in FROM..TO range.
5598 2018-10-19 Eric Botcazou <ebotcazou@adacore.com>
5600 * cfgexpand.c (expand_one_var): Use specific wording in error message
5601 for non-local frame variables.
5602 * stor-layout.c (layout_decl): Do not issue a warning for them.
5604 2018-10-19 Robin Dapp <rdapp@linux.ibm.com>
5606 * haifa-sched.c (priority): Add force_recompute parameter.
5607 (apply_replacement): Call priority () with force_recompute = true.
5608 (restore_pattern): Likewise.
5610 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
5612 * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
5613 HOST_BITS_PER_WIDE_INT.
5614 (test_vector_ops_duplicate): Likewise.
5616 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
5619 * config/i386/sse.md (VF_AVX512): New.
5620 (avx512bcst): Likewise.
5621 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
5623 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
5625 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
5628 2018-10-18 Jonathan Wakely <jwakely@redhat.com>
5630 * doc/invoke.texi (-dumpversion): Improve grammar.
5631 (-dumpfullversion): Make more consistent with -dumpversion.
5633 2018-10-18 Uros Bizjak <ubizjak@gmail.com>
5635 * config/i386/i386.c (ix86_emit_fp_unordered_jump):
5636 Set JUMP_LABEL to the jump insn.
5637 (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
5638 Predict emitted jump and add label to jump insn.
5640 2018-10-18 David Malcolm <dmalcolm@redhat.com>
5642 PR tree-optimization/87562
5643 * input.c (get_substring_ranges_for_loc): Use
5644 LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
5645 getting the linemap for the endpoint. Verify that it's either
5646 in the same linemap as the start point's spelling location, or
5647 at least in the same file.
5649 2018-10-18 Richard Biener <rguenther@suse.de>
5651 * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
5652 feed width-specific load/store costs through ix86_vec_cost.
5653 * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
5654 (k8_cost): Likewise.
5655 (bdver_cost): Likewise.
5656 (znver1_cost): Likewise.
5657 (btver1_cost): Likewise.
5658 (btver2_cost): Likewise.
5660 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
5662 * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
5663 to simplify subreg of vec_merge.
5665 2018-10-18 Richard Biener <rguenther@suse.de>
5667 * config/i386/i386.c: Fix costing of vector FMA.
5669 2018-10-18 Richard Biener <rguenther@suse.de>
5671 * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
5673 (ix86_builtin_vectorization_cost): For vec_construct properly
5674 cost insertion into SSE regs.
5675 (...): Adjust calls to ix86_vec_cost.
5677 2018-10-18 Richard Biener <rguenther@suse.de>
5681 2018-02-07 Richard Biener <rguenther@suse.de>
5683 PR tree-optimization/84204
5684 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5687 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
5690 * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
5692 (test_vector_ops_duplicate): Add test for a scalar subreg of a
5693 VEC_MERGE of a VEC_DUPLICATE.
5695 2018-10-17 Joseph Myers <joseph@codesourcery.com>
5697 * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
5698 * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
5699 * doc/standards.texi (C Language): Document C2X.
5700 * dwarf2out.c (highest_c_language), config/rl78/rl78.c
5701 (rl78_option_override): Handle "GNU C2X" language name.
5703 2018-10-17 Joseph Myers <joseph@codesourcery.com>
5705 * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
5706 Document C17 as published in 2018.
5708 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
5711 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
5712 bail out if both sides do not have the same storage order.
5714 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
5716 * bitmap.c (bitmap_head::dump): New.
5717 * bitmap.h (bitmap_head): Add dump().
5718 * gimple-ssa-evrp-analyze.c
5719 (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
5720 (evrp_range_analyzer::set_ssa_range_info): Same.
5721 (evrp_range_analyzer::record_ranges_from_phis): Same.
5722 (evrp_range_analyzer::record_ranges_from_stmt): Same.
5723 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
5724 * gimple-ssa-sprintf.c (get_int_range): Same.
5725 (format_integer): Same.
5726 (sprintf_dom_walker::handle_gimple_call): Same.
5727 * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
5728 (ipcp_vr_lattice::top_p): Same.
5729 (ipcp_vr_lattice::bottom_p): Same.
5730 (ipcp_vr_lattice::set_to_bottom): Same.
5731 (ipa_vr_operation_and_type_effects): Same.
5732 (propagate_vr_across_jump_function): Same.
5733 (ipcp_store_vr_results): Same.
5734 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
5735 (ipa_print_node_jump_functions_for_edge): Same.
5736 (ipa_get_value_range): Same.
5737 (ipa_compute_jump_functions_for_edge): Same.
5738 (ipa_write_jump_function): Same.
5739 * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
5740 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
5742 * vr-values.c (set_value_range_to_nonnegative): Same.
5743 (set_value_range_to_truthvalue): Same.
5744 (vr_values::get_value_range): Same.
5745 (vr_values::set_defs_to_varying): Same.
5746 (vr_values::update_value_range): Same.
5747 (symbolic_range_based_on_p): Same.
5748 (vr_values::op_with_boolean_value_range_p): Same.
5749 (vr_values::extract_range_for_var_from_comparison_expr): Same.
5750 (vr_values::extract_range_from_ssa_name): Same.
5751 (vr_values::extract_range_from_binary_expr): Same.
5752 (vr_values::extract_range_from_unary_expr): Same.
5753 (vr_values::extract_range_from_cond_expr): Same.
5754 (vr_values::extract_range_from_comparison): Same.
5755 (vr_values::check_for_binary_op_overflow): Same.
5756 (vr_values::extract_range_basic): Same.
5757 (vr_values::extract_range_from_assignment): Same.
5758 (compare_ranges): Same.
5759 (compare_range_with_value): Same.
5760 (vr_values::adjust_range_with_scev): Same.
5761 (vrp_valueize): Same.
5762 (vrp_valueize_1): Same.
5763 (vr_values::get_vr_for_comparison): Same.
5764 (vr_values::compare_name_with_value): Same.
5765 (vr_values::compare_names): Same.
5766 (vr_values::vrp_evaluate_conditional): Same.
5767 (find_case_label_ranges): Same.
5768 (vr_values::vrp_visit_switch_stmt): Same.
5769 (vr_values::extract_range_from_phi_node): Same.
5770 (vr_values::simplify_div_or_mod_using_ranges): Same.
5771 (vr_values::simplify_bit_ops_using_ranges): Same.
5772 (test_for_singularity): Same.
5773 (range_fits_type_p): Same.
5774 (vr_values::simplify_cond_using_ranges_1): Same.
5775 (vr_values::simplify_switch_using_ranges): Same.
5776 (vr_values::simplify_float_conversion_using_ranges): Same.
5777 (vr_values::two_valued_val_range_p): Same.
5778 (vr_values::add_equivalence): Move to value_range::equiv_add.
5779 * vr-values.h (vr_values::add_equivalence): Remove.
5780 (VR_INITIALIZER): Remove.
5781 * tree-vrp.c (value_range::set): New.
5782 (value_range::equiv_add): New.
5783 (value_range::value_range): New.
5784 (value_range::deep_copy): New.
5785 (value_range::check): New.
5786 (value_range::equal_p): New.
5787 (value_range::ignore_equivs_equal_p): New.
5788 (value_range::operator==): New.
5789 (value_range::operator!=): New.
5790 (value_range::symbolic_p): New.
5791 (value_range::numeric_p): New.
5792 (value_range::set_undefined): New.
5793 (value_range::set_varying): New.
5794 (value_range::may_contain_p): New.
5795 (value_range::equiv_clear): New.
5796 (value_range::singleton_p): New.
5797 (value_range::intersect): New.
5798 (value_range::dump): New.
5799 (value_range::set_and_canonicalize): New.
5800 (set_value_range): Adjust for value_range API.
5801 (set_value_range_to_undefined): Same.
5802 (set_value_range_to_varying): Same.
5803 (set_and_canonicalize_value_range): Same.
5804 (set_value_range_to_nonnull): Same.
5805 (set_value_range_to_null): Same.
5806 (range_is_null): Same.
5807 (range_is_nonnull): Same.
5808 (range_int_cst_p): Same.
5809 (range_int_cst_singleton_p): Same.
5810 (symbolic_range_p): Same.
5811 (range_includes_zero_p): Same.
5812 (value_range_constant_singleton): Same.
5813 (vrp_set_zero_nonzero_bits): Same.
5814 (ranges_from_anti_range): Same.
5815 (extract_range_into_wide_ints): Same.
5816 (extract_range_from_multiplicative_op): Same.
5817 (set_value_range_with_overflow): Same.
5818 (extract_range_from_binary_expr_1): Same.
5819 (extract_range_from_unary_expr): Same.
5820 (dump_value_range): Same.
5821 (debug_value_range): Same.
5822 (vrp_prop::check_array_ref): Same.
5823 (vrp_prop::check_mem_ref): Same.
5824 (vrp_prop::vrp_initialize): Same.
5825 (vrp_prop::visit_stmt): Same.
5826 (intersect_ranges): Same.
5827 (vrp_prop::visit_phi): Same.
5828 (vrp_prop::vrp_finalize): Same.
5829 (determine_value_range_1): Same.
5830 (determine_value_range): Same.
5831 (vrp_intersect_ranges_1): Rename to...
5832 (vrp_intersect_1): this.
5833 (vrp_intersect_ranges): Rename to...
5834 (value_range::intersect_helper): ...this.
5835 (vrp_meet_1): Rename to...
5836 (value_range::union_helper): ...this.
5837 (vrp_meet): Rename to...
5838 (value_range::union_): ...this.
5839 (copy_value_range): Remove.
5840 * tree-vrp.h (struct value_range): Rewrite into a proper class.
5841 (value_range::vrtype): New.
5842 (value_range::type): New.
5843 (value_range::equiv): New.
5844 (value_range::min): New.
5845 (value_range::max): New.
5846 (value_range::varying_p): New.
5847 (value_range::undefined_p): New.
5848 (value_range::null_p): New.
5849 (value_range::equiv_add): New.
5850 (copy_value_range): Remove.
5852 2018-10-17 David Malcolm <dmalcolm@redhat.com>
5854 * Makefile.in (SELFTEST_TARGETS): New.
5855 (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
5856 (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
5857 (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
5859 (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
5860 (selftest-c++-gdb, selftest-c++-valgrind): Move to
5862 * configure: Regenerate.
5863 * configure.ac (selftest_languages): New.
5865 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
5867 * tree-vrp.c (extract_range_from_multiplicative_op): Remove
5868 overflow wraps argument.
5869 (extract_range_from_binary_expr_1): Do not pass overflow wraps to
5870 wide_int_range_multiplicative_op.
5871 * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
5872 overflow wraps argument.
5873 (wide_int_range_multiplicative_op): Same.
5874 (wide_int_range_lshift): Same.
5875 (wide_int_range_div): Same.
5876 * wide-int-range.h (wide_int_range_multiplicative_op): Same.
5877 (wide_int_range_lshift): Same.
5878 (wide_int_range_div): Same.
5880 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
5882 * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
5883 use sign as argument.
5884 * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
5885 wide_int_range_shift_undefined_p.
5887 2018-10-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5889 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
5891 (@despeculate_copy<ALLI_TI:mode>): ... This.
5892 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
5895 2018-10-16 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5897 * config.gcc: Obsolete *-*-solaris2.10*.
5898 * doc/install.texi (Specific, *-*-solaris2*): Document it.
5900 2018-10-12 Jeff Law <law@redhat.com>
5902 * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
5903 reg + sym +- const_int addressing modes.
5905 2018-10-15 David Malcolm <dmalcolm@redhat.com>
5907 * common.opt (fdiagnostics-minimum-margin-width=): New option.
5908 * diagnostic-show-locus.c (layout::layout): Apply the minimum
5910 (layout::start_annotation_line): Only print up to 3 of the
5911 margin character, to avoid touching the left-hand side.
5912 (selftest::test_diagnostic_show_locus_fixit_lines): Update for
5913 minimum margin width, as set by test_diagnostic_context's ctor.
5914 (selftest::test_fixit_insert_containing_newline): Likewise.
5915 (selftest::test_fixit_insert_containing_newline_2): Likewise.
5916 (selftest::test_line_numbers_multiline_range): Clear
5917 dc.min_margin_width.
5918 * diagnostic.c (diagnostic_initialize): Initialize
5920 * diagnostic.h (struct diagnostic_context): Add field
5922 * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
5923 * opts.c (common_handle_option): Handle
5924 OPT_fdiagnostics_minimum_margin_width_.
5925 * selftest-diagnostic.c
5926 (selftest::test_diagnostic_context::test_diagnostic_context):
5927 Initialize min_margin_width to 6.
5928 * toplev.c (general_init): Initialize global_dc->min_margin_width.
5930 2018-10-15 David Malcolm <dmalcolm@redhat.com>
5932 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
5933 Fix usage of "error_at_rich_loc" in the comment.
5935 2018-10-15 Renlin Li <renlin.li@arm.com>
5938 * tree-vectorizer.c (try_vectorize_loop_1): Don't use
5939 if-conversioned loop when it contains ifn with types not
5940 supported by backend.
5941 * internal-fn.c (expand_direct_optab_fn): Add an assert.
5942 (direct_internal_fn_supported_p): New helper function.
5943 * internal-fn.h (direct_internal_fn_supported_p): Declare.
5945 2018-10-15 Jakub Jelinek <jakub@redhat.com>
5948 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
5949 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
5950 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
5952 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
5954 PR tree-optimization/87022
5955 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
5956 bits in dist vector rather than the first one.
5958 2018-10-15 Richard Biener <rguenther@suse.de>
5961 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
5962 (visit_loadstore): When a used restrict tag escaped verify that
5963 the points-to solution of "other" pointers do not include
5965 (compute_dependence_clique): If a used restrict tag escaped
5966 communicated that down to visit_loadstore.
5968 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
5970 * config/s390/s390.c (s390_expand_vec_init): Force vector element
5971 into reg if it isn't a general operand.
5973 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
5976 * config/i386/sse.md (*vec_dupv2di): Add register source to
5979 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
5982 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
5983 Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
5984 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
5985 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
5987 2018-10-13 Eric Botcazou <ebotcazou@adacore.com>
5989 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
5990 (notice_args_size): Set it in the current trace if no insn that can
5991 throw internally has been seen yet.
5992 (connect_traces): When connecting args_size between traces, allow the
5993 incoming values not to match if there is an insn setting it before the
5994 first insn that can throw internally; in that case, force the creation
5995 of a CFI note on this latter insn.
5997 2018-10-13 Jonathan Wakely <jwakely@redhat.com>
5999 * opt-problem.h (opt_wrapper): Use template-argument-list when naming
6000 the base class, because using the injected-class-name was not clearly
6001 specified until DR 176.
6003 2018-10-12 Paul Koning <ni1d@arrl.net>
6005 * config/pdp11/pdp11.md (doloop_end): New expander.
6006 (doloop_end_insn): renamed from "doloop_end".
6007 (addqi3): New pattern.
6008 (subqi3): New pattern.
6009 * config/pdp11/predicates.md (incdec_operand): New predicate.
6011 2018-10-12 Yury Gribov <tetra2005@gmail.com>
6014 * real.c (format_helper::can_represent_integral_type_p): New function
6015 * real.h (format_helper::can_represent_integral_type_p): Ditto.
6016 * match.pd: New pattern.
6018 2018-10-12 Alexandre Oliva <oliva@adacore.com>
6020 * configure.ac: Introduce --enable-large-address-aware
6021 to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
6022 * doc/install.texi: Document it.
6023 * configure, config.in: Rebuilt.
6024 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
6025 based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
6026 (LINK_SPEC): Insert it.
6027 * config/i386/mingw-w64.h: Likewise.
6029 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
6031 2018-10-12 Peter Bergner <bergner@linux.ibm.com>
6033 PR rtl-optimization/87600
6034 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
6036 2018-10-12 Paul Koning <ni1d@arrl.net>
6038 * doc/md.texi (doloop_end): Document that the pattern code may
6039 need to check operand mode.
6041 2018-10-12 Wilco Dijkstra <wdijkstr@arm.com>
6043 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
6044 to zero-extend between int and floating-point registers.
6045 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
6046 ldp into floating-point registers. Add type and arch attributes.
6047 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
6048 Use f_loads for type attribute.
6050 2018-10-11 Martin Sebor <msebor@redhat.com>
6052 * doc/extend.texi (attribute packed): Correct typos.
6054 2018-10-11 Martin Sebor <msebor@redhat.com>
6056 * doc/extend.texi (attribute flatten): Mention interaction with
6059 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
6062 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6063 Set new_decl virtual flag to zero.
6065 2018-10-11 Martin Sebor <msebor@redhat.com>
6068 * doc/extend.texi (attribute format_arg): Discuss using multiple
6069 attributes on a single function.
6071 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
6073 PR tree-optimization/86829
6074 * match.pd (sin (atan (x))): New simplification rules.
6075 (cos (atan (x))): Likewise.
6076 * real.c (build_sinatan_real): New function.
6077 * real.h (build_sinatan_real): Prototype.
6079 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
6081 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
6083 (fold_mergeeo_helper): New helper function.
6084 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
6085 intrinsics. Correct some whitespace indentation issues.
6087 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
6090 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
6091 Use HOST_WIDE_INT_1U for shift.
6093 2018-10-11 Doug Rupp <rupp@adacore.com>
6094 Olivier Hainque <hainque@adacore.com>
6096 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
6097 Pass --relax to the linker for RTPs.
6098 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
6100 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
6101 Jan Hubicka <jh@suse.cz>
6102 Martin Jambor <mjambor@suse.cz>
6104 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
6105 the same elements are repeated rather than printing all of them.
6106 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
6107 "repeated" elements.
6108 * read-rtl-function.c (test_loading_repeat): New function.
6109 (read_rtl_function_c_tests): Call test_loading_repeat.
6110 * rtl-tests.c (test_dumping_repeat): New function.
6111 (rtl_tests_c_tests): Call test_dumping_repeat.
6113 2018-10-11 Richard Biener <rguenther@suse.de>
6115 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
6116 bdver?_cost): Unify to ...
6117 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
6118 * config/i386/i386.c (processor_cost_table): Adjust.
6120 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
6123 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
6124 the thunk when expanding to GIMPLE.
6126 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
6128 * varasm.c (mergeable_string_section): Don't try to move zero-length
6129 strings to the merge section.
6131 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
6134 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
6136 2018-10-10 Jakub Jelinek <jakub@redhat.com>
6139 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
6140 to special_args set.
6142 2018-10-10 Richard Biener <rguenther@suse.de>
6144 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
6145 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
6146 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
6147 and recursing and pattern terminating the recursion on SSE
6148 vector width using ix86_expand_reduc.
6149 (reduc_sminmax_scal_<mode>): Split into part reducing to half
6150 width and recursing and SSE2 vector variant doing the final
6151 reduction with ix86_expand_reduc.
6152 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
6153 with terminating the recursion at AVX level, splitting that
6156 2018-10-09 David Malcolm <dmalcolm@redhat.com>
6158 * genmatch.c (error_cb): Rename to...
6159 (diagnostic_cb): ...this, converting int params to enums.
6160 (fatal_at): Update for renaming.
6161 (warning_at): Likewise.
6163 * input.c (selftest::ebcdic_execution_charset::apply):
6164 Update for renaming of...
6165 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
6167 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
6168 converting level and reason to enums.
6169 (class selftest::lexer_error_sink): Rename to...
6170 (class selftest::lexer_test_options): ...this, renaming field
6171 "m_errors" to "m_diagnostics".
6172 (selftest::lexer_test_options::apply): Update for renaming of...
6173 (selftest::lexer_test_options::on_error): ...this, renaming to...
6174 (selftest::lexer_test_options::on_diagnostic): ...this
6175 converting level and reason to enums.
6176 (selftest::test_lexer_string_locations_raw_string_unterminated):
6177 Update for renamings.
6178 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
6181 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
6183 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
6184 * config/rs6000/pmmintrin.h: New file.
6186 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
6188 PR tree-optimization/86659
6189 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
6191 2018-10-09 Richard Biener <rguenther@suse.de>
6193 PR tree-optimization/63155
6194 * tree-ssa-structalias.c: Include tree-ssa.h.
6195 (get_constraint_for_ssa_var): For undefs return nothing_id.
6196 (find_func_aliases): Cleanup PHI handling.
6198 2018-10-09 Richard Biener <rguenther@suse.de>
6200 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
6203 2018-10-09 Martin Liska <mliska@suse.cz>
6205 * asan.c (asan_emit_stack_protection): If a stack variable
6206 is located in a same file as current function, then emit
6207 line info into variable definition string.
6209 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
6211 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
6214 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
6216 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
6219 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
6221 PR tree-optimization/86659
6222 * gimple-match.h (struct gimple_match_op): Add reverse field.
6223 (gimple_match_op::set_op): New overloaded method.
6224 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
6225 the REF_REVERSE_STORAGE_ORDER flag on the value.
6226 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
6227 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
6229 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
6232 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
6233 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
6235 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
6238 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
6239 Defined with __builtin_ia32_vfmaddsubpd512_mask.
6241 2018-10-08 Richard Biener <rguenther@suse.de>
6243 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
6244 cost the same as AVX128 ones.
6246 2018-10-08 Paul Koning <ni1d@arrl.net>
6248 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
6249 (expand_block_move): New function.
6250 * config/pdp11/pdp11.c (output_block_move): Remove.
6251 (expand_block_move): New function.
6252 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
6253 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
6254 (*movmemhi1): Remove.
6256 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
6258 * config/s390/2827.md: Increase latencies for some FP instructions.
6260 2018-10-08 Richard Biener <rguenther@suse.de>
6262 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6264 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
6265 * tree-vectorizer.h (dump_stmt_cost): Adjust.
6266 (add_stmt_cost): Dump return value of the hook.
6268 2018-10-08 Richard Biener <rguenther@suse.de>
6270 PR tree-optimization/63155
6271 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
6272 (ssa_propagation_engine::ssa_propagate): Remove redundant
6273 bitmap bit clearing.
6275 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
6277 PR rtl-optimization/86939
6278 PR rtl-optimization/87479
6279 * ira.h (non_conflicting_reg_copy_p): New prototype.
6280 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
6281 (make_hard_regno_dead): Don't add conflicts for register
6282 ignore_reg_for_conflicts.
6283 (make_object_dead): Likewise.
6284 (non_conflicting_reg_copy_p): New function.
6285 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
6286 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
6287 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
6288 (make_hard_regno_dead): Don't add conflicts for register
6289 ignore_reg_for_conflicts. Remove special conflict handling of
6290 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
6291 check_pic_pseudo_p and update callers.
6292 (mark_pseudo_dead): Don't add conflicts for register
6293 ignore_reg_for_conflicts.
6294 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
6296 2018-10-05 Andrew Waterman <andrew@sifive.com>
6297 Jim Wilson <jimw@sifive.com>
6299 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
6300 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
6301 new pattern using HONOR_SNANS that emits one extra instruction.
6303 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
6305 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
6306 patterns): Merge SI and DI patterns to a GPR pattern.
6307 (unnamed define_insn and define_split for record form of that): Merge
6308 to a single define_insn_and_split pattern.
6310 2018-10-05 David Malcolm <dmalcolm@redhat.com>
6313 * input.c (expand_location_to_spelling_point): Add param "aspect"
6314 and use rather than hardcoding LOCATION_ASPECT_CARET.
6315 (get_substring_ranges_for_loc): Handle the case of a single token
6316 within a macro expansion.
6317 * input.h (expand_location_to_spelling_point): Add "aspect" param,
6318 defaulting to LOCATION_ASPECT_CARET.
6320 2018-10-05 Paul Koning <ni1d@arrl.net>
6322 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
6323 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
6324 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
6325 (pdp11_guard_type): New function.
6327 2018-10-05 Paul Koning <ni1d@arrl.net>
6329 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
6330 * config/pdp11/pdp11.opt (mfloat32): Remove.
6332 * doc/invoke.texi (pdp11 -mfloat32): Remove:
6333 (pdp11 -mfloat64): Remove.
6335 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
6337 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
6338 (*cmp<mode>_cc_i387): Ditto.
6339 (*cmpu<mode>_cc_i387): Ditto.
6340 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
6341 * config/i386/i386.c (ix86_expand_fp_compare): Remove
6343 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
6344 Emit x86_sahf_1 pattern.
6345 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
6346 (ix86_expand_carry_flag_compare): Ditto.
6348 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
6350 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
6351 to reg_or_0_operand. Add "C" constraint.
6352 (*cmpxf_cc_i387): Ditto.
6353 (*cmp<mode>_i387): Change operand 2 predicate
6354 to nonimm_or_0_operand. Add "C" constraint.
6355 (*cmp<mode>_cc_i387): Ditto.
6356 (*cmp<mode>_0_i387): Remove insn pattern.
6357 (*cmp<mode>_0_cc_i387): Ditto.
6359 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
6361 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
6362 * config/i386/predicates.md (nonimm_or_0_operand): Rename
6363 from vector_move_operand. Update all uses.
6365 2018-10-05 Martin Sebor <msebor@redhat.com>
6367 PR tree-optimization/87490
6368 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
6371 2018-10-05 Richard Biener <rguenther@suse.de>
6373 PR tree-optimization/63155
6374 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
6375 vertical space in dumpfiles.
6376 * tree-ssa-propagate.h
6377 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
6378 * tree-ssa-propagate.c (cfg_blocks_back): New global.
6379 (ssa_edge_worklist_back): Likewise.
6380 (curr_order): Likewise.
6381 (cfg_blocks_get): Remove abstraction.
6382 (cfg_blocks_add): Likewise.
6383 (cfg_blocks_empty_p): Likewise.
6384 (add_ssa_edge): Add to current or next worklist based on
6386 (add_control_edge): Likewise.
6387 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
6389 (ssa_propagation_engine::ssa_propagate): ... here. Unify
6390 iteration from CFG and SSA edge worklist so we process
6391 everything in RPO order, prioritizing forward progress
6393 (ssa_prop_init): Allocate new worklists, do not dump
6395 (ssa_prop_fini): Free new worklists.
6397 2018-10-05 Richard Biener <rguenther@suse.de>
6399 * tree-core.h (tree_block::abstract_flag): Remove.
6400 (tree_block::block_num): Make full 32bits.
6401 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
6402 * tree.h (BLOCK_ABSTRACT): Remove.
6403 * dwarf2out.c (gen_lexical_block_die): Remove dead code
6404 resulting from BLOCK_ABSTRACT being always false.
6405 (gen_inlined_subroutine_die): Likewise.
6406 (gen_block_die): Likewise.
6407 * tree.c (block_ultimate_origin): Likewise.
6408 * tree-pretty-print.c (dump_block_node): Remove code dealing
6409 with BLOCK_ABSTRACT.
6410 * tree-ssa-live.c (dump_scope_block): Likewise.
6411 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
6412 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
6414 2018-10-05 Richard Biener <rguenther@suse.de>
6416 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
6417 is asked for initialize mode to the component mode of the
6420 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
6423 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
6424 assembler for -mavx.
6425 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
6427 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
6430 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
6432 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
6433 to be DImode. When using mffscrn, force the operand to a register.
6435 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
6437 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
6438 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
6439 X87MODEF mode iterator.
6440 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
6441 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
6442 X87MODEF mode iterator.
6444 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
6446 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
6447 -Wno-prio-ctor-dtor.
6449 2018-10-04 David Malcolm <dmalcolm@redhat.com>
6451 * Makefile.in (OBJS): Add opt-problem.o.
6452 * dump-context.h: Include "selftest.h.
6453 (selftest::temp_dump_context): New forward decl.
6454 (class dump_context): Make friend of class
6455 selftest::temp_dump_context.
6456 (dump_context::dump_loc_immediate): New decl.
6457 (class dump_pretty_printer): Move here from dumpfile.c.
6458 (class temp_dump_context): Move to namespace selftest.
6459 (temp_dump_context::temp_dump_context): Add param
6460 "forcibly_enable_dumping".
6461 (selftest::verify_dumped_text):
6462 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
6463 (selftest::verify_item):
6464 (ASSERT_IS_TEXT): Move here from dumpfile.c.
6465 (ASSERT_IS_TREE): Likewise.
6466 (ASSERT_IS_GIMPLE): Likewise.
6467 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
6469 (dump_context::dump_loc_immediate): ...this new function.
6470 (class dump_pretty_printer): Move to dump-context.h.
6471 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
6472 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
6473 (temp_dump_context::temp_dump_context): Move to "selftest"
6474 namespace. Add param "forcibly_enable_dumping", and use it to
6475 conditionalize the use of m_pp;
6476 (selftest::verify_dumped_text): Make non-static.
6477 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
6478 (selftest::verify_item): Make non-static.
6479 (ASSERT_IS_TEXT): Move to dump-context.h.
6480 (ASSERT_IS_TREE): Likewise.
6481 (ASSERT_IS_GIMPLE): Likewise.
6482 (selftest::test_capture_of_dump_calls): Pass "true" for new
6483 param of temp_dump_context.
6484 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
6485 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
6487 * opt-problem.cc: New file.
6488 * opt-problem.h: New file.
6489 * optinfo-emit-json.cc
6490 (selftest::test_building_json_from_dump_calls): Pass "true" for
6491 new param of temp_dump_context.
6492 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
6493 (optinfo::emit_for_opt_problem): New function.
6494 (optinfo::emit): Clarity which emit_item is used.
6495 * optinfo.h (optinfo::get_dump_location): New accessor.
6496 (optinfo::emit_for_opt_problem): New decl.
6497 (optinfo::emit): Make const.
6498 * selftest-run-tests.c (selftest::run_tests): Call
6499 selftest::opt_problem_cc_tests.
6500 * selftest.h (selftest::opt_problem_cc_tests): New decl.
6501 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
6502 bool to opt_result, converting fprintf messages to
6503 opt_result::failure_at calls. Add "stmt" param for use by the
6505 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
6506 (runtime_alias_check_p): Convert return type from bool to
6507 opt_result, converting dump_printf calls to
6508 opt_result::failure_at, using the statement DDR_A for their
6510 (find_data_references_in_stmt): Convert return type from bool to
6511 opt_result, converting "return false" to opt_result::failure_at
6513 * tree-data-ref.h: Include "opt-problem.h".
6514 (dr_analyze_innermost): Convert return type from bool to opt_result,
6515 and add a const gimple * param.
6516 (find_data_references_in_stmt): Convert return type from bool to
6518 (runtime_alias_check_p): Likewise.
6519 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
6520 dr_analyze_innermost.
6521 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
6522 Convert return type from bool to opt_result, adding a message for
6523 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
6524 (vect_analyze_data_ref_dependence): Convert return type from bool
6525 to opt_result. Change sense of return type from "false"
6526 effectively meaning "no problems" to "false" meaning a problem,
6527 so that "return false" becomes "return opt_result::success".
6528 Convert "return true" calls to opt_result::failure_at, using
6529 the location of statement A rather than vect_location.
6530 (vect_analyze_data_ref_dependences): Convert return type from bool
6532 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
6533 calls to opt_result::failure_at, using the stmt location rather
6535 (vect_verify_datarefs_alignment): Convert return type from bool
6537 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
6538 into multiple more-tightly-scoped copies.
6539 (vect_analyze_data_refs_alignment): Convert return type from bool
6541 (vect_analyze_data_ref_accesses): Likewise, converting a
6542 "return false" to a "return opt_result::failure_at", adding a
6544 (vect_prune_runtime_alias_test_list): Convert return type from
6545 bool to opt_result, converting dump_printf_loc to
6546 opt_result::failure_at. Add a %G to show the pertinent statement,
6547 and use the stmt's location rather than vect_location.
6548 (vect_find_stmt_data_reference): Convert return type from
6549 bool to opt_result, converting dump_printf_loc to
6550 opt_result::failure_at, using stmt's location.
6551 (vect_analyze_data_refs): Convert return type from bool to
6552 opt_result. Convert "return false" to "return
6553 opt_result::failure_at", adding messages as needed.
6554 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
6555 type from bool to opt_result.
6556 (vect_determine_vf_for_stmt): Likewise.
6557 (vect_determine_vectorization_factor): Likewise, converting
6558 dump_printf_loc to opt_result::failure_at, using location of phi
6559 rather than vect_location.
6560 (vect_analyze_loop_form_1): Convert return type from bool to
6561 opt_result, converting dump_printf_loc calls, retaining the use of
6563 (vect_analyze_loop_form): Convert return type from loop_vec_info
6564 to opt_loop_vec_info.
6565 (vect_analyze_loop_operations): Convert return type from bool to
6566 opt_result, converting dump_printf_loc calls, using the location
6567 of phi/stmt rather than vect_location where available. Convert
6568 various "return false" to "return opt_result::failure_at" with
6569 "unsupported phi" messages.
6570 (vect_get_datarefs_in_loop): Convert return type from bool to
6571 opt_result. Add a message for the
6572 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
6573 (vect_analyze_loop_2): Convert return type from bool to
6574 opt_result. Ensure "ok" is set to a opt_result::failure_at before
6575 each "goto again;", adding new messages where needed.
6576 Add "unsupported grouped {store|load}" messages.
6577 (vect_analyze_loop): Convert return type from loop_vec_info to
6579 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
6581 * tree-vect-stmts.c (process_use): Likewise, converting
6582 dump_printf_loc call and using stmt location, rather than
6584 (vect_mark_stmts_to_be_vectorized): Likeise.
6585 (vect_analyze_stmt): Likewise, adding a %G.
6586 (vect_get_vector_types_for_stmt): Convert return type from bool to
6587 opt_result, converting dump_printf_loc calls and using stmt
6588 location, rather than vect_location.
6589 (vect_get_mask_type_for_stmt): Convert return type from tree to
6590 opt_tree, converting dump_printf_loc calls and using stmt location.
6591 * tree-vectorizer.c: Include "opt-problem.h.
6592 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
6593 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
6594 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
6595 enabled, use it to report at the top level "couldn't vectorize
6596 loop" followed by the problem.
6597 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
6598 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
6600 (vect_analyze_stmt): Likewise.
6601 (vect_get_vector_types_for_stmt): Likewise.
6602 (tree vect_get_mask_type_for_stmt): Likewise.
6603 (vect_analyze_data_ref_dependences): Likewise.
6604 (vect_enhance_data_refs_alignment): Likewise.
6605 (vect_analyze_data_refs_alignment): Likewise.
6606 (vect_verify_datarefs_alignment): Likewise.
6607 (vect_analyze_data_ref_accesses): Likewise.
6608 (vect_prune_runtime_alias_test_list): Likewise.
6609 (vect_find_stmt_data_reference): Likewise.
6610 (vect_analyze_data_refs): Likewise.
6611 (vect_analyze_loop): Convert return type from loop_vec_info to
6613 (vect_analyze_loop_form): Likewise.
6614 (vect_analyze_slp): Convert return type from bool to opt_result.
6616 2018-10-04 David Malcolm <dmalcolm@redhat.com>
6618 * doc/invoke.texi (-fopt-info): Document new "internals"
6620 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
6621 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
6623 (optinfo_verbosity_options): Add "internals".
6624 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
6625 (dump_context::apply_dump_filter_p): New member function.
6626 (dump_context::dump_loc): Use apply_dump_filter_p rather than
6627 explicitly masking the dump_kind.
6628 (dump_context::begin_scope): Increment the scope depth first. Use
6629 apply_dump_filter_p rather than explicitly masking the dump_kind.
6630 (dump_context::emit_item): Use apply_dump_filter_p rather than
6631 explicitly masking the dump_kind.
6632 (dump_dec): Likewise.
6633 (dump_hex): Likewise.
6634 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
6635 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
6636 (opt_info_switch_p): Update handling of default
6637 MSG_OPTIMIZED_LOCATIONS to cope with default of
6638 MSG_PRIORITY_USER_FACING.
6639 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
6640 masking the dump_kind.
6641 (selftest::test_capture_of_dump_calls): Update test_dump_context
6642 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
6643 than MSG_ALL. Generalize scope test to be run at all four
6644 combinations of with/without MSG_PRIORITY_USER_FACING and
6645 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
6646 for each of the two values.
6647 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
6648 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
6649 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
6650 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
6651 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
6652 with MSG_PRIORITY_*.
6653 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
6654 dump messages as MSG_PRIORITY_USER_FACING.
6655 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
6656 about the interaction with MSG_PRIORITY_*.
6658 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
6660 * varasm.c (output_constant): Add new parameter merge_strings.
6661 Make strings properly zero terminated in merge string sections.
6662 (mergeable_string_section): Don't fail if the last char is non-zero.
6663 (assemble_variable_contents): Handle merge string sections.
6664 (assemble_variable): Likewise.
6665 (assemble_constant_contents): Likewise.
6666 (output_constant_def_contents): Likewise.
6667 (output_constructor_array_range,
6668 output_constructor_regular_field): Adjust call to output_constant.
6669 (output_object_block): Adjust call to assemble_constant_contents
6670 and assemble_variable_contents.
6672 2018-10-04 Martin Liska <mliska@suse.cz>
6675 * cgraphunit.c (process_function_and_variable_attributes):
6676 Warn about a function with alias attribute and a body.
6678 2018-10-04 Martin Liska <mliska@suse.cz>
6681 * multiple_target.c (redirect_to_specific_clone): New function.
6682 (ipa_target_clone): Use it.
6683 * tree-inline.c: Fix comment.
6685 2018-10-04 David Malcolm <dmalcolm@redhat.com>
6687 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
6689 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
6690 (gcc::dump_manager::register_pass): New member function, adapted
6691 from loop body in gcc::pass_manager::register_pass, adding a
6692 call to update_dfi_for_opt_info.
6693 (gcc::dump_manager::opt_info_enable_passes): Store the
6694 -fopt-info options into the new fields. Move the loop
6696 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
6698 * dumpfile.h (struct opt_pass): New forward decl.
6699 (gcc::dump_manager::register_pass): New decl.
6700 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
6701 (class gcc::dump_manager): Add fields "m_optgroup_flags",
6702 "m_optinfo_flags", and "m_optinfo_filename".
6703 * passes.c (gcc::pass_manager::register_pass): Move all of the
6704 dump-handling code to gcc::dump_manager::register_pass.
6706 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
6708 PR rtl-optimization/87466
6709 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
6710 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
6711 * doc/tm.texi: Regenerate.
6712 * ira-lives.c (process_bb_node_lives): Use the new target hook.
6713 * lra-lives.c (process_bb_lives): Likewise.
6714 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
6717 2018-10-04 Tamar Christina <tamar.christina@arm.com>
6719 * params.c (add_params): Fix initialization.
6721 2018-10-04 Martin Liska <mliska@suse.cz>
6723 PR gcov-profile/84107
6724 * tree-profile.c (init_ic_make_global_vars):
6725 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
6726 Come up with new ic_tuple* variables. Emit
6727 __gcov_indirect_call{,_topn} variables.
6728 (gimple_gen_ic_profiler): Access the variable
6730 (gimple_gen_ic_func_profiler): Access
6731 __gcov_indirect_call.callee field.
6732 (gimple_init_gcov_profiler): Use ptr_type_node.
6733 * value-prof.c (gimple_ic): Use ptr_type_node.
6735 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6737 PR tree-optimization/85787
6738 * ipa-pure-const.c (malloc_candidate_p_1): Move most of
6739 malloc_candidate_p into this function and add support for
6740 detecting multiple phis.
6741 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
6743 2018-10-04 Martin Liska <mliska@suse.cz>
6746 * ipa-inline.c (inline_to_all_callers_1):
6747 Call ultimate_alias_target for node being inlined.
6749 2018-10-03 Jeff Law <law@redhat.com>
6751 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
6753 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
6754 * tree.h (get_typenode_from_name): Prototype.
6756 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
6758 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
6759 Change operand 2 predicate to nonimmediate_operand.
6760 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
6762 2018-10-03 Martin Sebor <msebor@redhat.com>
6763 Jeff Law <law@redhat.com>
6765 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
6767 (get_string_length): Detect unterminated arrays.
6768 (format_string): Same.
6769 (format_directive): Warn about unterminated arrays.
6770 (handle_gimple_call): Mark statements with no_warning as needed.
6772 2018-10-03 Jim Wilson <jimw@sifive.com>
6774 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
6775 also define __riscv_abi_rve. Delete trailing white space.
6777 2018-10-03 Paul Koning <ni1d@arrl.net>
6779 Enable LRA register allocator for PDP11.
6780 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
6783 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
6784 * config/pdp11/pdp11.opt (-mlra): New option.
6785 * doc/invoke.texi (PDP-11 Options): Document -mlra.
6787 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
6789 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
6790 (*<absneg:code>extend<mode>xf2): Ditto.
6792 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
6794 PR tree-optimization/87415
6795 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
6798 2018-10-02 Jeff Law <law@redhat.com>
6800 * gimple-fold.c (get_range_strlen): Only set *nonstr when
6801 an unterminated string is discovered. Bubble up range
6802 even for unterminated strings.
6803 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
6804 indicates the string was not terminated via NONSTR.
6806 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
6808 * tree-vrp.c (extract_range_from_unary_expr): Special case all
6809 pointer conversions.
6810 Do not do anything special for anti-ranges.
6812 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
6814 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
6815 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
6817 2018-10-03 Martin Liska <mliska@suse.cz>
6819 PR gcov-profile/86109
6820 * coverage.c (coverage_begin_function): Do not
6821 mark lambdas as artificial.
6822 * tree-core.h (struct GTY): Remove tm_clone_flag
6823 and introduce new lambda_function.
6824 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
6826 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
6829 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
6830 P8_VECTOR and VSX are enabled.
6832 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
6834 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
6835 0x3907 as CPU model number.
6837 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
6839 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
6840 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
6841 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
6842 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
6843 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
6844 * config/s390/s390.md: Likewise. Rename also the cpu attribute
6845 value from arch12 to z14.
6847 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
6849 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
6851 (isinf<mode>2): Ditto.
6853 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
6855 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
6856 before emitting fxam. Perform calculations in XFmode.
6858 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
6860 * match.pd (((X /[ex] A) +- B) * A): New transformation.
6862 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
6865 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
6866 * tree.c (signed_or_unsigned_type_for): Handle complex.
6868 2018-10-02 Jeff Law <law@redhat.com>
6870 * gimple-fold.c (get_range_strlen): Remove dead code.
6872 2018-10-02 Martin Sebor <msebor@redhat.com>
6873 Jeff Law <law@redhat.com>
6875 * builtins.c (unterminated_array): Add new arguments.
6876 If argument is not terminated, bubble up size and exact
6878 (expand_builtin_strnlen): Detect, avoid expanding
6879 and diagnose unterminated arrays.
6880 (c_strlen): Fill in offset of start of unterminated strings.
6881 * builtins.h (unterminated_array): Update prototype.
6883 2018-10-02 Richard Biener <rguenther@suse.de>
6885 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
6886 of haddv4df, first reduce to SSE width and exploit the fact
6887 that we only need element zero with the reduction result.
6888 (reduc_plus_scal_v2df): Likewise.
6890 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
6892 * dojump.h (do_jump): Delete.
6893 (do_jump_1): Likewise.
6894 (split_comparison): Move around.
6895 * dojump.c (do_jump): Make static.
6896 (do_jump_1): Likewise.
6897 (jumpifnot): Move around.
6898 (jumpifnot_1): Likewise.
6900 (jumpif_1): Likewise.
6901 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
6903 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
6905 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
6906 insns in the delay slot and add_insn_after for the jump insn.
6908 2018-10-02 Richard Biener <rguenther@suse.de>
6910 * tree-inline.c (expand_call_inline): Use the location of
6911 the callee declaration for the inline-entry marker.
6912 * final.c (notice_source_line): Remove special-casing of
6913 NOTE_INSN_INLINE_ENTRY.
6915 2018-10-01 Carl Love <cel@us.ibm.com>
6918 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
6919 (__builtin_mtfsb0): New.
6920 (__builtin_mtfsb1): New.
6921 ( __builtin_set_fpscr_rn): New.
6922 (__builtin_set_fpscr_drn): New.
6923 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
6924 (rs6000_expand_set_fpscr_rn_builtin): Add.
6925 (rs6000_expand_set_fpscr_drn_builtin): Add.
6926 (rs6000_expand_builtin): Add case statement entries for
6927 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
6928 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
6929 RS6000_BUILTIN_MFFSL.
6930 (rs6000_init_builtins): Add ftype initialization and def_builtin
6931 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
6932 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
6933 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
6934 rs6000_mffscdrn): Add define_insn.
6935 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
6936 * doc/extend.texi: Add documentation for the builtins.
6938 2018-10-01 Richard Biener <rguenther@suse.de>
6940 PR tree-optimization/87465
6941 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
6942 causing branch miscounts.
6944 2018-10-01 Tamar Christina <tamar.christina@arm.com>
6946 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
6947 aarch64_option_default_param): New.
6948 (params.h): Include.
6949 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
6950 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
6951 stack-clash protection validation code.
6953 2018-10-01 Tamar Christina <tamar.christina@arm.com>
6955 * params.c (validate_param): New.
6956 (add_params): Use it.
6957 (set_param_value): Refactor param validation into validate_param.
6958 (diagnostic.h): Include.
6959 * diagnostic.h (diagnostic_ready_p): New.
6961 2018-10-01 Tamar Christina <tamar.christina@arm.com>
6963 * params.c (set_param_value):
6964 Add index of parameter being validated.
6965 * common/common-target.def (option_validate_param): New.
6966 * common/common-targhooks.h (default_option_validate_param): New.
6967 * common/common-targhooks.c (default_option_validate_param): New.
6968 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
6969 * doc/tm.texi: Regenerate.
6971 2018-10-01 Tamar Christina <tamar.christina@arm.com>
6974 * config/aarch64/aarch64.c (aarch64_override_options_internal):
6975 Add validation for stack-clash parameters and set defaults.
6977 2018-10-01 Tamar Christina <tamar.christina@arm.com>
6980 * configure.ac: Add stack-clash-protection-guard-size.
6981 * doc/install.texi: Document it.
6982 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
6983 * params.def: Update comment for guard-size.
6984 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
6985 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
6986 * configure: Regenerate.
6988 2018-10-01 Tamar Christina <tamar.christina@arm.com>
6991 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
6992 STACK_DYNAMIC_OFFSET): New.
6993 * config/aarch64/aarch64.c (aarch64_layout_frame):
6994 Update outgoing args size.
6995 (aarch64_stack_clash_protection_alloca_probe_range,
6996 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
6998 2018-10-01 Tamar Christina <tamar.christina@arm.com>
7001 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
7003 * target.def (stack_clash_protection_alloca_probe_range): New.
7004 (stack_clash_protection_final_dynamic_probe): Remove.
7005 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
7006 (default_stack_clash_protection_final_dynamic_probe): Remove.
7007 * targhooks.c: Likewise.
7008 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
7009 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
7010 * doc/tm.texi: Regenerate.
7012 2018-10-01 Tamar Christina <tamar.christina@arm.com>
7015 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
7016 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
7017 aarch64_clamp_to_uimm12_shift): New.
7018 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
7019 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
7021 2018-10-01 Tamar Christina <tamar.christina@arm.com>
7024 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
7026 2018-10-01 Jeff Law <law@redhat.com>
7027 Richard Sandiford <richard.sandiford@linaro.org>
7028 Tamar Christina <tamar.christina@arm.com>
7031 * config/aarch64/aarch64.md
7032 (probe_stack_range): Add k (SP) constraint.
7033 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
7034 STACK_CLASH_MAX_UNROLL_PAGES): New.
7035 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
7036 stack probes for stack clash.
7037 (aarch64_allocate_and_probe_stack_space): New.
7038 (aarch64_expand_prologue): Use it.
7039 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
7040 (aarch64_sub_sp): Add emit_move_imm optional param.
7042 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
7044 PR tree-optimization/87261
7045 * match.pd: Remove trailing whitespace.
7046 Add (x & y) | ~(x | y) -> ~(x ^ y),
7047 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
7049 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
7051 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
7053 (ashlsi3_insn): Update instruction constraints.
7054 (ashrsi3_insn): Likewise.
7055 (rotrsi3): Likewise.
7056 (add_shift): Likewise.
7057 * config/arc/constraints.md (Csz): New 32 bit constraint. It
7058 avoids placing in the limm field small constants which, otherwise,
7059 could end into a small instruction.
7061 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
7063 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
7064 destination register is not odd-even.
7065 (umaddsidi4_split): Likewise.
7067 2018-10-01 Richard Biener <rguenther@suse.de>
7069 * tree-inline.c (expand_call_inline): Store origin of fn
7070 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
7071 * tree.c (block_ultimate_origin): Simplify and do some
7074 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
7076 * config/i386/mmx.md (EMMS): New int iterator.
7077 (emms): New int attribute.
7078 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
7079 EMMS int iterator. Explicitly declare clobbers.
7080 (mmx_emms): Remove expander.
7082 * config/i386/predicates.md (emms_operation): Remove predicate.
7083 (vzeroall_pattern): New predicate.
7084 (vzeroupper_pattern): Rename from vzeroupper_operation.
7085 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
7086 vzeroupper_pattern and vzeroall_pattern predicates.
7088 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
7090 PR rtl-optimization/86939
7091 * ira-lives.c (make_hard_regno_born): Rename from this...
7092 (make_hard_regno_live): ... to this. Remove update to conflict
7093 information. Update function comment.
7094 (make_hard_regno_dead): Add conflict information update. Update
7096 (make_object_born): Rename from this...
7097 (make_object_live): ... to this. Remove update to conflict information.
7098 Update function comment.
7099 (make_object_dead): Add conflict information update. Update function
7101 (mark_pseudo_regno_live): Call make_object_live.
7102 (mark_pseudo_regno_subword_live): Likewise.
7103 (mark_hard_reg_dead): Update function comment.
7104 (mark_hard_reg_live): Call make_hard_regno_live.
7105 (process_bb_node_lives): Likewise.
7106 * lra-lives.c (make_hard_regno_born): Rename from this...
7107 (make_hard_regno_live): ... to this. Remove update to conflict
7108 information. Remove now uneeded check_pic_pseudo_p argument.
7109 Update function comment.
7110 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
7111 to conflict information. Update function comment.
7112 (mark_pseudo_live): Remove update to conflict information. Update
7114 (mark_pseudo_dead): Add conflict information update.
7115 (mark_regno_live): Call make_hard_regno_live.
7116 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
7117 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
7119 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
7122 * config/i386/i386.c (construct_container): Use TImode for
7123 BLKmode values in 2 integer registers.
7125 2018-09-29 Jeff Law <law@redhat.com>
7127 * builtins.c (unterminated_array): Pass in c_strlen_data * to
7128 c_strlen rather than just a tree *.
7129 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
7130 Update recursive calls appropriately. If caller did not provide a
7131 suitable data pointer, create a local one. When a non-terminated
7132 string is discovered, bubble up information about the string via the
7133 c_strlen_data object.
7134 * builtins.h (c_strlen): Update prototype.
7135 (c_strlen_data): New structure.
7136 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
7137 For a type 2 call, if c_strlen indicates a non-terminated string
7138 use the length of the non-terminated string.
7139 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
7141 2018-09-29 Jakub Jelinek <jakub@redhat.com>
7144 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
7145 __m512d type for __A argument rather than __m512.
7147 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
7149 * match.pd (simple_comparison): Don't optimize if either operand is
7150 a function pointer when target needs function pointer canonicalization.
7152 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
7154 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
7155 power5 .. power9 to remove indirection.
7156 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
7157 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
7158 ASM_CPU_476_SPEC): Delete.
7159 (ASM_CPU_SPEC): Adjust.
7160 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
7161 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
7163 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
7165 * config.in: Delete HAVE_AS_DCI.
7166 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
7167 * config/rs6000/rs6000.h: Ditto.
7168 * configure.ac: Delete HAVE_AS_DCI.
7169 * configure: Regenerate.
7171 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
7173 * config.in (HAVE_AS_LWSYNC): Delete.
7174 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
7175 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
7177 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
7178 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
7180 * configure.ac: Delete HAVE_AS_LWSYNC.
7181 * configure: Regenerate.
7183 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
7184 Pierre-Marie de Rodat <derodat@adacore.com>
7186 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
7187 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
7188 (cgraph_node::create_thunk): Add indirect_offset parameter.
7189 (thunk_adjust): Likewise.
7190 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
7191 and initialize the corresponding field with it.
7192 (cgraph_node::dump): Dump indirect_offset field.
7193 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
7194 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
7195 (thunk_adjust): Add indirect_offset parameter and deal with it.
7196 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
7197 pass it to thunk_adjust. Do not call the target hook if it's non-zero
7198 or if the thunk is external or local. Fix formatting. Do not chain
7199 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
7200 if any, in the GIMPLE representation.
7201 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
7202 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
7203 (input_node): Read indirect_offset field.
7204 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
7205 call to thunk_adjust.
7206 * tree-nested.c (struct nesting_info): Add thunk_p field.
7207 (create_nesting_tree): Set it.
7208 (convert_all_function_calls): Copy static chain from targets to thunks.
7209 (finalize_nesting_tree_1): Return early for thunks.
7210 (unnest_nesting_tree_1): Do not finalize thunks.
7211 (gimplify_all_functions): Do not gimplify thunks.
7213 2018-09-28 David Malcolm <dmalcolm@redhat.com>
7215 * opt-suggestions.c (option_proposer::build_option_suggestions):
7216 Release "option_values".
7218 2018-09-28 David Malcolm <dmalcolm@redhat.com>
7220 * coverage.c (get_coverage_counts): Convert problem-reporting dump
7221 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
7222 * dumpfile.c (kind_as_string): New function.
7223 (dump_loc): Rather than a hardcoded prefix of "note: ", use
7224 kind_as_string to vary the prefix based on dump_kind.
7225 (selftest::test_capture_of_dump_calls): Update for above.
7227 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
7229 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
7230 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
7232 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
7234 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
7235 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
7236 INVALID_REGNUM instead of FPSR_REG.
7237 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
7238 * config/i386/i386.md: Update comment of FP compares.
7239 (fldenv): Do not clobber FPSR_REG.
7241 2018-09-28 Richard Biener <rguenther@suse.de>
7243 * tree.h (BLOCK_ORIGIN): New.
7244 * omp-expand.c (grid_expand_target_grid_body): Assign
7245 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
7246 * tree-inline.c (remap_block): Likewise.
7247 * auto-profile.c (get_function_decl_from_block): Simplify
7248 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
7249 * langhooks.c (lhd_print_error_function): Likewise.
7250 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
7252 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7253 * tree.c (block_nonartificial_location): Likewise.
7254 (block_ultimate_origin): Likewise.
7255 * tree-pretty-print.c (percent_K_format): Likewise. Remove
7256 no longer needed LTO case.
7258 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
7259 Jan Hubicka <jh@suse.cz>
7260 Martin Jambor <mjambor@suse.cz>
7262 * simplify-rtx.c (simplify_merge_mask): New function.
7263 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
7264 same masks are used in op1 or op2.
7265 (test_vec_merge): New function.
7266 (test_vector_ops): Call test_vec_merge.
7268 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
7270 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
7271 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
7272 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
7273 (cypress_costs): Set it.
7274 (supersparc_costs): Likewise.
7275 (hypersparc_costs): Likewise.
7276 (leon_cost): Likewise.
7277 (leon3_costs): Likewise.
7278 (sparclet_costs): Likewise.
7279 (ultrasparc_costs): Likewise.
7280 (ultrasparc_costs): Likewise.
7281 (niagara_costs): Likewise.
7282 (niagara2_costs): Likewise.
7283 (niagara3_costs): Likewise.
7284 (niagara4_costs): Likewise.
7285 (niagara7_costs): Likewise.
7286 (m8_costs): Likewise.
7287 (TARGET_CAN_FOLLOW_JUMP): Define.
7288 (pass_work_around_errata::gate): Minor tweak.
7289 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
7290 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
7291 Automaitcally clear MASK_FSMULD mask for V7 processors.
7292 (sparc_can_follow_jump): New static function.
7293 (output_ubranch): Deal with CROSSING_JUMP_P.
7294 (sparc_use_sched_lookahead): Rewrite using switch statement.
7295 (sparc_issue_rate): Reorder.
7296 (sparc_branch_cost): New function.
7298 2018-09-27 Martin Sebor <msebor@redhat.com>
7300 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
7301 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
7302 (int_fits_type_p): Same.
7304 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
7306 * config/i386/i386.md (FPCR_REG): Remove.
7307 (UNSPEC_FLDCW): Remove.
7308 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
7309 (x86_fldcw_1): Remove insn pattern.
7310 (fnstenv): Do not clobber FPCR_REG.
7312 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
7313 (CALL_USED_REGISTERS): Ditto.
7314 (REG_ALLOC_ORDER): Ditto.
7315 (REG_CLASS_CONTENTS): Ditto.
7316 (HI_REGISTER_NAMES): Ditto.
7317 (ADDITIONAL_REGISTER_NAMES): Use defines instead
7318 of numerical constants.
7319 * config/i386/i386.c (regclass_map): Remove fpsr register.
7320 (dbx_register_map): Ditto.
7321 (dbx64_register_map): Ditto.
7322 (svr4_dbx_register_map): Ditto.
7323 (print_reg): Do not handle FPCR_REG.
7325 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
7328 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
7329 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
7330 Delete, always treat as true.
7331 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
7332 Ditto. Simplify remaining code.
7333 * config/powerpcspe/powerpcspe.h: Ditto.
7334 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
7335 Simplify remaining code.
7336 (rs6000_expand_builtin): Ditto.
7337 * config/rs6000/rs6000.h: Ditto.
7338 * configure.ac: Ditto.
7339 * configure: Regenerate.
7341 2018-09-27 Martin Liska <mliska@suse.cz>
7343 * coverage.c (get_coverage_counts): Revert the formatting
7344 of missing profile opt info.
7346 2018-09-27 Richard Biener <rguenther@suse.de>
7350 * dwarf2out.c (set_block_origin_self): Do not mark outermost
7351 block as we do not output that.
7352 (gen_inlined_subroutine_die): Elide the originally outermost
7353 block, matching what we do for concrete instances.
7354 (decls_for_scope): Add parameter specifying whether to recurse
7357 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
7358 Tom de Vries <tom@codesourcery.com>
7362 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
7363 STORE_FLAG_VALUE == 1.
7365 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
7367 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
7368 constant definitions.
7369 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
7370 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
7371 ("speculation_barrier"): New expander definition.
7373 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
7375 PR gcov-profile/86957
7376 * common.opt: New warning option -Wmissing-profile.
7377 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
7378 * doc/invoke.texi: Document -Wmissing-profile.
7380 2018-09-26 Jim Wilson <jimw@sifive.com>
7382 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
7383 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
7385 2018-09-26 Martin Sebor <msebor@redhat.com>
7387 * tree.c (zerop): Change return type to bool.
7388 (integer_zerop, integer_onep, integer_each_onep): Same.
7389 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
7390 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
7391 (real_onep, real_minus_onep, chain_index): Same.
7392 (print_type_hash_statistics, type_list_equal): Same.
7393 * tree.h (zerop): Same.
7394 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
7395 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
7396 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
7397 (real_onep, real_minus_onep, chain_index): Same.
7398 (print_type_hash_statistics, type_list_equal): Same.
7400 2018-09-26 Jim Wilson <jimw@sifive.com>
7402 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
7404 2018-09-26 Jakub Jelinek <jakub@redhat.com>
7407 * config/i386/i386.c: Include debug.h and dwarf2out.h.
7408 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
7411 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
7413 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
7415 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
7417 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
7418 and FP_SECOND_SSE_REGS.
7419 (REG_CLASS_NAMES): Ditto.
7420 (REG_CLASS_CONTENTS): Ditto.
7421 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
7422 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
7423 (ix86_preferred_output_reload_class): Ditto.
7424 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
7425 clobber constraint to "=&f".
7426 (fix_truncdi_i387): Ditto.
7427 (lrintxfdi2): Ditto.
7428 (fistdi2_<rounding>): Ditto.
7429 (fpremxf4_i387): Change "=u" constraint to "=f".
7430 (fprem1xf4_i387): Ditto.
7432 (fptanxf4_i387): Ditto.
7433 (fxtractxf3_i387): Ditto.
7434 (fscalexf4_i387): Ditto.
7435 (atan2xf3): Change "u" constraint to "f".
7436 (fyl2xxf3_i387): Ditto.
7437 (fyl2xp1xf3_i387): Ditto.
7439 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
7442 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
7443 for removed I387_MASK_PM entity.
7446 2018-09-26 Jeff Law <law@redhat.com>
7448 2018-09-26 Alexey Neyman <stilor@att.net>
7450 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
7451 headers are no longer pulled in by <isl/val.h>.
7453 2018-09-26 Richard Biener <rguenther@suse.de>
7456 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
7457 or concrete instance DIE to the tree. Create abstract origin
7458 attributes also for concrete instances.
7460 2018-09-26 Alexey Neyman <stilor@att.net>
7462 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
7463 headers are no longer pulled in by <isl/val.h>.
7465 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
7467 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
7468 Use new helper functions.
7469 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
7470 Use new helper functions.
7471 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
7472 aarch_mm_needs_release): New declarations.
7473 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
7474 aarch_mm_needs_release): New.
7476 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
7478 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
7479 (arm32_output_mi_thunk): Deal with long calls.
7481 2018-09-26 Richard Biener <rguenther@suse.de>
7485 * tree-inline.c (expand_call_inline): When the location
7486 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
7487 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
7488 the inserted BLOCK to make inlined_function_outer_scope_p
7490 * dwarf2out.c (add_call_src_coords_attributes): Do not add
7491 coords for reserved locations.
7493 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
7495 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
7496 (*call_indirect_nonlocal_sysv<mode>): Ditto.
7497 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
7498 (*sibcall_nonlocal_sysv<mode>): Ditto.
7499 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
7500 (<bd>_<mode>): Ditto.
7501 (<bd>tf_<mode>): Ditto.
7503 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
7505 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
7506 control string as a list of templates instead of as C code.
7507 (*altivec_movti): Ditto.
7508 * config/rs6000/darwin.md (movdf_low_di): Ditto.
7510 2018-09-25 Jim Wilson <jimw@sifive.com>
7512 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
7513 when target symbol is weak.
7515 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
7518 * builtins.c (unterminated_array): Simplify.
7519 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
7520 where pointer arithmetic is safe.
7522 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
7525 * config/rs6000/altivec.md (altivec_vspltb): Use
7526 const_0_to_15_operand instead of u5bit_cint_operand.
7527 (*altivec_vspltb_internal): Ditto.
7528 (altivec_vspltb_direct): Ditto.
7529 (altivec_vsplth): Use const_0_to_7_operand instead of
7531 (*altivec_vsplth_internal): Ditto.
7532 (altivec_vsplth_direct): Ditto.
7533 (altivec_vspltw): Use const_0_to_3_operand instead of
7535 (*altivec_vspltw_internal): Ditto.
7536 (altivec_vspltw_direct): Ditto.
7537 (altivec_vspltsf): Ditto.
7538 (*altivec_vspltsf_internal): Ditto.
7539 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
7540 various splats with the proper size immediate. Reorder the various
7541 cases by ascending size of immediate, and put all such together.
7543 2018-09-25 Richard Biener <rguenther@suse.de>
7546 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
7548 (maybe_create_die_with_external_ref): Do not create
7549 DW_TAG_imported_unit here.
7550 (add_abstract_origin_attribute): Handle external BLOCK refs.
7551 (dwarf2out_abstract_function): Simplify LTO case.
7552 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
7553 rather than using maybe_create_die_with_external_ref.
7555 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
7558 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
7559 (frndintxf2_mask_pm_i387): Ditto.
7560 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
7561 Enable for !flag_trapping_math.
7562 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
7563 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
7564 Change operand 1 predicate to nonimmediate_operand.
7565 (attr "i387_cw"): Remove mask_pm.
7566 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
7567 (enum ix86_entity): Remove I387_MASK_PM.
7568 * config/i386/i386.c (ix86_i387_mode_needed): Do not
7569 handle I387_MASK_PM.
7570 (ix86_mode_needed): Ditto.
7571 (ix86_mode_after): Ditto.
7572 (ix86_mode_entry): Ditto.
7573 (ix86_mode_exit): Ditto.
7574 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
7576 2018-09-25 Jakub Jelinek <jakub@redhat.com>
7578 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
7579 to_update_switch_stmts to vNULL instead of calling create on them
7582 2018-09-25 Richard Biener <rguenther@suse.de>
7584 PR tree-optimization/87402
7585 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
7586 (visit_phi): Re-instantiate handling of supposed to be VARYING
7587 but non-VARYING backedge value.
7589 2018-09-25 Richard Biener <rguenther@suse.de>
7592 * dwarf2out.c (struct sym_off_pair): New.
7593 (external_die_map): New global.
7594 (lookup_decl_die): When in LTO create DIEs lazily from the
7596 (lookup_block_die): New function, create DIEs lazily in LTO.
7597 (equate_block_to_die): New function.
7598 (dwarf2out_die_ref_for_decl): During WPA get the association
7599 from the external DIE map.
7600 (dwarf2out_register_external_die): Record mapping into the
7602 (maybe_create_die_with_external_ref): New function split out from
7603 DIE generation part of old dwarf2out_register_external_die.
7604 (add_abstract_origin_attribute): Do not return the DIE. When
7605 in LTO reference externals directly.
7606 (dwarf2out_abstract_function): When in LTO ignore calls for
7607 decls with external DIEs (already present abstract instances).
7608 (gen_call_site_die): Adjust.
7609 (add_high_low_attributes): Likewise.
7610 (gen_lexical_block_die): Likewise.
7611 (gen_inlined_subroutine_die): Likewie.
7612 (gen_block_die): Likewise.
7613 (dwarf2out_inline_entry): Likewise.
7614 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
7617 2018-09-25 Martin Liska <mliska@suse.cz>
7619 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
7620 integers and not by a float value.
7622 2018-09-25 Martin Liska <mliska@suse.cz>
7625 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
7627 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
7628 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
7629 * godump.c (go_format_type): Remove extra parenthesis.
7631 2018-09-25 Martin Liska <mliska@suse.cz>
7633 * alias.c (set_dest_equal_p): Remove unused function.
7634 * config/i386/i386.c (def_builtin_pure2): Likewise.
7635 * diagnostic-show-locus.c (class layout): Remove
7637 (layout::layout): Likewise here.
7638 * dump-context.h (class temp_dump_context): Likewise.
7639 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
7640 (add_AT_loclistsptr): Likewise.
7641 (add_AT_offset): Likewise.
7642 (get_AT_hi_pc): Likewise.
7643 (is_comdat_die): Likewise.
7644 (type_is_enum): Likewise.
7645 (ceiling): Likewise.
7646 (add_AT_vms_delta): Likewise.
7647 (is_class_die): Likewise.
7648 * edit-context.c (class line_event): Remove unused field.
7649 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
7651 * ipa-cp.c (ipa_get_vr_lat): Likewise.
7652 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
7653 (ok_for_base_p_nonstrict): Likewise.
7654 * tree-chrec.c (is_not_constant_evolution): Likewise.
7655 (chrec_fold_poly_cst): Likewise.
7656 * tree-if-conv.c (has_pred_critical_p): Likewise.
7657 * tree-ssa-coalesce.c (print_exprs): Likewise.
7658 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
7659 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
7660 * tree-vrp.c (value_ranges_intersect_p): Likewise.
7661 (value_range_nonnegative_p): Likewise.
7663 2018-09-25 Martin Liska <mliska@suse.cz>
7665 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
7666 Do not handle "GNU Pascal".
7667 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
7669 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
7670 from documentation. Likewise.
7671 * dbxout.c (dbxout_range_type): Likewise.
7672 * doc/cpp.texi: Likewise.
7673 * doc/extend.texi: Likewise.
7674 * doc/frontends.texi: Likewise.
7675 * doc/invoke.texi: Remove Pascal entry.
7676 * tree.def (CLEANUP_POINT_EXPR): Likewise.
7677 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
7679 2018-09-25 Martin Liska <mliska@suse.cz>
7682 * doc/invoke.texi: Document all parameters and remove default
7685 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
7688 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
7689 contains HOST_WIDE_INTs when computing its size.
7691 2018-09-24 Jim Wilson <jimw@sifive.com>
7694 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
7696 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
7698 2018-09-24 Andrew Pinski <apinski@marvell.com>
7700 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
7701 access prev before checking it for NULLness in the
7702 AARCH64_FUSE_CMP_BRANCH case.
7704 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
7707 * config/i386/i386.c (rest_of_insert_endbranch): Set
7708 endbr_queued_at_entrance to true and don't insert ENDBR if
7709 x86_function_profiler will be called.
7710 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
7712 * config/i386/i386.h (machine_function): Add
7713 endbr_queued_at_entrance.
7715 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
7717 * genattrtab.c (mk_attr_alt): Use alternative_mask.
7718 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
7720 (check_attr_test): Use alternative_mask.
7721 (get_attr_value): Likewise.
7722 (compute_alternative_mask): Use alternative_mask and XWINT.
7723 (make_alternative_compare): Use alternative_mask.
7724 (attr_alt_subset_p): Use XWINT.
7725 (attr_alt_subset_of_compl_p): Likewise.
7726 (attr_alt_intersection): Use alternative_mask and XWINT.
7727 (attr_alt_union): Likewise.
7728 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
7729 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
7730 (simplify_test_exp): Use alternative_mask and XWINT.
7731 (write_test_expr): Use alternative_mask and XWINT, adjust bit
7732 number calculation to support 64 bits. Generate code that
7733 checks 64-bit masks.
7734 (main): Use alternative_mask.
7735 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
7737 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
7740 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
7741 RETURN+USE when returning via %r14.
7743 2018-09-24 Martin Liska <mliska@suse.cz>
7745 * gcov.c (output_lines): Print colorization legend
7746 for both flag_use_colors and flag_use_hotness_colors.
7749 2018-09-24 Martin Liska <mliska@suse.cz>
7751 * coverage.c (get_coverage_counts): Use warning_at
7752 with current_function_decl location. Use %qD in warning
7755 2018-09-24 Martin Liska <mliska@suse.cz>
7757 * memory-block.h (memory_block_pool::release): Annotate with
7758 valgrind that the memory is not accessible.
7760 2018-09-24 Martin Liska <mliska@suse.cz>
7763 * asan.c: Make asan_handled_variables extern.
7765 * cfgexpand.c (expand_stack_vars): Make sure
7766 a representative is unpoison if another
7767 variable in the partition is handled by
7768 use-after-scope sanitization.
7770 2018-09-24 Richard Biener <rguenther@suse.de>
7772 PR tree-optimization/63155
7773 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
7774 the worklist when the edge of the respective argument isn't
7777 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
7779 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
7780 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
7781 (MASK_CLASS_P): Update for rename.
7782 (MAYBE_MASK_CLASS_P): Ditto.
7783 (REG_CLASS_NAMES): Update.
7784 (REG_CLASS_CONTENT): Update.
7785 * config/i386/i386.c (regclass_map): Update for MASK_REG
7786 and ALL_MASK_REGS rename.
7787 * config/i386/constraints.md (Yk): Update for rename.
7790 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
7792 * config/i386/i386.h (enum reg_class): Remove
7793 EVEX_SSE_REGS and MOD4_SSE_REGS.
7794 (REG_CLASS_NAMES): Update.
7795 (REG_CLASS_CONTENT): Update.
7796 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
7797 registers as ALL_SSE_REGS.
7798 (ix86_additional_allocno_class_p): Remove.
7799 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
7800 (ix86_register_priority): Lower priority of EVEX SSE registers.
7801 Use IN_RANGE macro where appropriate.
7802 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
7803 AVX-5124VNNIW checks.
7804 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
7805 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
7806 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
7807 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
7808 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
7809 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
7810 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
7811 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
7812 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
7813 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
7814 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
7815 * config/i386/constraints.md (Yh): Remove.
7817 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
7819 * config/i386/i386.c (regclass_map): Declare integer REX registers
7822 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
7824 * doc/service.texi (Service): Switch the fsf.org link to https.
7826 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
7829 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
7830 Define to speculation_safe_value_not_needed.
7832 2018-09-21 Florian Weimer <fweimer@redhat.com>
7835 * doc/extend.texi (Common Function Attributes): Mention that
7836 noreturn suppresses tail call optimization.
7838 2018-09-21 Jeff Law <law@redhat.com>
7840 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
7841 vr_values::cleanup_edges_and_switches.
7842 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
7844 (identify_jump_threads): Remove EDGE_IGNORE handling.
7845 (execute_vrp): Move handling of to_remove_edges and
7846 to_update_switch_stmts into vr_values class member functions.
7847 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
7848 (to_update_switch_stmts): Likewise.
7849 * vr-values.c: Include cfghooks.h.
7850 (vr_values::vr_values): Initialize to_remove_edges and
7851 to_update_switch_stmts.
7852 (vr_values::~vr_values): Verify to_remove_edges and
7853 to_update_switch_stmts are empty.
7854 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
7855 (vr_values::cleanup_edges_and_switches): New member function.
7856 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
7857 function. Add new data members.
7859 2018-09-21 David Malcolm <dmalcolm@redhat.com>
7861 PR tree-optimization/87309
7862 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
7863 calls with pflags and alt_flags.
7864 (selftest::test_capture_of_dump_calls): Add test of interaction of
7865 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
7867 2018-09-21 Olivier Hainque <hainque@adacore.com>
7869 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
7871 2018-09-21 Olivier Hainque <hainque@adacore.com>
7873 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
7875 2018-09-21 Olivier Hainque <hainque@adacore.com>
7877 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
7878 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
7879 (VXWORKS_LIBS_RTP): Minor reordering.
7881 2018-09-21 Olivier Hainque <hainque@adacore.com>
7883 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
7884 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
7886 2018-09-21 Olivier Hainque <hainque@adacore.com>
7888 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
7889 (PTRDIFF_TYPE): Likewise.
7891 2018-09-21 Olivier Hainque <hainque@adacore.com>
7893 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
7894 triplet, similar to support for VxWorks7.
7895 * config/vxworks-dummy.h: Provide a default definition
7896 of TARGET_VXWORKS64 to 0.
7898 2018-09-21 Olivier Hainque <hainque@adacore.com>
7900 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
7901 * config/vxworks-dummy.h: here.
7903 2018-09-21 Olivier Hainque <hainque@adacore.com>
7905 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
7907 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
7908 Bo Zhou <zbo.zhou@hisilicon.com>
7910 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
7911 * config/aarch64/aarch64-tune.md: Regenerated.
7912 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
7913 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
7914 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
7916 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
7917 Julian Brown <julian@codesourcery.com>
7919 * builtins.c (get_builtin_sync_mem): Handle address spaces.
7921 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
7923 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
7924 if the call takes a static chain.
7926 2018-09-21 Martin Liska <mliska@suse.cz>
7928 * auto-profile.c (autofdo_source_profile::read): Do not
7930 (read_profile): Do not add working sets.
7931 (read_autofdo_file): Remove sum_all.
7932 (afdo_callsite_hot_enough_for_early_inline): Remove const
7934 * coverage.c (struct counts_entry): Remove gcov_summary.
7935 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
7936 do not support GCOV_TAG_PROGRAM_SUMMARY.
7937 (get_coverage_counts): Remove summary and expected
7939 * coverage.h (get_coverage_counts): Likewise.
7940 * doc/gcov-dump.texi: Remove -w option.
7941 * gcov-dump.c (dump_working_sets): Remove.
7942 (main): Do not support '-w' option.
7943 (print_usage): Likewise.
7944 (tag_summary): Likewise.
7945 * gcov-io.c (gcov_write_summary): Do not dump
7947 (gcov_read_summary): Likewise.
7948 (gcov_histo_index): Remove.
7949 (gcov_histogram_merge): Likewise.
7950 (compute_working_sets): Likewise.
7951 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
7953 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
7954 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
7955 (GCOV_HISTOGRAM_SIZE): Remove.
7956 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
7957 (struct gcov_summary): Simplify rapidly just
7958 to runs and sum_max fields.
7959 (gcov_histo_index): Remove.
7960 (NUM_GCOV_WORKING_SETS): Likewise.
7961 (compute_working_sets): Likewise.
7962 * gcov-tool.c (print_overlap_usage_message): Remove
7963 trailing empty line.
7964 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
7965 (output_lines): Remove program related line.
7966 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
7967 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
7969 (input_profile_summary): Do not read it.
7970 (merge_profile_summaries): And do not merge it.
7971 (input_symtab): Do not call removed function.
7972 * modulo-sched.c (sms_schedule): Do not print sum_max.
7973 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
7974 removed when histogram method was invented.
7975 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
7977 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
7979 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
7980 and dump selected value.
7981 * profile.c (add_working_set): Remove.
7982 (get_working_sets): Likewise.
7983 (find_working_set): Likewise.
7984 (get_exec_counts): Do not work with working sets.
7985 (read_profile_edge_counts): Do not inform as sum_max is removed.
7986 (compute_branch_probabilities): Likewise.
7987 (compute_value_histograms): Remove argument for call of
7988 get_coverage_counts.
7989 * profile.h: Do not make gcov_summary const.
7991 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
7993 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
7995 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
7997 PR tree-optimization/86990
7998 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
7999 Check that the entire merged store group is made of constants only for
8002 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
8004 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
8005 (VTABLE_VERIFICATION_SPEC): Likewise.
8006 (SANITIZER_EARLY_SPEC): Likewise.
8007 (SANITIZER_SPEC): Likewise.
8008 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
8009 * doc/invoke.texi (Link Options): Document -r.
8011 2018-09-20 Richard Biener <rguenther@suse.de>
8014 * gimplify.c (gimplify_expr): Retain alignment of
8015 addressable lvalue in dereference.
8017 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
8020 * configure.ac: Check for .loc is_stmt support.
8021 * configure, config.in: Rebuilt.
8022 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
8025 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
8027 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
8028 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
8031 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
8033 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
8035 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
8038 2018-09-20 Marek Polacek <polacek@redhat.com>
8040 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
8042 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
8044 PR tree-optimization/87288
8045 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
8046 into account when determining PEELING_FOR_NITERS.
8048 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
8050 PR tree-optimization/86877
8051 * tree-vect-loop.c (vect_analyze_loop_2): Call
8052 vect_verify_datarefs_alignment.
8054 2018-09-19 Marek Polacek <polacek@redhat.com>
8056 * doc/invoke.texi: Document -Wclass-conversion.
8058 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
8060 * config/pa/pa.c (pa_adjust_priority): Delete.
8061 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
8063 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
8064 (atomic_storehi): Likewise.
8065 (atomic_storesi): Likewise.
8066 (atomic_loaddi): Restore compare and swap exchange loop code.
8068 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
8070 PR rtl-optimization/86902
8071 * combine.c (try_combine): When changing the CC mode used, don't change
8072 an unrelated mode in other_insn to that new CC mode.
8074 2018-09-19 David Malcolm <dmalcolm@redhat.com>
8076 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
8077 with %T in place of calls to dump_generic_expr.
8078 (prune_runtime_alias_test_list): Likewise.
8079 (create_runtime_alias_checks): Likewise.
8080 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
8081 (vect_analyze_data_ref_dependence): Likewise.
8082 (vect_slp_analyze_data_ref_dependence): Likewise.
8083 (vect_record_base_alignment): Likewise. Use %G in place of call
8084 to dump_gimple_stmt.
8085 (vect_compute_data_ref_alignment): Likewise.
8086 (verify_data_ref_alignment): Likewise.
8087 (vect_find_same_alignment_drs): Likewise.
8088 (vect_analyze_group_access_1): Likewise.
8089 (vect_analyze_data_ref_accesses): Likewise.
8090 (dependence_distance_ge_vf): Likewise.
8091 (dump_lower_bound): Likewise.
8092 (vect_prune_runtime_alias_test_list): Likewise.
8093 (vect_find_stmt_data_reference): Likewise.
8094 (vect_analyze_data_refs): Likewise.
8095 (vect_create_addr_base_for_vector_ref): Likewise.
8096 (vect_create_data_ref_ptr): Likewise.
8097 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
8098 (vect_can_advance_ivs_p): Likewise.
8099 (vect_update_ivs_after_vectorizer): Likewise.
8100 (vect_gen_prolog_loop_niters): Likewise.
8101 (vect_prepare_for_masked_peels): Likewise.
8102 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
8103 (vect_determine_vectorization_factor): Likewise.
8104 (vect_is_simple_iv_evolution): Likewise.
8105 (vect_analyze_scalar_cycles_1): Likewise.
8106 (vect_analyze_loop_operations): Likewise.
8107 (report_vect_op): Likewise.
8108 (vect_is_slp_reduction): Likewise.
8109 (check_reduction_path): Likewise.
8110 (vect_is_simple_reduction): Likewise.
8111 (vect_create_epilog_for_reduction): Likewise.
8112 (vect_finalize_reduction:): Likewise.
8113 (vectorizable_induction): Likewise.
8114 (vect_transform_loop_stmt): Likewise.
8115 (vect_transform_loop): Likewise.
8116 (optimize_mask_stores): Likewise.
8117 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
8118 (vect_split_statement): Likewise.
8119 (vect_recog_over_widening_pattern): Likewise.
8120 (vect_recog_average_pattern): Likewise.
8121 (vect_determine_min_output_precision_1): Likewise.
8122 (vect_determine_precisions_from_range): Likewise.
8123 (vect_determine_precisions_from_users): Likewise.
8124 (vect_mark_pattern_stmts): Likewise.
8125 (vect_pattern_recog_1): Likewise.
8126 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
8127 (vect_record_max_nunits): Likewise.
8128 (vect_build_slp_tree_1): Likewise.
8129 (vect_build_slp_tree_2): Likewise.
8130 (vect_print_slp_tree): Likewise.
8131 (vect_analyze_slp_instance): Likewise.
8132 (vect_detect_hybrid_slp_stmts): Likewise.
8133 (vect_detect_hybrid_slp_1): Likewise.
8134 (vect_slp_analyze_operations): Likewise.
8135 (vect_slp_analyze_bb_1): Likewise.
8136 (vect_transform_slp_perm_load): Likewise.
8137 (vect_schedule_slp_instance): Likewise.
8138 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
8139 (vect_mark_stmts_to_be_vectorized): Likewise.
8140 (vect_init_vector_1): Likewise.
8141 (vect_get_vec_def_for_operand): Likewise.
8142 (vect_finish_stmt_generation_1): Likewise.
8143 (vect_check_load_store_mask): Likewise.
8144 (vectorizable_call): Likewise.
8145 (vectorizable_conversion): Likewise.
8146 (vectorizable_operation): Likewise.
8147 (vectorizable_load): Likewise.
8148 (vect_analyze_stmt): Likewise.
8149 (vect_is_simple_use): Likewise.
8150 (vect_get_vector_types_for_stmt): Likewise.
8151 (vect_get_mask_type_for_stmt): Likewise.
8152 * tree-vectorizer.c (increase_alignment): Likewise.
8154 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
8156 * doc/rtl.texi: Adjust vec_select description.
8157 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
8158 non-constant selectors.
8160 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
8162 * config/aarch64/aarch64-protos.h
8163 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
8164 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
8165 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
8166 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
8167 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
8168 (AARCH64_FL_PROFILE): Move index so flags are ordered.
8169 (AARCH64_ISA_RCPC8_4): New flag.
8170 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
8171 to aarch64_offset_9bit_signed_unscaled_p.
8172 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
8174 * config/aarch64/constraints.md (Ust): New constraint.
8175 * config/aarch64/predicates.md.
8176 (aarch64_9bit_offset_memory_operand): New predicate.
8177 (aarch64_rcpc_memory_operand): New predicate.
8179 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
8181 PR rtl-optimization/87361
8182 * rtlanal.c (nonzero_bits1): Revert accidental change.
8184 2018-09-19 Richard Biener <rguenther@suse.de>
8186 PR tree-optimization/87349
8187 PR tree-optimization/87342
8188 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
8190 2018-09-18 Marek Polacek <polacek@redhat.com>
8192 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
8193 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
8195 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
8197 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
8199 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
8201 PR rtl-optimization/86882
8202 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
8204 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
8206 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
8207 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
8209 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
8212 * doc/invoke.texi (Link Options): Fix formatting and grammar.
8214 2018-09-18 Richard Biener <rguenther@suse.de>
8217 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
8218 (compute_samebase_partition_bases): Likewise.
8219 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
8220 (gimple_can_coalesce_p): Simplify.
8222 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
8224 Handle a library implementation of ffs calling __builtin_ffs.
8225 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
8226 (mmix_init_libfuncs): New function: make __builtin_ffs expand
8229 2018-09-17 David Malcolm <dmalcolm@redhat.com>
8231 * diagnostic-show-locus.c (class layout_range): Add field
8233 (layout_range::layout_range): Add "original_idx" param and use it
8234 to initialize new field.
8235 (make_range): Use 0 for original_idx.
8236 (layout::layout): Pass in index to calls to
8237 maybe_add_location_range.
8238 (layout::maybe_add_location_range): Add param "original_idx" and
8239 pass it on to layout_range.
8240 (layout::print_any_labels): Pass on range->m_original_idx to
8242 (gcc_rich_location::add_location_if_nearby): Use 0 for
8244 * gcc-rich-location.h (text_range_label::get_text): Update for new
8246 (range_label_for_type_mismatch::get_text): Likewise.
8248 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
8250 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
8252 2018-09-17 David Malcolm <dmalcolm@redhat.com>
8254 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
8255 format_string_diagnostic_t.
8256 (fmtwarn_n): Likewise.
8257 * substring-locations.c
8258 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
8259 (format_warning_n_va): Convert to...
8260 (format_string_diagnostic_t::emit_warning_n_va): ...this.
8261 (format_warning_va): Convert to...
8262 (format_string_diagnostic_t::emit_warning_va): ...this.
8263 (format_warning_at_substring): Convert to...
8264 (format_string_diagnostic_t::emit_warning): ...this.
8265 (format_warning_at_substring_n): Convert to...
8266 (format_string_diagnostic_t::emit_warning_n): ...this.
8267 * substring-locations.h (class format_string_diagnostic_t): New
8269 (format_warning_va): Convert to
8270 format_string_diagnostic_t::emit_warning_va.
8271 (format_warning_n_va): Convert to
8272 format_string_diagnostic_t::emit_warning_n_va.
8273 (format_warning_at_substring): Convert to
8274 format_string_diagnostic_t::emit_warning.
8275 (format_warning_at_substring_n): Convert to
8276 format_string_diagnostic_t::emit_warning_n.
8278 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
8279 Bernd Schmidt <bernds_cb1@t-online.de>
8281 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
8284 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
8286 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
8287 operand 0 predicate to nonimmediate operand.
8288 (rint<mode>2_frndint): Remove insn pattern.
8289 (rint<mode>2): Change operand 1 predicate to general_operand.
8290 Extend operand 1 to XFmode and generate rintxf2 insn.
8291 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
8292 Do not use X87MODEF mode macro.
8293 (frndintxf2_<rounding>_i387): Rename from
8294 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
8295 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
8296 to XFmode and generate significandxf3 insn.
8298 2018-09-17 Richard Biener <rguenther@suse.de>
8300 PR tree-optimization/87328
8301 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
8302 visiting unexecutable backedges when not iterating.
8303 (do_rpo_vn): Mark all edges not executable even when not
8306 2018-09-17 Martin Jambor <mjambor@suse.cz>
8309 * doc/invoke.texi (Warning Options): Likewise.
8311 2018-09-17 Richard Biener <rguenther@suse.de>
8313 PR tree-optimization/87301
8314 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
8315 clean EH info from leftover copy assignments.
8317 2018-09-17 Martin Liska <mliska@suse.cz>
8319 PR gcov-profile/85871
8320 * gcov.c (output_intermediate_file): Fix out of bounds
8323 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
8325 * config/arc/arc.c: Object attributes for core4 not reflected
8327 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
8330 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
8332 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
8334 2018-09-17 Martin Liska <mliska@suse.cz>
8336 * doc/gcov.texi: Document new option --use-hotness-colors.
8337 * gcov.c (struct source_info): Declare new field.
8338 (source_info::source_info): Set default for maximum_count.
8339 (print_usage): Add new -q option.
8340 (process_args): Process it.
8341 (accumulate_line_info): Save src->maximum_count.
8342 (output_line_beginning): Make color line number if
8343 flag_use_hotness_colors is set.
8344 (output_line_details): Pass default argument value.
8345 (output_lines): Pass src->maximum_count.
8347 2018-09-17 Martin Liska <mliska@suse.cz>
8349 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
8350 Use processor_names table.
8351 * config/i386/i386.c (ix86_default_align): Use
8352 processor_cost_table for alignment values.
8353 (ix86_option_override_internal): Use processor_names.
8354 (ix86_function_specific_print): Likewise.
8355 * config/i386/i386.h (struct processor_costs):
8356 Add alignment values.
8357 (struct ptt): Remove and replace with const char *.
8358 * config/i386/x86-tune-costs.h (struct processor_costs):
8359 Declare default alignments for all costs.
8361 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
8363 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
8364 symbolics or VR_VARYING ranges for ABS_EXPR.
8365 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
8366 when range will wrap.
8368 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
8371 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
8372 before and after a JUMP_TABLE_DATA.
8374 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
8377 * dojump.c (do_compare_and_jump): Canonicalize function pointers
8378 when one operand is a function pointer. Use POINTER_TYPE_P and
8379 FUNC_OR_METHOD_TYPE_P.
8380 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
8381 * fold-const.c (build_range_check): Likewise.
8382 * match.pd (simple_comparison): Likewise.
8384 2018-09-14 David Malcolm <dmalcolm@redhat.com>
8387 * spellcheck.c (get_edit_distance_cutoff): New function.
8388 (selftest::test_edit_distance_unit_test_oneway): Rename to...
8389 (selftest::test_get_edit_distance_one_way): ...this.
8390 (selftest::test_get_edit_distance_unit): Rename to...
8391 (selftest::test_get_edit_distance_both_ways): ...this.
8392 (selftest::test_edit_distances): Move tests to this new function,
8393 and test some more pairs of strings. Update for above renaming.
8394 (selftest::get_old_cutoff): New function.
8395 (selftest::test_get_edit_distance_cutoff): New function.
8396 (selftest::assert_suggested_for): New function.
8397 (ASSERT_SUGGESTED_FOR): New macro.
8398 (selftest::assert_not_suggested_for): New function.
8399 (ASSERT_NOT_SUGGESTED_FOR): New macro.
8400 (selftest::test_suggestions): New function.
8401 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
8402 tests to selftest::test_edit_distances and call it. Add calls to
8403 selftest::test_get_edit_distance_cutoff and
8404 selftest::test_suggestions.
8405 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
8406 (best_match::consider): Replace hard-coded cutoff calculation with
8408 (best_match::get_cutoff): New declaration.
8409 (best_match::get_best_meaningful_candidate): Likewise.
8411 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
8413 * builtins.c (fold_builtin_strlen): Remove TODO comment.
8415 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
8418 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
8420 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
8421 terminated string literal.
8423 2018-09-14 Martin Sebor <msebor@redhat.com>
8425 * builtins.c (unterminated_array): Handle ARRAY_REF.
8426 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
8427 * builtins.h (unterminated_array): Declare extern.
8428 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
8430 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
8433 2018-09-14 Martin Sebor <msebor@redhat.com>
8434 Jeff Law <law@redhat.com>
8436 * builtins.c (unterminated_array): New.
8437 (expand_builtin_strcpy): Adjust.
8438 (expand_builtin_strcpy_args): Detect unterminated arrays.
8439 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
8440 unterminated arrays.
8441 * gimple-fold.h (get_maxval_strlen): Add argument.
8442 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
8444 * gimple-fold.c (get_range_strlen): Add argument.
8445 (get_maxval_strlen): Adjust.
8446 * gimple-fold.h (get_range_strlen): Add argument.
8448 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
8450 * config/i386/movdirintrin.h: Fix copyright year.
8452 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
8454 * reg-stack.c: Include regs.h.
8455 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
8456 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
8457 FIRST_STACK_REG, not DFmode.
8458 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
8459 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
8460 (change stack): Default register mode to the reg_raw_mode of
8461 FIRST_STACK_REG, not DFmode.
8462 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
8463 (*swapxf): Rename from swapxf.
8465 2018-09-14 Carl Love <cel@us.ibm.com>
8467 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
8468 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
8470 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
8473 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
8476 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
8479 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
8481 2018-09-14 Jason Merrill <jason@redhat.com>
8483 Fix --enable-gather-detailed-mem-stats.
8484 * hash-table.c (hash_table_usage): Change from variable to function.
8485 * hash-table.h: Adjust.
8486 * Makefile.in: Add missing dependencies on hash-table.h.
8488 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8490 PR tree-optimization/87259
8492 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
8493 execute_cse_reciprocals_1 has tried transforming.
8495 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
8497 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
8498 VR_VARYING for PLUS/MINUS_EXPR.
8500 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
8502 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
8505 2018-09-14 Richard Biener <rguenther@suse.de>
8508 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
8509 bits for the merged partition.
8511 2018-09-13 Martin Sebor <msebor@redhat.com>
8512 Bernd Edlinger <bernd.edlinger@hotmail.de>
8514 * builtins.h (c_srlen): Add argument.
8515 * builtins.c (warn_string_no_nul): New function.
8516 (c_strlen): Add argument and use it. Update recursive calls.
8517 Pass DECL argument to string_constant to get info on non
8518 terminated strings. Update *NONSTR as needed.
8519 (fold_builtin_strlen): Add argument to calls to c_strlen.
8520 Warn for unterminated arrays.
8521 (warn_string_no_null): Add prototype.
8522 * expr.c (string_constant): Update arguments. Update recursive
8523 calls appropriately. Detect missing NUL terminator and outermost
8524 declaration its missing in.
8525 Improve checks for arrays with nonzero lower bound or elements
8526 that are not a single byte. Simplify offset computation.
8527 Simplify checks for non-NUL terminated strings.
8528 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
8529 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
8531 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
8533 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
8535 * fold-const.c (c_getstr): Fix function comment. Remove unused third
8536 argument. Fix range checks.
8537 * fold-const.h (c_getstr): Adjust protoype.
8538 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
8539 string is constant but contains no NUL byte.
8541 * expr.c (string_constant): Adjust function comment.
8542 Remove bogus check for zero termination.
8544 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
8546 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
8547 (get_constant_size): Don't make STRING_CSTs larger than they are.
8548 (check_string_literal): New check function for STRING_CSTs.
8549 (output_constant): Use it.
8551 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
8554 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
8556 2018-09-13 Richard Biener <rguenther@suse.de>
8558 PR tree-optimization/87263
8559 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
8560 (struct unwind_state): Add max_rpo field.
8561 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
8562 Compute max_rpo, the max RPO number a block can be backwards reached
8563 from. Re-write non-iterating mode to a RPO ordered worklist approach,
8564 separating it from the iterating mode.
8566 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
8568 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
8569 (rfs_decision): New scheduling decision.
8571 2018-09-13 Richard Biener <rguenther@suse.de>
8574 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
8575 (vn_nary_op_insert_pieces_predicated): Do not write useless
8576 valid_dominated_by_p entry outside of the allocated storage.
8578 2018-09-13 Omar Sandoval <osandov@osandov.com>
8579 Tom de Vries <tdevries@suse.de>
8582 * dwarf2out.c (is_c): New function.
8583 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
8585 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
8588 * config/aarch64/aarch64.md (*aarch64_bfxil):
8590 * config/aarch64/constraints.md (Ulc): Define.
8591 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
8593 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
8596 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
8598 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
8599 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
8600 aarch64_layout_frame call.
8601 (aarch64_expand_epilogue): Likewise.
8602 (aarch64_initial_elimination_offset): Likewise.
8603 (aarch64_get_separate_components): Likewise.
8604 (aarch64_use_return_insn_p): Likewise.
8605 (aarch64_layout_frame): Remove unneeded check.
8607 2018-09-13 Jakub Jelinek <jakub@redhat.com>
8609 * configure.ac: Only append
8610 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
8611 gcc_config_arguments if it was never reconfigured or last reconfigure
8612 was with different arguments.
8613 * configure: Regenerated.
8615 2018-09-13 Jakub Jelinek <jakub@redhat.com>
8616 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8619 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
8620 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
8622 2018-09-13 Jakub Jelinek <jakub@redhat.com>
8624 PR tree-optimization/87287
8625 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
8626 X % C == 0 to X % (unsigned) C == 0 optimization to ...
8627 * match.pd (X % C == 0): ... here. New optimization.
8629 2018-09-12 Jakub Jelinek <jakub@redhat.com>
8632 * expr.h (maybe_optimize_mod_cmp): Declare.
8633 * expr.c (mod_inv): New function.
8634 (maybe_optimize_mod_cmp): New function.
8635 (do_store_flag): Use it.
8636 * cfgexpand.c (expand_gimple_cond): Likewise.
8638 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
8639 Julian Brown <julian@codesourcery.com>
8642 * gimplify.c (gimplify_scan_omp_clauses): Set
8643 target_firstprivatize_array_bases in OpenACC parallel and kernels
8644 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
8645 OpenACC data regions.
8647 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
8649 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
8650 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
8652 2018-09-12 Richard Biener <rguenther@suse.de>
8654 PR tree-optimization/87280
8655 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
8656 edge but unreachable target.
8657 (do_rpo_vn): For conservatively handling a PHI only mark
8658 the backedge executable but not the block reachable.
8660 2018-09-12 Richard Biener <rguenther@suse.de>
8662 PR tree-optimization/87266
8663 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
8666 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
8668 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
8670 ("trunc<BFP:mode><DFP_ALL:mode>2")
8671 ("trunc<DFP_ALL:mode><BFP:mode>2")
8672 ("extend<BFP:mode><DFP_ALL:mode>2")
8673 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
8674 according to the target operand type.
8676 2018-09-12 Jakub Jelinek <jakub@redhat.com>
8677 Andreas Krebbel <krebbel@linux.ibm.com>
8679 PR tree-optimization/86844
8680 * gimple-ssa-store-merging.c
8681 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
8682 there are any overlapping stores in between them, make sure they are
8683 also coalesced or we give up completely.
8685 2018-09-12 Jakub Jelinek <jakub@redhat.com>
8688 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
8689 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
8691 2018-09-12 Tom de Vries <tdevries@suse.de>
8693 * common.opt (gdescribe-dies): Add option.
8694 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
8695 attribute for artifical and nameless decls.
8696 (dwarf2out_register_external_die): Add description attribute to
8697 external reference die.
8698 (add_desc_attribute): New functions.
8699 (gen_subprogram_die): Add description attribute to
8700 DW_TAG_call_site_parameter.
8701 * tree-pretty-print.c (print_generic_expr_to_str): New function.
8702 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
8703 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
8705 (@item -gdescribe-dies): Add.
8707 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
8709 * tree-vrp.c (vrp_shift_undefined_p): Remove.
8710 (extract_range_from_binary_expr_1: Call
8711 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
8712 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
8715 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
8717 * gimple-ssa-warn-alloca.c
8718 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
8719 field for ALLOCA_BOUND_*_LARGE.
8721 2018-09-11 Nathan Sidwell <nathan@acm.org>
8723 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
8725 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
8727 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
8728 for clobbers. Remove obsolete comment.
8730 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
8732 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
8733 mpxchk, mpxld and mpxst types.
8734 (define_attr length_immediate): Remove all processing of mpx types.
8735 (define_attr prefix_0f): Ditto.
8736 (define_attr memory): Ditto.
8738 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
8740 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
8741 (log<mode>2): Change operand 1 predicate to general_operand.
8742 Extend operand 1 to XFmode and generate logxf3 insn.
8743 (log10<mode>2): Change operand 1 predicate to general_operand.
8744 Extend operand 1 to XFmode and generate log10xf3 insn.
8745 (log2<mode>2): Change operand 1 predicate to general_operand.
8746 Extend operand 1 to XFmode and generate log2xf3 insn.
8747 (fyl2xp1_extend<mode>xf3_i387): Remove.
8748 (log1p<mode>2): Change operand 1 predicate to general_operand.
8749 Extend operand 1 to XFmode and generate log1pxf3 insn.
8750 (fxtract_extend<mode>xf3_i387): Remove.
8751 (logb<mode>2): Change operand 1 predicate to general_operand.
8752 Extend operand 1 to XFmode and generate logbxf3 insn.
8753 (ilogb<mode>2): Change operand 1 predicate to general_operand.
8754 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
8755 (significand<mode>2): Change operand 1 predicate to general_operand.
8756 Extend operand 1 to XFmode and generate significandxf3 insn.
8758 2018-09-11 Nathan Sidwell <nathan@acm.org>
8760 * gcc.c (perror_with_name, pfatal_with_name): Delete.
8761 (load_specs): Use fatal_error.
8762 (DELETE_IF_ORDINARY, process_command): Use error.
8763 (execute, run_attempt): Use fatal_error.
8765 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
8767 * diagnostic-core.h (sorry_at): New prototype.
8768 * diagnostic.c (sorry_at): New function.
8770 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
8772 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
8773 by zero as VR_UNDEFINED.
8775 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
8777 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
8778 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
8779 (<sincos>mode2): New expander.
8780 (sincos_extend<mode>xf3_i387): Remove insn pattern.
8781 (sincos -> sin, cos splitters): Remove splitter patterns.
8782 (sincos<mode>3): Change operand 2 predicate to general_operand.
8783 Extend operand 2 to XFmode and generate sincosxf3 insn.
8784 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
8785 Change operand 3 predicate to const1_operand.
8786 (fptan_extend<mode>xf4_i387): Remove insn pattern.
8787 (tanxf2): Update operands in the call to fptanxf4_i387.
8788 (tan<mode>2): Change operand 1 predicate to general_operand.
8789 Extend operand 1 to XFmode and generate tanxf3 insn.
8790 (atan2xf3): Rename from *fpatanxf3_i387.
8791 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
8792 (atan2xf3): Remove expander.
8793 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
8794 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
8795 (atan<mode>2): Change operand 1 predicate to general_operand.
8796 Extend operand 1 to XFmode and generate atanxf3 insn.
8798 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
8800 * config/i386/i386.md (x87/SSE constant load splitter): Use
8801 memory_operand instead of nonimmediate_operand for input operand
8804 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
8806 * config/i386/i386.md (float partial SSE register stall splitter): Move
8807 splitter near its instruction pattern.
8808 (float_extend partial SSE register stall splitter): Ditto.
8809 (float_truncate partial SSE register stall splitter): Ditto.
8811 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
8814 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8815 to speculation_safe_value_not_needed.
8818 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
8819 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
8820 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
8821 leaf_function_p, instead use has_hard_reg_initial_val.
8823 2018-09-09 Nathan Sidwell <nathan@acm.org>
8825 * gcc.h (pfatal_with_name): Don't declare here.
8826 * gcc.c (pfatal_with_name): Make static.
8828 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
8830 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
8833 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
8835 PR rtl-optimization/85458
8836 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
8837 priority hook to reduce the priority of EXPR.
8839 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
8841 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
8842 DImode for x87 on 32bit targets. Conditionally disable x87 modes
8843 with X87_ENABLE_FLOAT. Remove preparation code.
8844 (*float<SWI48:mode><MODEF:mode>2): Rename from
8845 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
8846 math using "enabled" attribute.
8847 (*floatdi<MODEF:mode>2_i387): Rename from
8848 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
8849 enable for 32bit targets only.
8850 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
8852 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
8853 as operand 1 predicate. Rewrite as define_insn_and_split.
8854 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
8856 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
8858 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
8859 to fallthru to FLOAT case.
8861 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
8864 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
8865 around folding of vec_sl to handle out of range shift values.
8867 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
8869 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
8870 Update callers to gen_fix_trunc<mode>_i387_fisttp
8871 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
8872 nonimmediate_operand.
8873 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
8874 and corresponding splitters.
8875 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
8876 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
8877 (fix_truncdi_i387_with_temp): Remove insn pattern
8878 and corresponding splitters.
8879 (fix_trunc<mode>_i387): Change operand 0 predicate to
8880 nonimmediate_operand.
8881 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
8882 and corresponding splitters.
8883 (*fistdi2_1): Remove.
8885 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
8886 (lrintxfdi2): Remove expander. Reimplement as define_insn.
8887 (*fist<mode>2_1): Remove.
8888 (fist<mode>2): Ditto.
8889 (fist<mode>2_with_temp): Remove insn pattern and corresponding
8891 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
8892 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
8893 (fistdi2_<rounding>): Change operand 0 predicate to
8894 nonimmediate_operand.
8895 (fistdi2_<rounding>_with_temp): Remove insn pattern
8896 and corresponding splitters.
8897 (fist<mode>2_<rounding>): Change operand 0 predicate to
8898 nonimmediate_operand.
8899 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
8900 and corresponding splitters.
8902 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
8904 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
8906 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
8909 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
8911 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
8912 early gimple folding of vec_splat().
8913 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
8914 * gimple-fold.h: Add an extern define for tree_vec_extract().
8916 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
8918 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
8919 wrappers around TREE_TYPE comparisons.
8921 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
8924 * config/s390/predicates.md: Add nonsym_memory_operand.
8925 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
8926 contains a SYMBOL_REF, load it into an intermediate pseudo.
8927 (s390_emit_compare_and_swap): Legitimize operand.
8928 * config/s390/s390.md: Use the new nonsym_memory_operand
8929 with UNSPECV_CAS patterns.
8931 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
8934 * config/s390/s390-passes.def: New file.
8935 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
8937 (make_pass_s390_early_mach): Add declaration.
8938 * config/s390/s390.c (make_pass_s390_early_mach):
8939 (s390_option_override): Remove dynamic registration.
8940 * config/s390/t-s390: Add s390-passes.def.
8942 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
8944 * config/s390/s390.c (s390_decompose_constant_pool_ref):
8945 Remove UNSPEC_LTREL_BASE check.
8946 (annotate_constant_pool_refs): Likewise.
8947 (find_constant_pool_ref): Likewise.
8948 (find_ltrel_base): Removed.
8949 (replace_ltrel_base): Removed.
8950 (s390_mainpool_finish): Remove replace_ltrel_base call.
8951 (s390_chunkify_start): Remove pending LTREL_BASE logic.
8952 (s390_chunkify_finish): Remove replace_ltrel_base call.
8953 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
8955 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
8958 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
8959 to speculation_safe_value_not_needed.
8961 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
8962 Bernd Schmidt <bernds_cb1@t-online.de>
8964 * config/nvptx/nvptx-opts.h: New file.
8965 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
8966 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
8968 (TARGET_SM35): New macro.
8969 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
8971 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
8973 (misa=): New option.
8974 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
8976 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
8978 * config/i386/i386.md (truncdfsf2): Remove expander.
8979 (truncdfsf2_with_temp): Ditto.
8980 (truncxf<mode>2): Ditto.
8981 (*truncdfsf_fast_mixed): Remove insn pattern.
8982 (*truncdfsf_fast_i387): Ditto.
8983 (*truncdfsf_mixed): Ditto.
8984 (*truncdfsf_i387): Ditto.
8985 (*truncdfsf2_i387_1): Ditto.
8986 (*truncxfsf2_mixed): Ditto.
8987 (*truncxfdf2_mixed): Ditto.
8988 (*truncxf<mode>2_i387_noop): Ditto. Update callers
8989 to call gen_truncxf<mode>2 instead.
8990 (*truncxf<mode>2_i387): Remove.
8991 (reg->reg splitters): Remove splitter pattern.
8992 (reg->mem splitters): Ditto.
8994 (truncdfsf2): New insn pattern.
8995 (truncxf<mode>2): Ditto.
8997 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8999 * tree-ssa-math-opts.c (is_mult_by): New function.
9000 (is_square_of): Use the above.
9001 (optimize_recip_sqrt): New function.
9002 (pass_cse_reciprocals::execute): Use the above.
9004 2018-09-05 Richard Biener <rguenther@suse.de>
9007 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
9008 to zero-init the emplaced vec.
9010 2018-09-05 Martin Liska <mliska@suse.cz>
9012 PR tree-optimization/87205
9013 * tree-switch-conversion.c (pass_lower_switch::execute):
9014 Group cases for switch statements.
9016 2018-09-05 Richard Biener <rguenther@suse.de>
9018 PR tree-optimization/87217
9019 * tree-ssa-sccvn.c (vuse_valueize): New.
9020 (vn_reference_lookup_pieces): Use it.
9021 (vn_reference_lookup): Likewise.
9023 2018-09-05 Nathan Sidwell <nathan@acm.org>
9026 * stor-layout.c (place_field): Scan forwards to check last
9027 bitfield when ms_bitfield_placement is in effect.
9029 2018-09-05 Richard Biener <rguenther@suse.de>
9032 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
9035 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
9036 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
9038 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
9039 * config.gcc (extra_objs): Build it.
9040 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
9042 * config/aarch64/aarch64-passes.def
9043 (pass_tag_collision_avoidance): New pass.
9044 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
9045 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
9046 (aarch64_classify_address): Remove static qualifier.
9047 (aarch64_address_info, aarch64_address_type): Move to...
9048 * config/aarch64/aarch64-protos.h: ... here.
9049 (make_pass_tag_collision_avoidance): New function.
9050 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
9053 2018-09-05 Martin Liska <mliska@suse.cz>
9055 * doc/gcov.texi: Update documentation of humar
9057 * gcov.c (format_count): Print one decimal place, it provides
9058 more fine number of situations like '1G' vs. '1.4G'.
9060 2018-09-05 Martin Liska <mliska@suse.cz>
9063 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
9064 * optc-gen.awk: Allow 'Var' for Deprecated options in order
9065 to generate a MASK value.
9067 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
9070 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
9071 if frame pointer isn't used.
9072 (compute_frame_pointer_to_fb_displacement): Likewise.
9074 2018-09-04 Jakub Jelinek <jakub@redhat.com>
9077 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
9078 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
9079 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
9080 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
9081 and OPTION_MASK_ISA_XSAVEC_UNSET.
9083 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
9085 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
9086 XOR operations in NAND case.
9088 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
9090 * wide-int-range.cc (wide_int_range_convert): New.
9091 * wide-int-range.h (wide_int_range_convert): New.
9092 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
9093 code into wide_int_range_convert.
9094 (extract_range_into_wide_ints): Do not munge anti range constants
9095 into the entire domain. Just return the range back.
9097 2018-09-04 Martin Liska <mliska@suse.cz>
9099 * genmatch.c (output_line_directive): Add new argument
9101 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
9103 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
9105 * doc/invoke.texi (Option Summary): Add whitespace.
9107 * doc/invoke.texi (Option Summary): Add -Waligned-new.
9109 2018-09-04 Richard Biener <rguenther@suse.de>
9111 PR tree-optimization/87211
9112 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
9113 backedge value we're supposed to treat as VARYING also number
9114 the PHI to VARYING in case it got a different value-number already.
9116 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
9118 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
9119 (extract_range_from_binary_expr_1): Do not call
9120 vrp_can_optimize_bit_op.
9121 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
9123 (wide_int_range_get_mask_and_bounds): New.
9124 (wide_int_range_optimize_bit_op): New.
9125 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
9126 (wide_int_range_bit_and): Same.
9127 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
9128 (wide_int_range_optimize_bit_op): New.
9129 (wide_int_range_get_mask_and_bounds): New.
9131 2018-09-04 Richard Biener <rguenther@suse.de>
9133 PR tree-optimization/87176
9134 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
9135 variable. When value-numbering a virtual PHI node make sure
9136 to not value-number to the backedge value.
9138 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
9140 * doc/extend.texi (Long Long, Hex Floats): Document support for
9141 long long and hex floats in more recent versions of ISO C++.
9143 2018-09-03 Richard Biener <rguenther@suse.de>
9145 PR tree-optimization/87177
9146 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
9149 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
9151 * bb-reorder.c (edge_order): Convert to C-qsort-style
9152 tri-state comparator.
9153 (reorder_basic_blocks_simple): Change std::stable_sort to
9156 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
9158 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
9159 tri-state comparator.
9160 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
9162 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
9164 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
9165 (mergesort): ... here as maximum count for using netsort.
9166 (gcc_qsort): Set nlim to 3 if stable sort is requested.
9167 (gcc_stablesort): New.
9168 * system.h (gcc_stablesort): Declare.
9170 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
9172 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
9173 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
9174 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
9176 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
9178 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
9179 lxsdx and stxsdx alternatives.
9180 (*mov<mode>_hardfloat64): Ditto.
9181 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
9183 2018-09-03 Richard Biener <rguenther@suse.de>
9185 PR tree-optimization/87200
9186 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
9189 2018-09-03 Martin Liska <mliska@suse.cz>
9191 PR tree-optimization/87201
9192 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
9193 Fix parenthesis in an expression.
9195 2018-09-03 Richard Biener <rguenther@suse.de>
9197 PR tree-optimization/87197
9198 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
9199 visited. CSE the VN_INFO hashtable lookup.
9201 PR tree-optimization/87169
9202 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
9203 iterating make sure there's no extra backedges from irreducible
9204 regions feeding the header. Mark the destination block
9207 2018-09-03 Martin Liska <mliska@suse.cz>
9210 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
9211 * common/common-targhooks.c (default_get_valid_option_values):
9213 * common/common-targhooks.h (default_get_valid_option_values):
9215 * common/config/i386/i386-common.c: Move processor_target_table
9217 (ix86_get_valid_option_values): New function.
9218 (TARGET_GET_VALID_OPTION_VALUES): New macro.
9219 * config/i386/i386.c (struct ptt): Move to i386-common.c.
9220 (PTA_*): Move all defined masks into i386-common.c.
9221 (ix86_function_specific_restore): Use new processor_cost_table.
9222 * config/i386/i386.h (struct ptt): Moved from i386.c.
9223 (struct pta): Likewise.
9224 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
9225 * doc/tm.texi.in: Likewise.
9226 * opt-suggestions.c (option_proposer::suggest_option):
9227 Pass prefix to build_option_suggestions.
9228 (option_proposer::get_completions): Likewise.
9229 (option_proposer::build_option_suggestions): Use the new target
9231 * opts.c (struct option_help_tuple): New struct.
9232 (print_filtered_help): Use the new target hook.
9234 2018-09-03 Martin Liska <mliska@suse.cz>
9237 * predict.c (set_even_probabilities): Add likely_edges
9238 argument and handle cases where we have precisely one
9240 (combine_predictions_for_bb): Catch also likely_edges.
9241 (tree_predict_by_opcode): Handle gswitch statements.
9242 * tree-cfg.h (find_case_label_for_value): New declaration.
9243 (find_taken_edge_switch_expr): Likewise.
9244 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
9245 Find pivot in decision tree based on probabily, not by number of
9248 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
9250 * doc/standards.texi (Standards): Update Objective-C reference.
9252 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
9254 * doc/install.texi (Prerequisites): Update link for MPC.
9256 2018-09-01 Michael Matz <matz@suse.de>
9258 PR tree-optimization/87074
9259 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
9260 PHIs for outer-loop uses.
9262 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
9264 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
9265 * doc/invoke.texi (C Dialect Options): Ditto.
9267 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
9269 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
9271 2018-08-31 Richard Biener <rguenther@suse.de>
9273 PR tree-optimization/87168
9274 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
9275 (rpo_elim::eliminate_avail): When OP was not visited it must
9278 2018-08-31 David Malcolm <dmalcolm@redhat.com>
9280 * tree-vrp.c (copy_value_range): Convert param "from" from
9281 "value_range *" to "const value_range *".
9282 (range_is_null): Likewise for param "vr".
9283 (range_int_cst_p): Likewise.
9284 (range_int_cst_singleton_p): Likewise.
9285 (symbolic_range_p): Likewise.
9286 (value_ranges_intersect_p): Likewise for both params.
9287 (value_range_nonnegative_p): Likewise for param "vr".
9288 (value_range_constant_singleton): Likewise.
9289 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
9290 (extract_range_into_wide_ints): Likewise for param "vr".
9291 (extract_range_from_multiplicative_op): Likewise for params "vr0"
9293 (vrp_can_optimize_bit_op): Likewise.
9294 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
9296 (extract_range_from_unary_expr): Likewise.
9297 (debug_value_range): Likewise for param "vr".
9298 (value_range::dump): Add "const" qualifier.
9299 (vrp_prop::check_array_ref): Convert local "vr" from
9300 "value_range *" to "const value_range *".
9301 (vrp_prop::check_mem_ref): Likewise.
9302 (vrp_prop::visit_stmt): Likewise for local "old_vr".
9303 (vrp_intersect_ranges_1): Likewise for param "vr_1".
9304 (vrp_intersect_ranges): Likewise.
9305 (simplify_stmt_for_jump_threading): Likewise for local "vr".
9306 (vrp_prop::vrp_finalize): Likewise.
9307 * tree-vrp.h (value_range::dump): Add "const" qualifier.
9308 (vrp_intersect_ranges): Add "const" qualifier to params as above.
9309 (extract_range_from_unary_expr): Likewise.
9310 (value_range_constant_singleton): Likewise.
9311 (symbolic_range_p): Likewise.
9312 (copy_value_range): Likewise.
9313 (extract_range_from_binary_expr_1): Likewise.
9314 (range_int_cst_p): Likewise.
9315 (vrp_set_zero_nonzero_bits): Likewise.
9316 (range_int_cst_singleton_p): Likewise.
9318 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
9320 * config/aarch64/arm_neon.h (vabsd_s64): New.
9321 (vnegd_s64): Likewise.
9323 2018-08-31 Martin Jambor <mjambor@suse.cz>
9325 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
9327 2018-08-31 Martin Liska <mliska@suse.cz>
9329 * ipa-icf.c (sem_item::add_type): Use
9330 sem_item::m_type_hash_cache.
9331 * ipa-icf.h: Move the cache from sem_item_optimizer
9334 2018-08-31 Nathan Sidwell <nathan@acm.org>
9336 * doc/extend.texi (Backwards Compatibility): Remove implicit
9337 extern C leeway of () being (...).
9339 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9341 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
9343 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
9347 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
9349 2018-08-31 Jakub Jelinek <jakub@redhat.com>
9352 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
9353 gen_int_mode. Formatting fixes.
9355 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
9357 * target.def (custom_function_descriptors): Improve documentation.
9358 * doc/tm.texi.in (Trampolines): Expand discussion of function
9359 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
9360 beginning of the section.
9361 * doc/tm.texi: Regenerated.
9363 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
9365 * cfg.h (class auto_edge_flag): Spell out the template-id of the
9366 base class in the initializer list. This is a workaround for
9367 building with older GCC.
9368 (class auto_bb_flag): Likewise.
9370 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
9372 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
9373 (altivec_vcmpequ<VI_char>_p): Remove star.
9374 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
9376 (expand_strncmp_vec_sequence): New function.
9377 (emit_final_str_compare_vec): New function.
9378 (expand_strn_compare): Add support for vector strncmp.
9379 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
9380 length specification to bytes.
9381 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
9382 (vcmpnezb_p): New pattern.
9383 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
9384 for option -mstring-compare-inline-limit.
9386 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
9388 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
9389 (PTA_SKYLAKE): Add PTA_AES.
9390 (PTA_GOLDMONT): Likewise.
9392 2018-08-29 Jan Hubicka <jh@suse.cz>
9395 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
9396 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
9398 2018-08-29 Jan Hubicka <jh@suse.cz>
9400 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
9402 (hash_tree): Do not visit TYPE_STUB_DECL.
9403 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
9404 stream TYPE_STUB_DECL.
9405 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
9406 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
9407 after free_lang_data.
9408 (type_in_anonymous_namespace_p): Likewise.
9410 2018-08-29 Jan Hubicka <jh@suse.cz>
9412 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
9413 comment that it has to be even number.
9414 (class sreal): Change m_sig type to int32_t.
9415 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
9416 int64_t for temporary calculations.
9417 (sreal_verify_basics): Drop one bit from minimum and maximum.
9419 2018-08-30 Richard Biener <rguenther@suse.de>
9421 PR tree-optimization/87147
9422 * tree-ssa-sccvn.c (SSA_VISITED): New function.
9423 (visit_phi): When the degenerate result is from the backedge and
9424 we didn't visit its definition yet drop to VARYING.
9425 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
9427 2018-08-29 Jan Hubicka <jh@suse.cz>
9429 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
9431 (hash_tree): Likewise.
9433 2018-08-29 Jan Hubicka <jh@suse.cz>
9435 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
9436 and TYPE_NEXT_REF_TO.
9438 2018-08-29 Jan Hubicka <jh@suse.cz>
9440 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
9441 comment that it has to be even number.
9442 (class sreal): Change m_sig type to int32_t.
9443 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
9444 int64_t for temporary calculations.
9445 (sreal_verify_basics): Drop one bit from minimum and maximum.
9447 2018-08-30 Tamar Christina <tamar.christina@arm.com>
9449 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
9451 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
9454 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
9455 if to_add is negative.
9457 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
9460 * builtins.c (c_strlen): Improve range checks.
9462 2018-08-29 Martin Sebor <msebor@redhat.com>
9463 Jeff Law <law@redhat.com>
9465 PR tree-optimization/86714
9466 PR tree-optimization/86711
9467 * builtins.c (c_strlen): Add arguments to call to string_constant.
9468 * expr.c (string_constant): Add argument. Detect missing nul
9469 terminator and outermost declaration it's missing in.
9470 * expr.h (string_constant): Add argument.
9471 * fold-const.c (read_from_constant_string): Add arguments to call to
9473 (c_getstr): Likewise.
9474 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
9476 * tree-ssa-strlen.c (get_stridx): Likewise.
9478 2018-08-29 Jan Hubicka <jh@suse.cz>
9480 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
9481 Do not stream DECL_VINDEX.
9482 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
9483 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
9484 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
9486 2018-08-29 Richard Biener <rguenther@suse.de>
9488 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
9489 virtual operands that are not default defs to honor region
9491 (rpo_vn_valueize): Remove ineffective code here.
9493 2018-08-29 Richard Biener <rguenther@suse.de>
9495 PR tree-optimization/87132
9496 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
9497 when skipping defs reachable over backedges.
9499 2018-08-29 Richard Biener <rguenther@suse.de>
9501 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
9502 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
9503 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
9504 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
9505 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
9506 (vn_reference_lookup_3): Remove use of const_parms.
9507 (free_rpo_vn): Do not free const_parms.
9508 (do_rpo_vn): Do not call init_const_parms.
9509 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
9510 SSA_NAME_POINTS_TO_READONLY_MEMORY.
9511 (call_may_clobber_ref_p_1): Likewise.
9513 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
9516 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
9518 (-ftree-scev-cprop): Document.
9520 2018-08-29 Jan Hubicka <jh@suse.cz>
9522 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
9524 (sreal constructor): Update.
9525 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
9526 sreal:operator/): Update.
9528 2018-08-29 Martin Liska <mliska@suse.cz>
9530 * tree-switch-conversion.c (switch_conversion::expand):
9531 Strenghten assumption about gswitch statements.
9533 2018-08-29 Richard Biener <rguenther@suse.de>
9535 PR tree-optimization/87117
9536 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
9537 re-value-number released SSA VDEFs.
9539 2018-08-29 Richard Biener <rguenther@suse.de>
9541 PR tree-optimization/87126
9542 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
9544 2018-08-28 Jim Wilson <jimw@sifive.com>
9546 * config/riscv/pic.md: Rewrite.
9547 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
9549 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
9550 (SOFTF, default_load, softload, softstore): New.
9552 2018-08-28 Jeff Law <law@redhat.com>
9554 * fold-const.c (fold_binary_loc): Remove recently added assert.
9556 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
9558 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
9559 to OP parmeter of generated function.
9561 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
9563 PR tree-optimization/87009
9564 * match.pd: Add boolean optimizations.
9566 2018-08-28 Martin Sebor <msebor@redhat.com>
9569 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
9570 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
9571 (pass_walloca::gate): Use it.
9572 (alloca_call_type): Same.
9573 (pass_walloca::execute): Same.
9574 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
9576 2018-08-28 David Malcolm <dmalcolm@redhat.com>
9578 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
9579 GCC_VERSION for usage of "__gcc_dump_printf__" format from
9582 2018-08-28 Richard Biener <rguenther@suse.de>
9584 PR tree-optimization/87124
9585 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
9586 constants before looking up avail.
9588 2018-08-28 Jakub Jelinek <jakub@redhat.com>
9591 * calls.c (maybe_warn_nonstring_arg): Punt early if
9592 warn_stringop_overflow is zero. Don't call get_range_strlen
9593 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
9594 Swap comparison operands to have constants on rhs. Only use
9595 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
9596 increment lenrng[0].
9598 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
9600 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
9601 use of tree_to_shwi. Remove duplicated test for the size being
9602 a whole number of bytes.
9604 2018-08-28 Richard Biener <rguenther@suse.de>
9606 PR tree-optimization/87117
9607 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
9608 Handle removed stmt without LHS (GIMPLE_NOP).
9610 2018-08-28 Richard Biener <rguenther@suse.de>
9612 PR tree-optimization/87117
9613 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
9614 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
9616 2018-08-28 Richard Biener <rguenther@suse.de>
9618 PR tree-optimization/87117
9619 * tree-ssa-pre.c (compute_avail): Do not make expressions
9620 with predicated values available.
9621 (get_expr_value_id): Assert we do not run into predicated value
9624 2018-08-28 Richard Biener <rguenther@suse.de>
9626 PR tree-optimization/87117
9627 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
9628 get virtual operands.
9629 (get_expr_operands): Handle STRING_CST like other decls.
9631 2018-08-28 Martin Liska <mliska@suse.cz>
9633 * tree.h: Update documentation of fndecl_built_in_p
9637 2018-08-27 Jeff Law <law@redhat.com>
9638 PR tree-optimization/87110
9639 * tree-ssa-dse.c (compute_trims): Handle non-constant
9642 2018-08-27 Martin Sebor <msebor@redhat.com>
9644 PR tree-optimization/86914
9645 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
9647 2018-08-27 Martin Sebor <msebor@redhat.com>
9649 PR tree-optimization/87112
9650 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
9651 the type of the bound argument.
9653 2018-08-27 Jeff Law <law@redhat.com>
9655 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
9656 type does not have a TYPE_SIZE_UNIT.
9658 2018-08-27 Steve Ellcey <sellcey@cavium.com>
9660 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
9661 with include of backend.h.
9663 2018-08-27 Richard Biener <rguenther@suse.de>
9665 PR tree-optimization/86927
9666 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
9667 use const cond reduction code.
9669 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
9671 PR tree-optimization/85758
9672 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
9674 2018-08-27 David Malcolm <dmalcolm@redhat.com>
9677 * diagnostic-show-locus.c (class layout_range): Update for
9678 conversion of show_caret_p to a tri-state.
9679 (layout_range::layout_range): Likewise.
9680 (make_range): Likewise.
9681 (layout::maybe_add_location_range): Likewise.
9682 (layout::should_print_annotation_line_p): Don't show annotation
9683 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
9684 (layout::get_state_at_point): Update for conversion of
9685 show_caret_p to a tri-state. Bail out early for
9686 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
9687 underlining or source colorization.
9688 (gcc_rich_location::add_location_if_nearby): Update for conversion
9689 of show_caret_p to a tri-state.
9690 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
9691 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
9693 (selftest::test_one_liner_labels): Likewise.
9694 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
9695 conversion of show_caret_p to a tri-state.
9696 * pretty-print.c (text_info::set_location): Likewise.
9697 * pretty-print.h (text_info::set_location): Likewise.
9698 * substring-locations.c (format_warning_n_va): Likewise.
9699 * tree-diagnostic.c (default_tree_printer): Likewise.
9700 * tree-pretty-print.c (newline_and_indent): Likewise.
9702 2018-08-27 David Malcolm <dmalcolm@redhat.com>
9705 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
9706 line above for line-insertion fix-it hints.
9707 (selftest::test_fixit_insert_containing_newline): Update the
9708 expected results, and add a test with line-numbering enabled.
9710 2018-08-27 Martin Liska <mliska@suse.cz>
9713 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
9714 params with DECL_HAS_VALUE_EXPR_P.
9716 2018-08-27 Martin Liska <mliska@suse.cz>
9718 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
9719 selected expansion strategy.
9721 2018-08-27 Martin Liska <mliska@suse.cz>
9723 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
9724 * builtins.c (is_builtin_fn): Likewise.
9725 * attribs.c (diag_attr_exclusions): Use new function
9726 fndecl_built_in_p and remove check for FUNCTION_DECL if
9728 (builtin_mathfn_code): Likewise.
9729 (fold_builtin_expect): Likewise.
9730 (fold_call_expr): Likewise.
9731 (fold_builtin_call_array): Likewise.
9732 (fold_call_stmt): Likewise.
9733 (set_builtin_user_assembler_name): Likewise.
9734 (is_simple_builtin): Likewise.
9735 * calls.c (gimple_alloca_call_p): Likewise.
9736 (maybe_warn_nonstring_arg): Likewise.
9737 * cfgexpand.c (expand_call_stmt): Likewise.
9738 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
9739 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
9740 (cgraph_node::verify_node): Likewise.
9741 * cgraphclones.c (build_function_decl_skip_args): Likewise.
9742 (cgraph_node::create_clone): Likewise.
9743 * config/arm/arm.c (arm_insert_attributes): Likewise.
9744 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
9745 * dse.c (scan_insn): Likewise.
9746 * expr.c (expand_expr_real_1): Likewise.
9747 * fold-const.c (operand_equal_p): Likewise.
9748 (fold_binary_loc): Likewise.
9749 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
9750 * gimple-low.c (lower_stmt): Likewise.
9751 * gimple-pretty-print.c (dump_gimple_call): Likewise.
9752 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9754 * gimple.c (gimple_build_call_from_tree): Likewise.
9755 (gimple_call_builtin_p): Likewise.
9756 (gimple_call_combined_fn): Likewise.
9757 * gimplify.c (gimplify_call_expr): Likewise.
9758 (gimple_boolify): Likewise.
9759 (gimplify_modify_expr): Likewise.
9760 (gimplify_addr_expr): Likewise.
9761 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
9762 * ipa-cp.c (determine_versionability): Likewise.
9763 * ipa-fnsummary.c (compute_fn_summary): Likewise.
9764 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
9765 * ipa-split.c (visit_bb): Likewise.
9766 (split_function): Likewise.
9767 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
9768 * lto-cgraph.c (input_node): Likewise.
9769 * lto-streamer-out.c (write_symbol): Likewise.
9770 * omp-low.c (setjmp_or_longjmp_p): Likewise.
9771 (lower_omp_1): Likewise.
9772 * predict.c (strip_predict_hints): Likewise.
9773 * print-tree.c (print_node): Likewise.
9774 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
9775 * trans-mem.c (is_tm_irrevocable): Likewise.
9776 (is_tm_load): Likewise.
9777 (is_tm_simple_load): Likewise.
9778 (is_tm_store): Likewise.
9779 (is_tm_simple_store): Likewise.
9780 (is_tm_abort): Likewise.
9781 (tm_region_init_1): Likewise.
9782 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
9783 * tree-cfg.c (verify_gimple_call): Likewise.
9784 (move_stmt_r): Likewise.
9785 (stmt_can_terminate_bb_p): Likewise.
9786 * tree-eh.c (lower_eh_constructs_2): Likewise.
9787 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
9788 * tree-inline.c (remap_gimple_stmt): Likewise.
9789 (copy_bb): Likewise.
9790 (estimate_num_insns): Likewise.
9791 (fold_marked_statements): Likewise.
9792 * tree-sra.c (scan_function): Likewise.
9793 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
9794 (optimize_stack_restore): Likewise.
9795 (pass_fold_builtins::execute): Likewise.
9796 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
9797 (mark_all_reaching_defs_necessary_1): Likewise.
9798 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
9799 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
9800 (pass_forwprop::execute): Likewise.
9801 * tree-ssa-loop-im.c (stmt_cost): Likewise.
9802 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
9803 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
9804 * tree-ssa-strlen.c (get_string_length): Likewise.
9805 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
9806 (find_func_aliases_for_call): Likewise.
9807 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
9808 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
9809 * tree-tailcall.c (find_tail_calls): Likewise.
9810 * tree.c (need_assembler_name_p): Likewise.
9811 (free_lang_data_in_decl): Likewise.
9812 (get_call_combined_fn): Likewise.
9813 * ubsan.c (is_ubsan_builtin_p): Likewise.
9814 * varasm.c (incorporeal_function_p): Likewise.
9815 * tree.h (DECL_BUILT_IN): Remove and replace with
9817 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
9818 (fndecl_built_in_p): New.
9820 2018-08-27 Martin Liska <mliska@suse.cz>
9822 PR tree-optimization/86847
9823 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
9824 Dump also subtree probability.
9825 (switch_decision_tree::do_jump_if_equal): New function.
9826 (switch_decision_tree::emit_case_nodes): Handle special
9827 situations in balanced tree that can be emitted much simpler.
9828 Fix calculation of probabilities that happen in tree expansion.
9829 * tree-switch-conversion.h (struct cluster): Add
9831 (struct simple_cluster): Likewise.
9832 (struct case_tree_node): Add new function has_child.
9833 (do_jump_if_equal): New.
9835 2018-08-27 Martin Liska <mliska@suse.cz>
9837 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
9838 Add new argument to bit_test_cluster constructor.
9839 (bit_test_cluster::emit): Set bits really number of values
9841 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
9842 probability argument.
9843 * tree-switch-conversion.h (struct bit_test_cluster):
9844 Add m_handles_entire_switch.
9846 2018-08-27 Martin Liska <mliska@suse.cz>
9848 PR tree-optimization/86702
9849 * tree-switch-conversion.c (jump_table_cluster::emit):
9850 Make probabilities even for values in jump table
9851 according to number of cases handled.
9852 (switch_decision_tree::compute_cases_per_edge): Pass
9853 argument to reset_out_edges_aux function.
9854 (switch_decision_tree::analyze_switch_statement): Likewise.
9855 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
9858 2018-08-27 Martin Liska <mliska@suse.cz>
9860 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
9861 cfun argument explicitly.
9862 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
9863 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
9864 function gimple_switch_default_bb.
9865 (convert_switch_statements):
9867 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
9868 * stmt.c (label_to_block_fn): Use label_to_block and pass
9869 cfun argument explicitly and use gimple_switch_label_bb.
9870 (expand_case): Likewise.
9871 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
9872 cfun argument explicitly. Likewise.
9873 (make_edges_bb): Likewise.
9874 (make_cond_expr_edges): Likewise.
9875 (get_cases_for_edge): Likewise.
9876 (make_gimple_switch_edges): Likewise.
9877 (label_to_block_fn): Likewise.
9878 (label_to_block): Likewise.
9879 (make_goto_expr_edges): Likewise.
9880 (make_gimple_asm_edges): Likewise.
9881 (main_block_label): Likewise.
9882 (group_case_labels_stmt): Likewise.
9883 (find_taken_edge_computed_goto): Likewise.
9884 (find_taken_edge_switch_expr): Likewise.
9885 (gimple_verify_flow_info): Likewise.
9886 (gimple_redirect_edge_and_branch): Likewise.
9887 (gimple_switch_label_bb): New function.
9888 (gimple_switch_default_bb): Likewise.
9889 (gimple_switch_edge): Likewise.
9890 (gimple_switch_default_edge): Likewise.
9891 * tree-cfg.h (label_to_block_fn): Remove and replace ...
9892 (label_to_block): ... with this.
9893 (gimple_switch_label_bb): New.
9894 (gimple_switch_default_bb): Likewise.
9895 (gimple_switch_edge): Likewise.
9896 (gimple_switch_default_edge): Likewise.
9897 * tree-cfgcleanup.c (convert_single_case_switch): Use
9898 new gimple functions and pass new argument to label_to_block.
9899 (cleanup_control_flow_bb):
9900 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
9901 cfun argument explicitly.
9902 (make_eh_edges): Likewise.
9903 (redirect_eh_dispatch_edge): Likewise.
9904 (lower_resx): Likewise.
9905 (lower_eh_dispatch): Likewise.
9906 (maybe_remove_unreachable_handlers): Likewise.
9907 (unsplit_eh): Likewise.
9908 (cleanup_empty_eh): Likewise.
9909 (verify_eh_edges): Likewise.
9910 (verify_eh_dispatch_edge): Likewise.
9911 * tree-ssa-dom.c (record_edge_info): Likewise.
9912 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
9913 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
9914 (thread_through_normal_block): Likewise.
9915 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
9916 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
9917 * tree-switch-conversion.c (switch_conversion::collect): Use new
9919 (switch_conversion::check_final_bb): Likewise.
9920 (switch_conversion::gather_default_values): Pass new argument
9922 (switch_conversion::build_constructors): Likewise.
9923 (switch_decision_tree::compute_cases_per_edge): Use new
9924 gimple_switch_edge function.
9925 (switch_decision_tree::analyze_switch_statement): Pass new argument
9927 (switch_decision_tree::try_switch_expansion): Use
9928 gimple_switch_default_edge.
9929 * tree-vrp.c (find_switch_asserts): Pass new argument
9931 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
9932 (vr_values::simplify_switch_using_ranges): Likewise.
9934 2018-08-27 Richard Biener <rguenther@suse.de>
9936 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
9937 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
9939 * tree-ssa-sccvn.h (struct vn_pval): New structure.
9940 (struct vn_nary_op_s): Add unwind_to member. Add
9941 predicated_values flag and put result into a union together
9942 with a linked list of vn_pval.
9943 (struct vn_ssa_aux): Add name member to make maintaining
9944 a map of SSA name to vn_ssa_aux possible. Remove no longer
9945 needed info, dfsnum, low, visited, on_sccstack, use_processed
9946 and range_info_anti_range_p members.
9947 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
9948 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
9950 (vn_valueize): New global.
9951 (vn_context_bb): Likewise.
9952 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
9953 VN_INFO_PTR_INFO): Remove.
9954 * tree-ssa-sccvn.c: ... (rewrite)
9955 (pass_fre::execute): For -O2+ initialize loops and run
9956 RPO VN in optimistic mode (iterating). For -O1 and -Og
9957 run RPO VN in non-optimistic mode.
9958 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
9959 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
9960 * doc/invoke.texi (sccvn-max-scc-size): Remove.
9961 (rpo-vn-max-loop-depth): Document.
9962 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
9963 when valuezing the VUSE signals we walked out of the region.
9964 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
9965 (phi_translate): Set VN context block to use for availability
9967 (compute_avail): Likewise.
9968 (pre_valueize): New function.
9969 (pass_pre::execute): Adjust to the RPO VN API.
9971 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
9972 (propagate_constants_for_unrolling): Remove.
9973 (tree_unroll_loops_completely): Perform value-numbering
9974 on the unrolled bodies loop parent.
9976 2018-08-27 Richard Biener <rguenther@suse.de>
9978 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
9979 for partial antic compute.
9981 2018-08-27 Jakub Jelinek <jakub@redhat.com>
9983 PR rtl-optimization/87065
9984 * combine.c (simplify_if_then_else): Formatting fix.
9985 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
9987 (known_cond): Don't return const_true_rtx for vector modes. Use
9988 CONST0_RTX instead of const0_rtx. Formatting fixes.
9990 2018-08-27 Martin Liska <mliska@suse.cz>
9992 PR gcov-profile/87069
9993 * gcov.c (process_file): Record files already processed
9994 and warn about a file being processed multiple times.
9996 2018-08-27 Martin Liska <mliska@suse.cz>
9999 * config/aarch64/aarch64.c (aarch64_override_options_internal):
10000 Set default values for x_aarch64_*_string strings.
10001 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
10002 prefix. For -mabi do not print '=ABI' in help and use
10003 <option_value> format for -msve-vector-bits and -moverride
10006 2018-08-26 Jeff Law <law@redhat.com>
10008 * config/mips/frame-header-opt.c: Include "backend.h" rather than
10011 2018-08-26 Marek Polacek <polacek@redhat.com>
10013 PR c++/87029, Implement -Wredundant-move.
10014 * doc/invoke.texi: Document -Wredundant-move.
10016 2018-08-25 Martin Sebor <msebor@redhat.com>
10018 PR tree-optimization/87059
10019 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
10020 to the same type as the other.
10021 * fold-const.c (fold_binary_loc): Assert expectation.
10023 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
10025 * config/darwin.c (machopic_legitimize_pic_address): Clean up
10026 extraneous parentheses, dead code section and formatting.
10028 2018-08-24 David Malcolm <dmalcolm@redhat.com>
10031 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
10032 wide enough for jumps in the line-numbering to be visible.
10033 (layout::print_gap_in_line_numbering): New member function.
10034 (layout::calculate_line_spans): When using line numbering, merge
10035 line spans that are only 1 line apart.
10036 (diagnostic_show_locus): When printing line numbers, show gaps in
10037 line numbering directly, rather than printing headers.
10038 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
10039 line-numbering with multiple line spans.
10040 (selftest::test_fixit_insert_containing_newline_2): Add test of
10041 line-numbering, in which the spans are close enough to be merged.
10043 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
10045 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
10046 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
10047 * tree-vrp.c (range_is_nonnull): Remove.
10048 (range_includes_zero_p): Accept value_range instead of min/max.
10049 (extract_range_from_binary_expr_1): Do not early bail on
10051 Use range_includes_zero_p instead of range_is_nonnull.
10052 (extract_range_from_unary_expr): Use range_includes_zero_p instead
10053 of range_is_nonnull.
10054 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
10055 special case VR_ANTI_RANGE.
10056 (vrp_finalize): Same.
10057 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
10058 instead of min/max.
10059 (range_is_nonnull): Remove.
10060 * vr-values.c (vrp_stmt_computes_nonzero): Use
10061 range_includes_zero_p instead of range_is_nonnull.
10062 (extract_range_basic): Pass value_range to range_includes_zero_p
10063 instead of range_is_nonnull.
10065 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
10067 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
10068 * emit-rtl.h (rtl_data): Remove return_bnd.
10069 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
10070 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
10071 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
10072 (POINTER_BOUNDS_MODE): Remove definition.
10073 (make_pointer_bounds_mode): Remove.
10074 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
10075 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
10076 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
10077 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
10078 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
10079 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
10080 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
10082 * config/i386/i386-modes.def (BND32, BND64): Remove.
10083 * config/i386/i386.c (dbx_register_map): Remove bound registers.
10084 (dbx64_register_map): Ditto.
10085 (svr4_dbx_register_map): Ditto.
10086 (indirect_thunk_bnd_needed): Remove.
10087 (indirect_thunks_bnd_used): Ditto.
10088 (indirect_return_bnd_needed): Ditto.
10089 (indirect_return_via_cx_bnd): Ditto.
10090 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
10091 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
10092 (output_indirect_thunk): Ditto. Remove need_prefix argument.
10093 (output_indirect_thunk_function): Remove handling of
10094 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
10095 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
10096 (ix86_save_reg): Remove handling of crtl->return_bnd.
10097 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
10098 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
10099 and UNSPEC_BNDLX_ADDR.
10100 (ix86_output_indirect_branch_via_reg): Remove handling of
10101 indirect_thunk_prefix_bnd.
10102 (ix86_output_indirect_branch_via_push): Ditto.
10103 (ix86_output_function_return): Ditto.
10104 (ix86_output_indirect_function_return): Ditto.
10105 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
10106 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
10107 (CALL_USED_REGISTERS): Ditto.
10108 (REG_ALLOC_ORDER): Update for removal of bound registers.
10109 (HI_REGISTER_NAMES): Ditto.
10110 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
10111 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
10112 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
10113 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
10114 (FIRST_PSEUDO_REG): Update.
10115 (BND): Remove mode iterator.
10116 * config/i386/predicates.md (bnd_mem_operator): Remove.
10118 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
10120 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
10123 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
10125 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
10126 the case in which the permute needs only a single element and
10127 repeats for every vector of the result. Extend that case to
10128 handle variable-length vectors.
10129 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
10131 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
10134 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
10135 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
10137 2018-08-24 Richard Biener <rguenther@suse.de>
10139 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
10140 bb_flags_allocated members.
10141 (auto_flag): New RAII class for allocating flags.
10142 (auto_edge_flag): New RAII class for allocating edge flags.
10143 (auto_bb_flag): New RAII class for allocating bb flags.
10144 * cfgloop.c (verify_loop_structure): Allocate temporary edge
10146 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
10147 in favor of temporarily allocated BB flag.
10148 * hsa-brig.c: Re-order includes.
10149 * hsa-dump.c: Likewise.
10150 * hsa-regalloc.c: Likewise.
10151 * print-rtl.c: Likewise.
10152 * profile-count.c: Likewise.
10154 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
10157 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
10160 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
10163 * wide-int-range.cc (wide_int_range_div): Do not ignore result
10164 from wide_int_range_multiplicative_op.
10166 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10168 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
10171 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
10173 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
10176 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
10178 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
10179 full GENERIC expression used for replacement.
10181 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
10183 * tree-vrp.c (abs_extent_range): Remove.
10184 (extract_range_into_wide_ints): Pass wide ints by reference.
10185 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
10186 Pass wide ints by reference in all calls to
10187 extract_range_into_wide_ints.
10188 * wide-int-range.cc (wide_int_range_div): New.
10189 * wide-int-range.h (wide_int_range_div): New.
10190 (wide_int_range_includes_zero_p): New.
10191 (wide_int_range_zero_p): New.
10193 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
10195 * config/aarch64/aarch64.md (arches): New enum.
10196 (arch): New enum attr.
10197 (arch_enabled): New attr.
10198 (enabled): Now uses arch_enabled only.
10199 (simd, sve, fp16): Removed attribute.
10200 (fp): Attr now defined in terms of 'arch'.
10201 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
10202 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
10203 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
10204 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
10205 attributes into 'arch'.
10206 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
10207 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
10208 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
10209 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
10210 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
10211 'simd' attribute into 'arch'.
10212 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
10213 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
10214 Convert use of 'fp' attribute to 'arch'.
10215 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
10216 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
10218 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
10219 (different modes) Merge 'fp' and 'simd' into 'arch'.
10220 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
10221 'simd' into 'arch'.
10223 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
10225 PR rtl-optimization/87026
10226 * expmed.c (canonicalize_comparison): If we can no longer create
10227 pseudoregisters, don't.
10229 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
10232 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
10234 * config/arm/arm.c (speculation_barrier_libfunc): New static
10236 (arm_init_libfuncs): Initialize it.
10237 (arm_emit_speculation_barrier): New function.
10238 * config/arm/arm.md (speculation_barrier): Call
10239 arm_emit_speculation_barrier for architectures that do not have
10241 (speculation_barrier_insn): Only match on Armv7 or later.
10243 2018-08-23 Richard Biener <rguenther@suse.de>
10245 PR middle-end/87024
10246 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
10249 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
10251 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
10252 of single-vector TBLs.
10253 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
10254 one input is given.
10256 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
10259 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
10260 aarch64_evpc_tbl guard.
10262 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
10264 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
10267 2018-08-22 Martin Sebor <msebor@redhat.com>
10269 PR middle-end/87052
10270 * tree-pretty-print.c (pretty_print_string): Add argument.
10271 (dump_generic_node): Call to pretty_print_string with string size.
10273 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
10275 PR rtl-optimization/86771
10276 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
10277 of two SETs into those two SETs, one to be placed at i2, if that SETs
10278 destination is modified between i2 and i3.
10280 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
10282 PR tree-optimization/86725
10283 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
10285 (vect_analyze_scalar_cycles_1): Check it.
10287 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
10289 PR tree-optimization/86725
10290 * tree-vect-loop.c (vect_is_simple_reduction): When treating
10291 an outer loop phi as a double reduction, make sure that the
10292 single user of the phi result is an inner loop phi.
10294 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
10296 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
10297 grouped stores with gaps to a strided group.
10299 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
10301 * tree-vect-stmts.c (get_group_load_store_type)
10302 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
10303 first statement in a group.
10305 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
10307 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
10308 the sequence used in gcc/gcc.c.
10310 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
10313 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
10316 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
10318 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
10319 Darwin10-specific unwinder-shim.
10320 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
10321 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
10322 New to cater for Darwin10 Rosetta.
10324 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
10326 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
10327 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
10329 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
10334 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
10335 (dwarf2out_switch_text_section): Generate a local label for the
10336 second function sub-section and apply it as the second FDE start
10338 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
10339 second sub-section start.
10341 2018-08-22 Richard Biener <rguenther@suse.de>
10343 PR tree-optimization/86988
10344 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
10346 2018-08-22 Richard Biener <rguenther@suse.de>
10348 PR tree-optimization/86945
10349 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
10351 2018-08-22 Alexandre Oliva <oliva@adacore.com>
10353 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
10354 a comment about how uses of r2 for .sdata2 come about.
10356 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
10358 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
10360 2018-08-21 Marek Polacek <polacek@redhat.com>
10362 PR c++/86981, Implement -Wpessimizing-move.
10363 * doc/invoke.texi: Document -Wpessimizing-move.
10365 2018-08-21 Jan Hubicka <jh@suse.cz>
10367 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
10368 * tree.h (is_redundant_typedef): Remove.
10369 * dwarf2out.c (is_redundant_typedef): Turn into static function.
10371 2018-08-21 Jan Hubicka <jh@suse.cz>
10373 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
10376 2018-08-21 Tamar Christina <tamar.christina@arm.com>
10378 * expmed.c (extract_low_bits): Reject invalid subregs early.
10380 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
10382 PR middle-end/86121
10383 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
10386 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
10388 * config/vxworks.h: Guard vxworks_asm_out_constructor and
10389 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
10390 * config/vxworks.c: Likewise.
10392 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
10394 * config/vxworks.c: Set targetm.have_ctors_dtors
10395 if HAVE_INITFINI_ARRAY_SUPPORT.
10396 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
10397 if HAVE_INITFINI_ARRAY_SUPPORT.
10399 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
10401 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
10402 default search path for VxWorks < 7.
10404 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
10406 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
10407 (get_format_string): Refer to c_getstr.
10409 2018-08-21 Tom de Vries <tdevries@suse.de>
10411 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
10412 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
10413 (debuginfo_early_stop): Declare.
10414 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
10415 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
10416 (debuginfo_early_stop): New function.
10417 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
10418 and debuginfo_early_stop.
10419 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
10420 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
10421 (general_init): Call debuginfo_early_init.
10422 (finalize): Call debuginfo_fini.
10423 (do_compile): Call debuginfo_init.
10424 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
10425 -fdump-early-debug.
10426 (@item -fdump-debug, @item -fdump-earlydebug): Add.
10428 2018-08-21 Tom de Vries <tdevries@suse.de>
10430 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
10431 flag_dump_noaddr and flag_dump_unnumbered.
10433 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
10435 * wide-int-range.cc (wide_int_range_abs): New.
10436 (wide_int_range_order_set): Rename from wide_int_range_min_max.
10437 * wide-int-range.h (wide_int_range_abs): New.
10438 (wide_int_range_min_max): New.
10439 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
10440 case to call wide_int_range_abs.
10441 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
10442 (extract_range_from_abs_expr): Delete.
10444 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
10447 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
10448 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
10451 2018-08-20 Nathan Sidwell <nathan@acm.org>
10452 Jeff Law <law@redhat.com>
10454 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
10455 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
10457 2018-08-20 David Malcolm <dmalcolm@redhat.com>
10460 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
10461 (decl_attributes): Likewise.
10462 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
10464 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
10465 * diagnostic-core.h (class auto_diagnostic_group): New class.
10466 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
10467 (diagnostic_report_diagnostic): Handle the first diagnostics within
10469 (emit_diagnostic): Add auto_diagnostic_group instance.
10470 (inform): Likewise.
10471 (inform_n): Likewise.
10472 (warning): Likewise.
10473 (warning_at): Likewise.
10474 (warning_n): Likewise.
10475 (pedwarn): Likewise.
10476 (permerror): Likewise.
10478 (error_n): Likewise.
10479 (error_at): Likewise.
10481 (fatal_error): Likewise.
10482 (internal_error): Likewise.
10483 (internal_error_no_backtrace): Likewise.
10484 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
10485 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
10486 * diagnostic.h (struct diagnostic_context): Add fields
10487 "diagnostic_group_nesting_depth",
10488 "diagnostic_group_emission_count", "begin_group_cb",
10490 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
10491 Add auto_diagnostic_group instance(s).
10492 (find_explicit_erroneous_behavior): Likewise.
10493 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
10494 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
10495 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
10496 (gimplify_va_arg_expr): Likewise.
10497 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
10498 (HSA_SORRY_AT): Likewise.
10499 * ipa-devirt.c (compare_virtual_tables): Likewise.
10500 (warn_odr): Likewise.
10501 * multiple_target.c (expand_target_clones): Likewise.
10502 * opts-common.c (cmdline_handle_error): Likewise.
10503 * reginfo.c (globalize_reg): Likewise.
10504 * substring-locations.c (format_warning_n_va): Likewise.
10505 * tree-inline.c (expand_call_inline): Likewise.
10506 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
10507 * tree-ssa-loop-niter.c
10508 (do_warn_aggressive_loop_optimizations): Likewise.
10509 * tree-ssa-uninit.c (warn_uninit): Likewise.
10510 * tree.c (warn_deprecated_use): Likewise.
10512 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
10515 * config/i386/i386.md (eh_return): Always update EH return
10516 address in word_mode.
10518 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
10520 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
10521 TARGET_SPLIT_COMPLEX_ARG is defined.
10523 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
10525 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
10527 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
10530 * expr.c (expand_assignment): Assert that bitpos is positive.
10531 (store_field): Likewise
10532 (expand_expr_real_1): Make sure that bitpos is positive.
10533 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
10536 2018-08-20 Nathan Sidwell <nathan@acm.org>
10538 * Makefile.in (CPP_ID_DATA_H): Delete.
10539 (CPP_INTERNAL_H): Don't add it.
10540 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
10541 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
10543 2018-08-20 Richard Biener <rguenther@suse.de>
10545 PR tree-optimization/78655
10546 * tree-vrp.c (extract_range_from_binary_expr_1): Make
10547 pointer + offset nonnull if either operand is nonnull work.
10549 2018-08-20 Tom de Vries <tdevries@suse.de>
10551 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
10552 unless the referenced die describes the added property using
10553 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
10554 Otherwise, add a DW_AT_location to the referenced die.
10556 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
10559 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
10560 register_operand when calling ix86_set_reg_reg_cost.
10561 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
10562 Set *total to 0 for operands that satisfy x86_64_immediate_operand
10563 predicate and to 1 otherwise.
10565 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
10567 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
10568 emit a diagnostic that it is not supported and reset the option.
10569 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
10570 supported and consume the option. (ASM_FINAL_SPEC): New.
10572 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
10574 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
10577 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
10579 C-SKY port: Documentation
10581 * doc/extend.texi (C-SKY Function Attributes): New section.
10582 * doc/invoke.texi (Option Summary): Add C-SKY options.
10583 (C-SKY Options): New section.
10584 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
10586 2018-08-17 Jojo <jijie_rong@c-sky.com>
10587 Huibin Wang <huibin_wang@c-sky.com>
10588 Sandra Loosemore <sandra@codesourcery.com>
10589 Chung-Lin Tang <cltang@codesourcery.com>
10591 C-SKY port: Backend implementation
10593 * config/csky/*: New.
10594 * common/config/csky/*: New.
10596 2018-08-17 Jojo <jijie_rong@c-sky.com>
10597 Huibin Wang <huibin_wang@c-sky.com>
10598 Sandra Loosemore <sandra@codesourcery.com>
10599 Chung-Lin Tang <cltang@codesourcery.com>
10600 Andrew Jenner <andrew@codesourcery.com>
10602 C-SKY port: Configury
10604 * config.gcc (csky-*-*): New.
10605 * configure.ac: Add csky to targets for dwarf2 debug_line support.
10606 * configure: Regenerated.
10608 2018-08-17 David Malcolm <dmalcolm@redhat.com>
10610 * dump-context.h: Include "dumpfile.h".
10611 (dump_context::dump_printf_va): Convert final param from va_list
10612 to va_list *. Convert from ATTRIBUTE_PRINTF to
10613 ATTRIBUTE_GCC_DUMP_PRINTF.
10614 (dump_context::dump_printf_loc_va): Likewise.
10615 * dumpfile.c: Include "stringpool.h".
10616 (make_item_for_dump_printf_va): Delete.
10617 (make_item_for_dump_printf): Delete.
10618 (class dump_pretty_printer): New class.
10619 (dump_pretty_printer::dump_pretty_printer): New ctor.
10620 (dump_pretty_printer::emit_items): New member function.
10621 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
10623 (dump_pretty_printer::emit_item): New member function.
10624 (dump_pretty_printer::stash_item): New member function.
10625 (dump_pretty_printer::format_decoder_cb): New member function.
10626 (dump_pretty_printer::decode_format): New member function.
10627 (dump_context::dump_printf_va): Reimplement in terms of
10628 dump_pretty_printer.
10629 (dump_context::dump_printf_loc_va): Convert final param from va_list
10631 (dump_context::begin_scope): Reimplement call to
10632 make_item_for_dump_printf.
10633 (dump_printf): Update for change to dump_printf_va.
10634 (dump_printf_loc): Likewise.
10635 (selftest::test_capture_of_dump_calls): Convert "stmt" from
10636 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
10637 with %T, %E, and %G.
10638 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
10639 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
10640 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
10641 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
10642 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
10643 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
10644 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
10645 within a dump_printf_loc call to "%wu".
10646 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
10647 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
10648 missing space after "=".
10649 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
10650 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
10651 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
10652 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
10653 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
10654 duplicate "vectorized" from message.
10656 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
10658 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
10659 polyNxK_t element's TYPE_STRING_FLAG.
10661 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
10663 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
10664 (they were unnamed before). Fix comments.
10666 2018-08-17 Nathan Sidwell <nathan@acm.org>
10668 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
10670 2018-08-17 Richard Biener <rguenther@suse.de>
10672 PR tree-optimization/86841
10673 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
10675 2018-08-17 Martin Liska <mliska@suse.cz>
10677 * common.opt: Remove Warn, Init and Report for options with
10678 Ignore/Deprecated flag. Warning is done automatically for
10680 * config/i386/i386.opt: Likewise.
10681 * config/ia64/ia64.opt: Likewise.
10682 * config/rs6000/rs6000.opt: Likewise.
10683 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
10684 Remove usage of flag_check_pointer_bounds.
10685 * lto-wrapper.c (merge_and_complain): Do not handle
10686 OPT_fcheck_pointer_bounds.
10687 (append_compiler_options): Likewise.
10688 * opt-functions.awk: Do not handle Deprecated.
10689 * optc-gen.awk: Check that Var, Report and Init are not
10690 used for an option with Ignore/Deprecated flag.
10691 * opts-common.c (decode_cmdline_option): Do not report
10693 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
10695 * opts.h (struct cl_option): Remove cl_deprecated flag.
10696 (CL_ERR_DEPRECATED): Remove error enum value.
10698 2018-08-17 Richard Biener <rguenther@suse.de>
10700 PR middle-end/86505
10701 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
10702 across a va-arg-pack using call adjust its return value accordingly.
10704 2018-08-16 Martin Sebor <msebor@redhat.com>
10706 PR tree-optimization/86853
10707 * gimple-ssa-sprintf.c (struct format_result): Rename member.
10708 (struct fmtresult): Add member and initialize it in ctors.
10709 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
10710 (format_string): Handle %S the same as %ls. Set MAYFAIL.
10711 (format_directive): Set POSUNDER4K when MAYFAIL is set.
10712 (parse_directive): Handle %C same as %c.
10713 (sprintf_dom_walker::compute_format_length): Adjust.
10714 (is_call_safe): Adjust.
10716 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
10718 * builtins.c (c_strlen): Add new parameter eltsize. Use it
10719 for determining how to count the elements.
10720 * builtins.h (c_strlen): Adjust prototype.
10721 * expr.c (string_constant): Add new parameter mem_size.
10722 Set *mem_size appropriately.
10723 * expr.h (string_constant): Adjust protoype.
10724 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
10725 * gimple-fold.h (get_range_strlen): Adjust prototype.
10726 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
10727 (format_string): Call get_string_length with eltsize.
10729 2018-08-16 David Malcolm <dmalcolm@redhat.com>
10731 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
10732 to emit the span, rather than setting it as the prefix.
10734 2018-08-16 David Malcolm <dmalcolm@redhat.com>
10736 * diagnostic-show-locus.c (layout::start_annotation_line): Add
10737 "margin_char" parameter, defaulting to space. Use it in place
10738 of pp_space for the initial part of the margin.
10739 (layout::print_leading_fixits): Use '+' when filling the margin
10740 of line-insertion fix-it hints.
10742 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
10744 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
10747 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
10749 * config/rs6000/altivec.md: Don't set length attribute to the default
10751 * config/rs6000/darwin.md: Ditto.
10752 * config/rs6000/dfp.md: Ditto.
10753 * config/rs6000/htm.md: Ditto.
10754 * config/rs6000/rs6000.md: Ditto.
10755 * config/rs6000/sync.md: Ditto.
10756 * config/rs6000/vsx.md: Ditto.
10758 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
10760 * config/rs6000/altivec.md: Don't set length attribute to the default
10761 value, for branch instructions.
10762 * config/rs6000/darwin.md: Ditto.
10763 * config/rs6000/rs6000.md: Ditto.
10765 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
10767 * config/rs6000/rs6000.md (length): Always define as const_int 4.
10768 (unnamed conditional branch define_insn): Set length to 4 or 8
10769 depending on offset.
10770 (<bd>_<mode>): Similar, for alternative 0.
10771 (<bd>tf_<mode>): Ditto.
10773 2018-08-16 Tamar Christina <tamar.christina@arm.com>
10775 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
10777 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
10779 * doc/rtl.texi: Replace old RTX class names with new names.
10782 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
10784 * expmed.h (canonicalize_comparison): New declaration.
10785 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
10786 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
10787 * optabs.c (prepare_cmp_insn): Likewise.
10788 * rtl.h (unsigned_condition_p): New function which checks if a
10789 comparison operator is unsigned.
10791 2018-08-16 Nathan Sidwell <nathan@acm.org>
10793 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
10794 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
10796 2018-08-16 Tamar Christina <tamar.christina@arm.com>
10799 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
10800 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
10801 (mov<mov>): ..this and enable unconditionally.
10803 2018-08-16 Tamar Christina <tamar.christina@arm.com>
10805 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
10807 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
10809 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
10810 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
10811 "Common" with "Target".
10813 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
10815 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
10816 * doc/invoke.texi (mmitigate-rop): Remove.
10817 * config/i386/i386.c: Do not include "regrename.h".
10818 (ix86_rop_should_change_byte_p, reg_encoded_number)
10819 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
10821 (ix86_reorg): Remove call to ix86_mitigate_rop.
10822 * config/i386/i386.md (attr "modrm_class"): Remove.
10823 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
10824 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
10825 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
10827 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
10829 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
10830 allow folding of mergeh() and mergel() for the float and double types.
10831 (fold_mergehl_helper): Rework to handle building a permute tree
10834 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
10836 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
10839 2018-08-15 David Malcolm <dmalcolm@redhat.com>
10841 * common.opt (fdiagnostics-show-labels): New option.
10842 * diagnostic-show-locus.c (class layout_range): Add field
10844 (class layout): Add field "m_show_labels_p".
10845 (layout_range::layout_range): Add param "label" and use it to
10846 initialize m_label.
10847 (make_range): Pass in NULL for new "label" param of layout_range's
10849 (layout::layout): Initialize m_show_labels_p.
10850 (layout::maybe_add_location_range): Pass in loc_range->m_label
10851 when constructing layout_range instances.
10852 (struct line_label): New struct.
10853 (layout::print_any_labels): New member function.
10854 (layout::print_line): Call it if label-printing is enabled.
10855 (selftest::test_one_liner_labels): New test.
10856 (selftest::test_diagnostic_show_locus_one_liner): Call it.
10857 * diagnostic.c (diagnostic_initialize): Initialize
10858 context->show_labels_p.
10859 * diagnostic.h (struct diagnostic_context): Add field
10861 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
10862 -fno-diagnostics-show-labels.
10863 * dwarf2out.c (gen_producer_string): Add
10864 OPT_fdiagnostics_show_labels to the ignored options.
10865 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
10867 (gcc_rich_location::maybe_add_expr): Likewise.
10868 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
10869 label" param, defaulting to NULL.
10870 (gcc_rich_location::add_expr): Add "label" param.
10871 (gcc_rich_location::maybe_add_expr): Likewise.
10872 (class text_range_label): New class.
10873 (class range_label_for_type_mismatch): New class.
10874 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
10875 of format_warning_va.
10876 (fmtwarn_n): Likewise for new params of format_warning_n_va.
10877 * lto-wrapper.c (merge_and_complain): Add
10878 OPT_fdiagnostics_show_labels to the "pick one setting" options.
10879 (append_compiler_options): Likewise to the dropped options.
10880 (append_diag_options): Likewise to the passed-on options.
10881 * opts.c (common_handle_option): Handle the new option.
10882 * selftest-diagnostic.c
10883 (test_diagnostic_context::test_diagnostic_context): Enable
10885 * substring-locations.c: Include "gcc-rich-location.h".
10886 (format_warning_n_va): Add "fmt_label" and "param_label" params
10887 and use them as appropriate.
10888 (format_warning_va): Add "fmt_label" and "param_label" params,
10889 passing them on to format_warning_n_va.
10890 (format_warning_at_substring): Likewise.
10891 (format_warning_at_substring_n): Likewise.
10892 * substring-locations.h (format_warning_va): Add "fmt_label" and
10893 "param_label" params.
10894 (format_warning_n_va): Likewise.
10895 (format_warning_at_substring): Likewise.
10896 (format_warning_at_substring_n): Likewise.
10897 * toplev.c (general_init): Initialize global_dc->show_labels_p.
10899 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
10902 * builtins.c (expand_builtin_memcmp): Do not expand the call
10903 when overflow is detected.
10905 2018-08-15 Martin Sebor <msebor@redhat.com>
10907 PR tree-optimization/71625
10908 * config/aarch64/aarch64-builtins.c
10909 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
10911 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
10913 * config/s390/s390.c (s390_reorg): Remove loop.
10915 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
10918 (darwin_function_switched_text_sections): Delete.
10919 * gcc/config/darwin.h
10920 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
10922 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
10925 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
10926 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
10927 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
10929 2018-08-15 Martin Liska <mliska@suse.cz>
10931 PR tree-optimization/86925
10932 * predict.c (expr_expected_value_1): When taking
10933 later predictor, assign also probability.
10934 Use fold_build2_initializer_loc in order to fold
10935 the expression in -frounding-math.
10937 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
10939 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
10941 (expand_vec_perm_1): Try the new method.
10943 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
10946 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
10947 Check whether lra_live_max_point is 0 before dividing.
10949 2018-08-14 Martin Sebor <msebor@redhat.com>
10951 PR tree-optimization/86650
10952 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
10953 (vrp_prop::check_mem_ref): Same.
10955 2018-08-13 Liu Hao <lh_mouse@126.com>
10957 * pretty-print.c (eat_esc_sequence): Swap the foreground and
10958 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
10959 and clear it thereafter, as it only works for DBCS.
10961 2018-08-13 Liu Hao <lh_mouse@126.com>
10963 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
10964 handle returned by _get_osf_handle().
10966 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
10968 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
10969 for folding vec_perm.
10971 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
10973 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
10974 Add support for gimple-folding of vec_pack() and vec_unpack()
10977 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
10979 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
10980 vec_xst variants to the list.
10981 (rs6000_gimple_fold_builtin): Add support for folding unaligned
10982 vector loads and stores.
10984 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
10986 * config.gcc (rs6000-ibm-aix4.x): Delete.
10987 (rs6000-ibm-aix5.1): Delete.
10988 (rs6000-ibm-aix5.2): Delete.
10989 (rs6000-ibm-aix5.3): Delete.
10990 * config/rs6000/aix43.h: Delete.
10991 * config/rs6000/aix51.h: Delete.
10992 * config/rs6000/aix52.h: Delete.
10993 * config/rs6000/t-aix43: Delete.
10995 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
10997 * config/s390/s390.c (s390_decompose_constant_pool_ref):
10999 (s390_decompose_address): Factor out constant pool ref
11002 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
11004 * config/nds32/nds32-predicates.c
11005 (nds32_can_use_bclr_p): Change return type as bool.
11006 (nds32_can_use_bset_p): Ditto.
11007 (nds32_can_use_btgl_p): Ditto.
11008 (nds32_can_use_bitci_p): Ditto.
11009 * config/nds32/nds32-protos.h
11010 (nds32_can_use_bclr_p): Change declaration.
11011 (nds32_can_use_bset_p): Ditto.
11012 (nds32_can_use_btgl_p): Ditto.
11013 (nds32_can_use_bitci_p): Ditto.
11015 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
11017 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
11018 Support -msched-prolog-epilog option.
11019 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
11021 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
11023 * common/config/nds32/nds32-common.c
11024 (nds32_option_optimization_table): Enalbe -malways-align.
11026 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
11028 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
11030 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
11031 OPT_misr_secure_ case.
11032 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
11033 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
11034 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
11036 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
11037 (nds32_isr_info): New field security_level.
11038 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
11039 * config/nds32/nds32.md (return_internal): Consider critical attribute.
11040 * config/nds32/nds32.opt (misr-secure): New option.
11041 * config/nds32/nds32_init.inc: New file.
11042 * config/nds32/nds32_isr.h: New file.
11044 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
11046 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
11047 Update comment for atomic instructions.
11048 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
11049 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
11051 (atomic_loaddi): Revise fence expansion to only emit fence prior to
11052 load for __ATOMIC_SEQ_CST model.
11053 (atomic_loaddi_1): Remove float register target.
11054 (atomic_storedi): Handle CONST_INT values.
11055 (atomic_storedi_1): Remove float register source. Add special case
11057 (memory_barrier): New expander and insn.
11059 2018-08-11 Jakub Jelinek <jakub@redhat.com>
11061 PR tree-optimization/86835
11062 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
11063 new_stmt after def_gsi, make sure to insert new_square_stmt after
11064 that stmt, not 2 stmts before it.
11066 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
11069 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
11072 2018-08-10 Martin Liska <mliska@suse.cz>
11075 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
11077 * builtins.c (expand_builtin_expect_with_probability):
11079 (expand_builtin_expect_with_probability): New function.
11080 (build_builtin_expect_predicate): Add new argumnet probability
11081 for BUILT_IN_EXPECT_WITH_PROBABILITY.
11082 (fold_builtin_expect):
11085 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
11086 * builtins.h (fold_builtin_expect): Set new argument.
11087 * doc/extend.texi: Document __builtin_expect_with_probability.
11088 * doc/invoke.texi: Likewise.
11089 * gimple-fold.c (gimple_fold_call): Pass new argument.
11090 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
11091 also BUILT_IN_EXPECT_WITH_PROBABILITY.
11092 * predict.c (get_predictor_value): New function.
11093 (expr_expected_value): Add new argument probability. Assume
11094 that predictor and probability are always non-null.
11095 (expr_expected_value_1): Likewise. For __builtin_expect and
11096 __builtin_expect_with_probability set probability. Handle
11097 combination in binary expressions.
11098 (tree_predict_by_opcode): Simplify code by simply calling
11099 get_predictor_value.
11100 (pass_strip_predict_hints::execute): Add handling of
11101 BUILT_IN_EXPECT_WITH_PROBABILITY.
11102 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
11104 * tree.h (DECL_BUILT_IN_P): New function.
11106 2018-08-10 Martin Liska <mliska@suse.cz>
11108 PR tree-optimization/85799
11109 * passes.def: Add argument for pass_strip_predict_hints.
11110 * predict.c (class pass_strip_predict_hints): Add new argument
11112 (strip_predictor_early): New function.
11113 (pass_strip_predict_hints::execute): Call the function to
11115 (strip_predict_hints): New function.
11116 * predict.def: Fix comment.
11118 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
11120 * Makefile.in: Clarify which tm.texi to copy over to assert the
11121 right to grant a GFDL license for all.
11123 2018-08-09 Jeff Law <law@redhat.com>
11125 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
11128 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
11130 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
11133 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
11135 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
11136 reductions for variable-length vectors.
11138 2018-08-09 David Malcolm <dmalcolm@redhat.com>
11141 * common.opt (fdiagnostics-show-line-numbers): New option.
11142 * diagnostic-show-locus.c (class layout): Add fields
11143 "m_show_line_numbers_p" and "m_linenum_width";
11144 (num_digits): New function.
11145 (test_num_digits): New function.
11146 (layout::layout): Initialize new fields. Update m_x_offset
11147 logic to handle any left margin.
11148 (layout::print_source_line): Print line number when requested.
11149 (layout::start_annotation_line): New member function.
11150 (layout::print_annotation_line): Call it.
11151 (layout::print_leading_fixits): Likewise.
11152 (layout::print_trailing_fixits): Likewise. Update calls to
11153 move_to_column for new parameter.
11154 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
11155 it to potentially call start_annotation_line.
11156 (layout::show_ruler): Call start_annotation_line.
11157 (selftest::test_line_numbers_multiline_range): New selftest.
11158 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
11159 and selftest::test_line_numbers_multiline_range.
11160 * diagnostic.c (diagnostic_initialize): Initialize
11161 show_line_numbers_p.
11162 * diagnostic.h (struct diagnostic_context): Add field
11163 "show_line_numbers_p".
11164 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
11165 -fno-diagnostics-show-line-numbers.
11166 * dwarf2out.c (gen_producer_string): Add
11167 OPT_fdiagnostics_show_line_numbers to the ignored options.
11168 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
11169 one setting" options.
11170 (append_compiler_options): Likewise to the dropped options.
11171 (append_diag_options): Likewise to the passed-on options.
11172 * opts.c (common_handle_option): Handle the new option.
11173 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
11175 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
11177 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
11178 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
11179 third argument of type "const signed char" to descriptions of
11180 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
11181 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
11182 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
11183 __builtin_bcdsub_ov functions.
11185 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
11187 PR tree-optimization/86858
11188 * tree-vect-loop.c (vect_is_simple_reduction): Restore
11189 flow_bb_inside_loop_p calls.
11191 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
11193 PR tree-optimization/86871
11194 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
11195 instead of gimple_assign_lhs.
11197 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
11200 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
11201 register constraint to operand 0.
11202 (add<mode>3_carryinC): Likewise.
11203 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
11205 2018-08-09 Martin Liska <mliska@suse.cz>
11208 * common.opt: Remove extra line.
11210 2018-08-09 Martin Liska <mliska@suse.cz>
11212 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
11213 at the end of a line, make first letter capital and end up
11214 a sentence with a dot.
11215 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
11216 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
11217 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
11218 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
11219 (PARAM_MAX_ISL_OPERATIONS): Likewise.
11220 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
11221 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
11222 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
11223 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
11224 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
11225 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
11226 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
11227 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
11228 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
11229 (PARAM_TREE_REASSOC_WIDTH): Likewise.
11230 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
11231 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
11232 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
11234 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
11237 * config/s390/s390.c (s390_option_override_internal): Reduce the
11238 stack-clash-protection-probe-interval param if it would be too big
11241 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
11244 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
11245 * config/m68k/m68k.c (handle_move_double): Don't call
11246 m68k_final_prescan_insn.
11247 (m68k_adjust_decorated_operand): Renamed from
11248 m68k_final_prescan_insn, remove first and third operand and
11250 (print_operand): Call it.
11251 (print_operand_address): Call it.
11253 2018-08-08 Nathan Sidwell <nathan@acm.org>
11255 * diagnostic.c (diagnostic_report_current_module): Use
11256 linemap_included_from & linemap_included_from_linemap.
11258 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
11260 * config/aarch64/aarch64-cores.def: Add phecda core.
11261 * config/aarch64/aarch64-tune.md: Regenerate.
11262 * doc/invoke.texi: Add phecda core.
11264 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
11267 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
11269 * config/s390/s390.md ("movti"): Add more alternatives for
11270 constant to GPR copies.
11272 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
11274 * config/s390/s390.c: Fix whitespace damage throughout the file.
11275 * config/s390/s390.h: Likewise.
11276 * config/s390/tpf.h: Likewise.
11278 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
11280 * config/s390/s390.c (s390_loadrelative_operand_p):
11281 Remove TARGET_CPU_ZARCH usages.
11282 (s390_rtx_costs): Likewise.
11283 (s390_legitimate_constant_p): Likewise.
11284 (s390_cannot_force_const_mem): Likewise.
11285 (legitimate_reload_constant_p): Likewise.
11286 (s390_preferred_reload_class): Likewise.
11287 (legitimize_pic_address): Likewise.
11288 (legitimize_tls_address): Likewise.
11289 (s390_split_branches): Removed.
11290 (s390_add_execute): Removed.
11291 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
11292 (s390_mainpool_start): Likewise.
11293 (s390_mainpool_finish): Likewise.
11294 (s390_mainpool_cancel): Removed.
11295 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
11296 (s390_chunkify_cancel): Likewise.
11297 (s390_return_addr_rtx): Likewise.
11298 (s390_register_info): Remove split_branches_pending_p uages.
11299 (s390_optimize_register_info): Likewise.
11300 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
11301 split_branches_pending_p usages.
11302 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
11303 (s390_load_got): Likewise.
11304 (s390_expand_split_stack_prologue): Likewise.
11305 (output_asm_nops): Likewise.
11306 (s390_function_profiler): Likewise.
11307 (s390_emit_call): Likewise.
11308 (s390_conditional_register_usage): Likewise.
11309 (s390_optimize_prologue): Likewise.
11310 (s390_reorg): Remove TARGET_CPU_ZARCH and
11311 split_branches_pending_p usages.
11312 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
11314 (s390_output_indirect_thunk_function): Likewise.
11315 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
11316 (TARGET_CPU_ZARCH_P): Removed.
11317 (struct machine_function): Remove split_branches_pending_p.
11318 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
11320 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
11322 * common/config/s390/s390-common.c (processor_flags_table):
11324 * config.gcc: Remove with_arch/with_tune support.
11325 * config/s390/2064.md: Remove cpu attribute comparisons.
11326 * config/s390/driver-native.c (s390_host_detect_local_cpu):
11328 * config/s390/linux.h (ASM_SPEC):
11329 Remove -march support.
11330 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
11331 Use a table to get an arch level.
11332 * config/s390/s390-opts.h (enum processor_type):
11333 Remove enum values.
11334 * config/s390/s390.c
11335 (processor_table): Remove entries, add arch_level values.
11336 (s390_issue_rate): Remove cases.
11337 (s390_option_override): Adjust
11338 s390_option_override_internal() call.
11339 (s390_option_override_internal): Remove deprecation warning.
11340 (s390_valid_target_attribute_tree): Adjust
11341 s390_option_override_internal() call.
11342 * config/s390/s390.h (struct s390_processor):
11343 Share with s390-c.c, add arch_level field.
11344 * config/s390/s390.md:
11345 Remove occurrences in cpu attribute.
11346 * config/s390/s390.opt: Remove -march/-mtune support.
11347 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
11348 * doc/invoke.texi: Remove deprecation warning.
11350 2018-08-08 Luis Machado <luis.machado@linaro.org>
11352 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
11354 (qdf24xx_tunings): Set vector cost structure to
11355 qdf24xx_vector_cost.
11357 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
11358 <register_sextend>: Set to 3.
11360 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
11363 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
11364 * config/aarch64/aarch64-simd.md
11365 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
11366 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
11367 (aarch64_frecpx<mode>): ...this new pattern.
11368 * config/aarch64/aarch64-simd-builtins.def: Remove comment
11369 about aarch64_frecp<FRECP:frecp_suffix><mode>.
11371 2018-08-07 Martin Liska <mliska@suse.cz>
11373 PR middle-end/83023
11374 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
11375 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
11376 * predict.def (PRED_MALLOC_NONNULL): New predictor.
11377 * doc/extend.texi: Document that malloc attribute adds
11380 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
11383 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11384 Define to speculation_safe_value_not_needed.
11386 2018-08-06 Jeff Law <law@redhat.com>
11388 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
11389 the vr_values instance to cprop_into_stmt.
11390 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
11391 (cprop_operand): Also query EVRP to determine if OP is a constant.
11393 2018-08-06 Nathan Sidwell <nathan@acm.org>
11395 * diagnostic.c (diagnostic_report_current_module): Reroll
11396 included-at loop. Translate text.
11398 2018-08-06 David Malcolm <dmalcolm@redhat.com>
11400 * function-tests.c (selftest::test_expansion_to_rtl): Call
11401 free_after_compilation.
11403 2018-08-06 Alan Hayward <alan.hayward@arm.com>
11405 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
11407 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
11409 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
11410 loops with memory block operations from getting unrolled.
11412 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
11415 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11416 Define to speculation_safe_value_not_needed.
11418 2018-08-06 Jeff Law <law@redhat.com>
11420 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
11423 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11426 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
11427 with all enabled __intN types.
11429 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
11431 2018-08-06 Alan Hayward <alan.hayward@arm.com>
11433 * alias.c (record_set): Check for clobber high.
11434 * cfgexpand.c (expand_gimple_stmt): Likewise.
11435 * combine-stack-adj.c (single_set_for_csa): Likewise.
11436 * combine.c (find_single_use_1): Likewise.
11437 (set_nonzero_bits_and_sign_copies): Likewise.
11438 (get_combine_src_dest): Likewise.
11439 (is_parallel_of_n_reg_sets): Likewise.
11440 (try_combine): Likewise.
11441 (record_dead_and_set_regs_1): Likewise.
11442 (reg_dead_at_p_1): Likewise.
11443 (reg_dead_at_p): Likewise.
11444 * dce.c (deletable_insn_p): Likewise.
11445 (mark_nonreg_stores_1): Likewise.
11446 (mark_nonreg_stores_2): Likewise.
11447 * df-scan.c (df_find_hard_reg_defs): Likewise.
11448 (df_uses_record): Likewise.
11449 (df_get_call_refs): Likewise.
11450 * dwarf2out.c (mem_loc_descriptor): Likewise.
11451 * haifa-sched.c (haifa_classify_rtx): Likewise.
11452 * ira-build.c (create_insn_allocnos): Likewise.
11453 * ira-costs.c (scan_one_insn): Likewise.
11454 * ira.c (equiv_init_movable_p): Likewise.
11455 (rtx_moveable_p): Likewise.
11456 (interesting_dest_for_shprep): Likewise.
11457 * jump.c (mark_jump_label_1): Likewise.
11458 * postreload-gcse.c (record_opr_changes): Likewise.
11459 * postreload.c (reload_cse_simplify): Likewise.
11460 (struct reg_use): Add source expr.
11461 (reload_combine): Check for clobber high.
11462 (reload_combine_note_use): Likewise.
11463 (reload_cse_move2add): Likewise.
11464 (move2add_note_store): Likewise.
11465 * print-rtl.c (print_pattern): Likewise.
11466 * recog.c (decode_asm_operands): Likewise.
11467 (store_data_bypass_p): Likewise.
11468 (if_test_bypass_p): Likewise.
11469 * regcprop.c (kill_clobbered_value): Likewise.
11470 (kill_set_value): Likewise.
11471 * reginfo.c (reg_scan_mark_refs): Likewise.
11472 * reload1.c (maybe_fix_stack_asms): Likewise.
11473 (eliminate_regs_1): Likewise.
11474 (elimination_effects): Likewise.
11475 (mark_not_eliminable): Likewise.
11476 (scan_paradoxical_subregs): Likewise.
11477 (forget_old_reloads_1): Likewise.
11478 * reorg.c (find_end_label): Likewise.
11479 (try_merge_delay_insns): Likewise.
11480 (redundant_insn): Likewise.
11481 (own_thread_p): Likewise.
11482 (fill_simple_delay_slots): Likewise.
11483 (fill_slots_from_thread): Likewise.
11484 (dbr_schedule): Likewise.
11485 * resource.c (update_live_status): Likewise.
11486 (mark_referenced_resources): Likewise.
11487 (mark_set_resources): Likewise.
11488 * rtl.c (copy_rtx): Likewise.
11489 * rtlanal.c (reg_referenced_p): Likewise.
11490 (single_set_2): Likewise.
11491 (noop_move_p): Likewise.
11492 (note_stores): Likewise.
11493 * sched-deps.c (sched_analyze_reg): Likewise.
11494 (sched_analyze_insn): Likewise.
11496 2018-08-06 Alan Hayward <alan.hayward@arm.com>
11498 * cse.c (invalidate_reg): New function extracted from...
11499 (invalidate): ...here.
11500 (canonicalize_insn): Check for clobber high.
11501 (invalidate_from_clobbers): invalidate clobber highs.
11502 (invalidate_from_sets_and_clobbers): Likewise.
11503 (count_reg_usage): Check for clobber high.
11504 (insn_live_p): Likewise.
11505 * cselib.c (cselib_expand_value_rtx_1):Likewise.
11506 (cselib_invalidate_regno): Check for clobber in setter.
11507 (cselib_invalidate_rtx): Pass through setter.
11508 (cselib_invalidate_rtx_note_stores):
11509 (cselib_process_insn): Check for clobber high.
11510 * cselib.h (cselib_invalidate_rtx): Add operand.
11512 2018-08-06 Alan Hayward <alan.hayward@arm.com>
11514 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
11515 (mark_not_eliminable): Likewise.
11516 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
11517 * lra-lives.c (process_bb_lives): Check for clobber high.
11518 * lra.c (new_insn_reg): Remember clobber highs.
11519 (collect_non_operand_hard_regs): Check for clobber high.
11520 (lra_set_insn_recog_data): Likewise.
11521 (add_regs_to_insn_regno_info): Likewise.
11522 (lra_update_insn_regno_info): Likewise.
11524 2018-08-06 Alan Hayward <alan.hayward@arm.com>
11526 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
11527 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
11529 2018-08-06 Alan Hayward <alan.hayward@arm.com>
11531 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
11532 (copy_insn_1): Likewise.
11533 (gen_hard_reg_clobber_high): New gen function.
11534 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
11535 * genemit.c (gen_exp): Likewise.
11536 (gen_emit_seq): Pass through info.
11537 (gen_insn): Check for CLOBBER_HIGH.
11538 (gen_expand): Pass through info.
11539 (gen_split): Likewise.
11540 (output_add_clobbers): Likewise.
11541 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
11542 (remove_clobbers): Likewise.
11543 * rtl.h (gen_hard_reg_clobber_high): New declaration.
11545 2018-08-06 Alan Hayward <alan.hayward@arm.com>
11547 * doc/rtl.texi (clobber_high): Add.
11548 (parallel): Add in clobber high
11549 * rtl.c (rtl_check_failed_code3): Add function.
11550 * rtl.def (CLOBBER_HIGH): Add expression.
11551 * rtl.h (RTL_CHECKC3): Add macro.
11552 (rtl_check_failed_code3): Add declaration.
11553 (XC3EXP): Add macro.
11555 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
11558 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
11559 cfun->machine->max_used_stack_alignment if needed.
11561 2018-08-04 Martin Sebor <msebor@redhat.com>
11563 PR tree-optimization/86571
11564 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
11567 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
11570 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11573 2018-08-03 Jeff Law <law@redhat.com>
11576 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11577 Define to speculation_safe_value_not_needed.
11579 2018-08-03 David Malcolm <dmalcolm@redhat.com>
11581 * doc/gcov.texi (-x): Remove duplicate "to".
11582 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
11583 (-Wif-not-aligned): Remove duplicate "is".
11584 (-flto): Remove duplicate "the".
11585 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
11586 duplicate "v5.00.b".
11587 (MSP430 Options): Remove duplicate "and" from the description
11588 of "-mgprel-sec=regexp".
11589 (x86 Options): Remove duplicate copies of "vmldLog102" and
11590 vmlsLog104 from description of "-mveclibabi=type".
11592 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
11594 * internal-fn.h (first_commutative_argument): Declare.
11595 * internal-fn.c (first_commutative_argument): New function.
11596 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
11597 restrictions for pattern statements. Use first_commutative_argument
11598 to look for commutative operands in calls to internal functions.
11600 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
11602 * Makefile.in (wide-int-range.o): New.
11603 * tree-vrp.c: Move all the wide_int_* functions to...
11604 * wide-int-range.cc: ...here.
11605 * tree-vrp.h: Move all the wide_int_* prototypes to...
11606 * wide-int-range.h: ...here.
11608 2018-08-03 Tom de Vries <tdevries@suse.de>
11610 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
11612 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
11613 * except.c (output_function_exception_table): Do early exit if
11614 targetm_common.except_unwind_info (&global_options) == UI_NONE.
11616 2018-08-03 Martin Liska <mliska@suse.cz>
11618 * predict.c (dump_prediction): Change to 2 digits
11621 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
11623 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
11625 (falkor_am_1_gtov_gtov): ... a new insn reservation.
11627 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
11629 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
11630 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
11631 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
11632 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
11633 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
11635 2018-08-02 David Malcolm <dmalcolm@redhat.com>
11637 * diagnostic-show-locus.c (diagnostic_show_locus): Use
11638 pp_take_prefix when saving the existing prefix.
11639 * diagnostic.c (diagnostic_append_note): Likewise.
11640 * langhooks.c (lhd_print_error_function): Likewise.
11641 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
11642 param's type. Free the existing prefix.
11643 (pp_take_prefix): New function.
11644 (pretty_printer::pretty_printer): Drop the prefix parameter.
11645 Rename the length parameter to match the comment.
11646 (pretty_printer::~pretty_printer): Free the prefix.
11647 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
11649 (struct pretty_printer): Drop the "const" from "prefix" field's
11650 type and clarify memory management.
11651 (pp_set_prefix): Drop the "const" from the 2nd param.
11652 (pp_take_prefix): New decl.
11654 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
11656 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
11657 for word_mode_ok here instead of passing as argument.
11658 (expand_block_compare): Change select_block_compare_mode() call.
11659 (expand_strncmp_gpr_sequence): New function.
11660 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
11662 2018-08-02 Jeff Law <law@redhat.com>
11665 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11666 Define to speculation_safe_value_not_needed.
11669 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11670 Define to speculation_safe_value_not_needed.
11672 2018-08-02 Tom de Vries <tdevries@suse.de>
11675 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
11676 function. Return UI_TARGET unconditionally.
11677 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
11678 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
11680 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
11682 * genemit.c (print_overload_test): Fix typo.
11684 2018-08-02 Richard Biener <rguenther@suse.de>
11686 PR tree-optimization/86816
11687 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
11688 which checks for value availability before querying it.
11689 (gvn_uses_equal): Use it.
11690 (same_succ_hash): Likewise.
11691 (gimple_equal_p): Likewise.
11693 2018-08-02 Nick Clifton <nickc@redhat.com>
11696 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11697 Define to speculation_safe_value_not_needed.
11700 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11701 Define to speculation_safe_value_not_needed.
11704 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11705 Define to speculation_safe_value_not_needed.
11708 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11709 Define to speculation_safe_value_not_needed.
11712 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11713 Define to speculation_safe_value_not_needed.
11716 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11717 Define to speculation_safe_value_not_needed.
11720 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11721 Define to speculation_safe_value_not_needed.
11724 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
11725 Define to speculation_safe_value_not_needed.
11728 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
11729 speculation_safe_value_not_needed.
11732 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
11733 to speculation_safe_value_not_needed.
11735 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
11737 * doc/md.texi: Expand the documentation of instruction names
11738 to mention port-local uses. Document '@' in pattern names.
11739 * read-md.h (overloaded_instance, overloaded_name): New structs.
11740 (mapping): Declare.
11741 (md_reader::handle_overloaded_name): New member function.
11742 (md_reader::get_overloads): Likewise.
11743 (md_reader::m_first_overload): New member variable.
11744 (md_reader::m_next_overload_ptr): Likewise.
11745 (md_reader::m_overloads_htab): Likewise.
11746 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
11747 m_next_overload_ptr and m_overloads_htab.
11748 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
11749 (get_mode_token, get_code_token, get_int_token): New functions.
11750 (map_attr_string): Add an optional argument that passes back
11751 the associated iterator.
11752 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
11753 (md_reader::handle_overloaded_name, add_overload_instance): New
11755 (apply_iterators): Handle '@' names. Report an error if '@'
11756 is used without iterators.
11757 (initialize_iterators): Initialize the new iterator_group fields.
11758 * genopinit.c (handle_overloaded_code_for)
11759 (handle_overloaded_gen): New functions.
11760 (main): Use them to print declarations of maybe_code_for_* and
11761 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
11762 * genemit.c (print_overload_arguments, print_overload_test)
11763 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
11764 (main): Use it to print definitions of maybe_code_for_* and
11765 maybe_gen_* functions.
11766 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
11767 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
11768 instead of explicit mode checks.
11769 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
11770 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
11771 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
11772 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
11773 (aarch64_expand_compare_and_swap): Likewise
11774 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
11775 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
11776 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
11777 (aarch64_constant_pool_reload_icode): Delete.
11778 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
11779 instead of aarch64_constant_pool_reload_icode. Use
11780 code_for_aarch64_reload_mov instead of explicit mode checks.
11781 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
11782 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
11783 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
11784 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
11785 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
11786 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
11787 (aarch64_atomic_load_op_code): Delete.
11788 (aarch64_emit_atomic_load_op): Likewise.
11789 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
11790 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
11791 instead of aarch64_emit_atomic_load_op.
11792 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
11793 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
11794 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
11795 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
11796 character before the pattern name.
11797 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
11798 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
11799 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
11800 (aarch64_frecps<mode>): Likewise.
11801 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
11802 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
11803 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
11804 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
11805 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
11807 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
11809 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
11810 Allow HFmode constants if TARGET_FP_F16INST.
11812 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
11815 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
11816 No longer check last store for clobber of address register.
11818 2018-08-02 Martin Liska <mliska@suse.cz>
11820 PR gcov-profile/86817
11821 * gcov.c (process_all_functions): New function.
11823 (process_file): Move functions processing to
11824 process_all_functions.
11826 2018-08-02 David Malcolm <dmalcolm@redhat.com>
11828 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
11829 "const" to the "gimple *" and "rtx_insn *" parameters.
11830 * dumpfile.h (dump_user_location_t::dump_user_location_t):
11832 (dump_location_t::dump_location_t): Likewise.
11834 2018-08-01 Martin Sebor <msebor@redhat.com>
11836 PR tree-optimization/86650
11837 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
11838 rather than a "gcall *". Directly pass the data of interest
11839 to percent_K_format, rather than building a temporary CALL_EXPR
11841 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
11842 (gimple_fold_builtin_strncat): Adjust.
11843 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
11844 gcall* argument with gimple*.
11845 * gimple-ssa-warn-restrict.c (check_call): Same.
11846 (wrestrict_dom_walker::before_dom_children): Same.
11847 (builtin_access::builtin_access): Same.
11848 (check_bounds_or_overlap): Same
11849 (maybe_diag_overlap): Same.
11850 (maybe_diag_offset_bounds): Same.
11851 * tree-diagnostic.c (default_tree_printer): Move usage of
11852 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
11854 * tree-pretty-print.c (percent_K_format): Add argument.
11855 * tree-pretty-print.h: Add argument.
11856 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
11857 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
11858 (maybe_diag_stxncpy_trunc): Same.
11859 (handle_builtin_stxncpy): Same.
11860 (handle_builtin_strcat): Same.
11862 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
11864 * match.pd: Optimise pointer range checks.
11866 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
11868 PR tree-optimization/86758
11869 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
11870 to remove pattern statements.
11872 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
11874 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
11875 result of dfs_enumerate_from when constructing stmt_vec_infos,
11876 instead of additionally calling get_loop_body.
11878 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
11880 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
11882 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
11883 When creating an iv, assert that the step is not known to be zero.
11884 (vect_setup_realignment): Update call accordingly.
11885 * tree-vect-stmts.c (vectorizable_store): Likewise.
11886 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
11888 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
11890 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
11891 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
11892 (vectorizable_reduction): Likewise.
11893 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11894 (vect_detect_hybrid_slp_stmts): Likewise.
11895 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11897 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
11899 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
11900 (wide_int_set_zero_nonzero_bits): ...this.
11901 (zero_nonzero_bits_from_vr): Rename to...
11902 (vrp_set_zero_nonzero_bits): ...this.
11903 (extract_range_from_multiplicative_op_1): Abstract wide int
11905 (wide_int_range_multiplicative_op): ...here.
11906 (extract_range_from_binary_expr_1): Extract wide int binary
11907 operations into their own functions.
11908 (wide_int_range_lshift): New.
11909 (wide_int_range_can_optimize_bit_op): New.
11910 (wide_int_range_shift_undefined_p): New.
11911 (wide_int_range_bit_xor): New.
11912 (wide_int_range_bit_ior): New.
11913 (wide_int_range_bit_and): New.
11914 (wide_int_range_trunc_mod): New.
11915 (extract_range_into_wide_ints): New.
11916 (vrp_shift_undefined_p): New.
11917 (extract_range_from_multiplicative_op): New.
11918 (vrp_can_optimize_bit_op): New.
11919 * tree-vrp.h (value_range::dump): New.
11920 (wide_int_range_multiplicative_op): New.
11921 (wide_int_range_lshift):New.
11922 (wide_int_range_shift_undefined_p): New.
11923 (wide_int_range_bit_xor): New.
11924 (wide_int_range_bit_ior): New.
11925 (wide_int_range_bit_and): New.
11926 (wide_int_range_trunc_mod): New.
11927 (zero_nonzero_bits_from_bounds): Rename to...
11928 (wide_int_set_zero_nonzero_bits): ...this.
11929 (zero_nonzero_bits_from_vr): Rename to...
11930 (vrp_set_zero_nonzero_bits): ...this.
11931 (range_easy_mask_min_max): Rename to...
11932 (wide_int_range_can_optimize_bit_op): this.
11933 * vr-values.c (simplify_bit_ops_using_ranges): Rename
11934 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
11936 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
11938 * tree-vectorizer.h (vect_orig_stmt): New function.
11939 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
11940 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
11941 (vect_create_epilog_for_reduction): Likewise.
11942 (vectorizable_live_operation): Likewise.
11943 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
11944 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
11945 * tree-vect-stmts.c (vectorizable_call): Likewise.
11946 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
11948 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
11950 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
11952 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
11953 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
11954 (vect_transform_loop): Likewise.
11955 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
11957 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
11959 * tree-vectorizer.h (vect_schedule_slp): Return void.
11960 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
11961 (vect_schedule_slp): Likewise.
11963 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
11965 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
11967 (vect_transform_loop): Update calls accordingly. Schedule SLP
11968 instances before the main loop, if any exist.
11970 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
11972 PR tree-optimization/86749
11973 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
11974 If the lhs is used in a COND_EXPR, check that it is being used
11975 as the "then" or "else" value.
11977 2018-08-01 Tom de Vries <tdevries@suse.de>
11980 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
11981 speculation_safe_value_not_needed.
11983 2018-08-01 Richard Biener <rguenther@suse.de>
11985 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
11986 as base and offset.
11988 2018-08-01 Martin Liska <mliska@suse.cz>
11990 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
11991 format how successful transformation is dumped.
11992 (gimple_mod_pow2_value_transform): Likewise.
11993 (gimple_mod_subtract_transform): Likewise.
11994 (gimple_stringops_transform): Likewise.
11996 2018-08-01 Martin Liska <mliska@suse.cz>
11998 PR value-prof/35543
11999 * value-prof.c (interesting_stringop_to_profile_p):
12000 Simplify the code and add BUILT_IN_MEMMOVE.
12001 (gimple_stringops_transform): Likewise.
12003 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
12005 * config/aarch64/aarch64-simd.md
12006 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
12007 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
12008 use GPI iterator instead of SI mode.
12010 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
12012 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
12013 rs6000_speculation_barrier.
12014 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
12015 new barrier pattern name.
12017 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
12019 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
12020 (speculation_barrier): New insn.
12022 2018-08-01 Richard Biener <rguenther@suse.de>
12025 * graphite.h: Include isl/id.h and isl/space.h to allow build
12028 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
12031 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
12032 mode for COFF targets.
12033 * defaults.h (TARGET_COFF): Define.
12034 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
12035 TARGET_COFF): Define.
12036 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
12037 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
12038 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
12040 2018-07-31 Alexandre Oliva <oliva@adacore.com>
12042 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
12043 * gimple-streamer-out.c (output_bb): Save it.
12044 * lto-streamer-in.c (input_struct_function_base): Restore
12045 instance discriminator if available. Create map on demand.
12046 * lto-streamer-out.c (output_struct_function_base): Save it if
12048 * final.c (decl_to_instance_map): Document LTO strategy.
12050 2018-07-31 Alexandre Oliva <oliva@adacore.com>
12051 Olivier Hainque <hainque@adacore.com>
12053 * debug.h (decl_to_instance_map_t): New type.
12054 (decl_to_instance_map): Declare.
12055 (maybe_create_decl_to_instance_map): New inline function.
12056 * final.c (bb_discriminator, last_bb_discriminator): New statics,
12057 to track basic block discriminators.
12058 (final_start_function_1): Initialize them.
12059 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
12061 (decl_to_instance_map): New variable.
12062 (map_decl_to_instance, maybe_set_discriminator): New functions.
12063 (notice_source_line): Set discriminator.
12065 2018-07-31 Ian Lance Taylor <iant@golang.org>
12067 * targhooks.c (default_have_speculation_safe_value): Add
12070 2018-07-31 David Malcolm <dmalcolm@redhat.com>
12072 * dump-context.h: Include "pretty-print.h".
12073 (dump_context::refresh_dumps_are_enabled): New decl.
12074 (dump_context::emit_item): New decl.
12075 (class dump_context): Add fields "m_test_pp" and
12077 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
12078 (temp_dump_context::get_dumped_text): New decl.
12079 (class temp_dump_context): Add field "m_pp".
12080 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
12081 (dump_context::refresh_dumps_are_enabled): ...and add a test for
12083 (set_dump_file): Update for above change.
12084 (set_alt_dump_file): Likewise.
12085 (dump_loc): New overload, taking a pretty_printer *.
12086 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
12087 to any test pretty-printer.
12088 (make_item_for_dump_gimple_stmt): New function, adapted from
12089 optinfo::add_gimple_stmt.
12090 (dump_context::dump_gimple_stmt): Call it, and use the result,
12091 eliminating the direct usage of dump_file and alt_dump_file in
12092 favor of indirectly using them via emit_item.
12093 (make_item_for_dump_gimple_expr): New function, adapted from
12094 optinfo::add_gimple_expr.
12095 (dump_context::dump_gimple_expr): Call it, and use the result,
12096 eliminating the direct usage of dump_file and alt_dump_file in
12097 favor of indirectly using them via emit_item.
12098 (make_item_for_dump_generic_expr): New function, adapted from
12100 (dump_context::dump_generic_expr): Call it, and use the result,
12101 eliminating the direct usage of dump_file and alt_dump_file in
12102 favor of indirectly using them via emit_item.
12103 (make_item_for_dump_printf_va): New function, adapted from
12104 optinfo::add_printf_va.
12105 (make_item_for_dump_printf): New function.
12106 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
12107 and use the result, eliminating the direct usage of dump_file and
12108 alt_dump_file in favor of indirectly using them via emit_item.
12109 (make_item_for_dump_dec): New function.
12110 (dump_context::dump_dec): Call it, and use the result,
12111 eliminating the direct usage of dump_file and alt_dump_file in
12112 favor of indirectly using them via emit_item.
12113 (make_item_for_dump_symtab_node): New function, adapted from
12114 optinfo::add_symtab_node.
12115 (dump_context::dump_symtab_node): Call it, and use the result,
12116 eliminating the direct usage of dump_file and alt_dump_file in
12117 favor of indirectly using them via emit_item.
12118 (dump_context::begin_scope): Reimplement, avoiding direct usage
12119 of dump_file and alt_dump_file in favor of indirectly using them
12121 (dump_context::emit_item): New member function.
12122 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
12123 Set up test pretty-printer on the underlying context. Call
12124 refresh_dumps_are_enabled.
12125 (temp_dump_context::~temp_dump_context): Call
12126 refresh_dumps_are_enabled.
12127 (temp_dump_context::get_dumped_text): New member function.
12128 (selftest::verify_dumped_text): New function.
12129 (ASSERT_DUMPED_TEXT_EQ): New macro.
12130 (selftest::test_capture_of_dump_calls): Run all tests twice, with
12131 and then without optinfo enabled. Add uses of
12132 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
12133 * dumpfile.h: Update comment for the dump_* API.
12134 * optinfo-emit-json.cc
12135 (selftest::test_building_json_from_dump_calls): Update for new
12136 param for temp_dump_context ctor.
12137 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
12138 and "m_owned" field.
12139 (optinfo_item::~optinfo_item): Likewise.
12140 (optinfo::add_item): New member function.
12141 (optinfo::emit): Update comment.
12142 (optinfo::add_string): Delete.
12143 (optinfo::add_printf): Delete.
12144 (optinfo::add_printf_va): Delete.
12145 (optinfo::add_gimple_stmt): Delete.
12146 (optinfo::add_gimple_expr): Delete.
12147 (optinfo::add_tree): Delete.
12148 (optinfo::add_symtab_node): Delete.
12149 (optinfo::add_dec): Delete.
12150 * optinfo.h (class dump_context): New forward decl.
12151 (optinfo::add_item): New decl.
12152 (optinfo::add_string): Delete.
12153 (optinfo::add_printf): Delete.
12154 (optinfo::add_printf_va): Delete.
12155 (optinfo::add_gimple_stmt): Delete.
12156 (optinfo::add_gimple_expr): Delete.
12157 (optinfo::add_tree): Delete.
12158 (optinfo::add_symtab_node): Delete.
12159 (optinfo::add_dec): Delete.
12160 (optinfo::add_poly_int): Delete.
12161 (optinfo_item::optinfo_item): Remove "owned" param.
12162 (class optinfo_item): Remove field "m_owned".
12164 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12166 PR middle-end/86705
12167 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
12168 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
12169 requested variable alignment.
12170 (expand_one_ssa_partition): Likewise.
12171 (expand_one_var): Likewise.
12173 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
12175 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
12176 to speculation_safe_value_not_needed.
12178 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
12180 * targhooks.h (speculation_safe_value_not_needed): New prototype.
12181 * targhooks.c (speculation_safe_value_not_needed): New function.
12182 * target.def (have_speculation_safe_value): Update documentation.
12183 * doc/tm.texi: Regenerated.
12185 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
12187 * config/aarch64/iterators.md (ALLI_TI): New iterator.
12188 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
12190 (despeculate_copy<ALLI:mode>_insn): New insn.
12191 (despeculate_copyti_insn): New insn.
12192 (despeculate_simple<ALLI:mode>): New insn
12193 (despeculate_simpleti): New insn.
12194 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
12196 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
12197 aarch64_speculation_safe_value.
12198 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
12200 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
12202 * config/aarch64/aarch64-speculation.cc: New file.
12203 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
12204 before pass_reorder_blocks.
12205 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
12207 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
12208 X14 and X15 when tracking speculation.
12209 * config/aarch64/aarch64.md (register name constants): Add
12210 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
12211 (unspec): Add UNSPEC_SPECULATION_TRACKER.
12212 (speculation_barrier): New insn attribute.
12213 (cmp<mode>): Allow SP in comparisons.
12214 (speculation_tracker): New insn.
12215 (speculation_barrier): Add speculation_barrier attribute.
12216 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
12217 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
12218 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
12220 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
12222 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
12223 aarch64_track_speculation is true.
12224 (tb<optab><mode>1): Likewise.
12225 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
12226 generate CB[N]Z when tracking speculation.
12227 (aarch64_split_compare_and_swap): Likewise.
12228 (aarch64_split_atomic_op): Likewise.
12230 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
12232 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
12234 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
12236 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
12237 (speculation_barrier): New insn.
12239 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
12241 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
12242 * config/arm/arm.md (speculation_barrier): New expand.
12243 (speculation_barrier_insn): New pattern.
12245 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
12247 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
12248 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
12249 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
12250 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
12252 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
12253 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
12254 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
12255 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
12256 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
12257 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
12258 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
12259 * builtins.c (expand_speculation_safe_value): New function.
12260 (expand_builtin): Call it.
12261 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
12262 * doc/extend.texi: Document __builtin_speculation_safe_value.
12263 * doc/md.texi: Document "speculation_barrier" pattern.
12264 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
12265 TARGET_HAVE_SPECULATION_SAFE_VALUE.
12266 * doc/tm.texi: Regenerated.
12267 * target.def (have_speculation_safe_value, speculation_safe_value): New
12269 * targhooks.c (default_have_speculation_safe_value): New function.
12270 (default_speculation_safe_value): New function.
12271 * targhooks.h (default_have_speculation_safe_value): Add prototype.
12272 (default_speculation_safe_value): Add prototype.
12274 2018-07-31 David Malcolm <dmalcolm@redhat.com>
12276 * dump-context.h (dump_context::dump_loc): New decl.
12277 * dumpfile.c (dump_context::dump_loc): New member function.
12278 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
12279 and dump_gimple_stmt.
12280 (dump_context::dump_gimple_expr_loc): Likewise, using
12282 (dump_context::dump_generic_expr_loc): Likewise, using
12284 (dump_context::dump_printf_loc_va): Likewise, using
12286 (dump_context::begin_scope): Explicitly using the global function
12287 "dump_loc", rather than the member function.
12289 2018-07-31 Martin Sebor <msebor@redhat.com>
12291 PR tree-optimization/86741
12292 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
12294 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
12296 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
12297 depend on whether prefetch instructions will be emitted or not.
12298 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
12299 will be emitted or not.
12300 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
12301 (TARGET_SETMEM_PFD): New macros.
12303 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12305 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
12306 (NULL_STMT_VEC_INFO): Delete.
12307 (stmt_vec_info::operator*): Likewise.
12308 (stmt_vec_info::operator gimple *): Likewise.
12309 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
12310 of NULL_STMT_VEC_INFO.
12311 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
12312 (vect_reassociating_reduction_p): Likewise.
12313 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
12314 (vectorizable_store): Likewise.
12315 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
12316 (vec_info::free_stmt_vec_infos): Likewise.
12318 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12320 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
12321 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
12323 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12325 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
12326 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
12327 (vec_info::free_stmt_vec_info): New private member functions.
12328 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
12329 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
12330 * tree-parloops.c (gather_scalar_reductions): Remove calls to
12331 set_stmt_vec_info_vec and free_stmt_vec_infos.
12332 * tree-vect-loop.c (_loop_vec_info): Remove call to
12333 set_stmt_vec_info_vec.
12334 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
12335 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
12336 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
12337 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
12338 (vec_info::free_stmt_vec_info): ...these new functions. Remove
12339 assignments in {vec_info::,}new_stmt_vec_info that are redundant
12340 with the clearing in the xcalloc.
12341 (stmt_vec_info_vec): Delete.
12342 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
12343 (vectorize_loops): Likewise.
12344 (vec_info::~vec_info): Remove argument from call to
12345 free_stmt_vec_infos.
12346 (vec_info::add_stmt): Remove vinfo argument from call to
12349 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12351 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
12352 rather than a gimple stmt.
12353 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
12354 information for pattern statements when passed the original
12355 statement; instead wait to be passed the pattern statement itself.
12356 Don't call set_vinfo_for_stmt here.
12357 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
12358 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
12359 stmt_vec_infos here.
12360 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
12361 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
12362 stmt_vec_infos entry.
12364 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12366 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
12367 * tree-vectorizer.c (vec_info::replace_stmt): New function.
12368 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
12369 * tree-vect-stmts.c (vectorizable_call): Likewise.
12370 (vectorizable_simd_clone_call): Likewise.
12372 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12374 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
12375 * tree-vectorizer.c (vec_info::remove_stmt): New function.
12376 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
12377 * tree-vect-loop.c (vect_transform_loop): Likewise.
12378 * tree-vect-slp.c (vect_schedule_slp): Likewise.
12379 * tree-vect-stmts.c (vect_remove_stores): Likewise.
12381 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12383 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
12384 (vect_dr_stmt): Delete.
12385 * tree-vectorizer.c (vec_info::lookup_dr): New function.
12386 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
12388 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12389 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
12390 (vect_verify_datarefs_alignment, vect_peeling_supportable)
12391 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
12392 (vect_analyze_data_refs): Likewise.
12393 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
12395 (vect_find_same_alignment_drs): Likewise.
12396 (vect_slp_analyze_node_dependences): Update calls accordingly.
12397 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
12398 instead of DR_VECT_AUX.
12399 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
12400 of a vector data references. Use vec_info::lookup_dr instead of
12402 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
12403 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
12404 instead of DR_VECT_AUX.
12406 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12408 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
12410 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
12412 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
12413 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
12414 (vect_gen_prolog_loop_niters): Likewise.
12416 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12418 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
12419 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
12420 (vect_known_alignment_in_bytes, vect_dr_behavior)
12421 (vect_get_scalar_dr_size): Take references as dr_vec_infos
12422 instead of data_references. Update calls to other routines for
12423 which the same change has been made.
12424 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
12425 dr_vec_infos instead of stmt_vec_infos.
12426 (vect_analyze_data_ref_dependence): Update call accordingly.
12427 (vect_slp_analyze_data_ref_dependence)
12428 (vect_record_base_alignments): Use DR_VECT_AUX.
12429 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
12430 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
12431 (vector_alignment_reachable_p, vect_get_data_access_cost)
12432 (vect_peeling_supportable, vect_analyze_group_access_1)
12433 (vect_analyze_group_access, vect_analyze_data_ref_access)
12434 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
12435 (vect_compile_time_alias, vect_small_gap_p)
12436 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
12437 (vect_supportable_dr_alignment): Take references as dr_vec_infos
12438 instead of data_references. Update calls to other routines for
12439 which the same change has been made.
12440 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
12441 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
12442 (vect_slp_analyze_and_verify_node_alignment)
12443 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
12444 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
12445 (vect_setup_realignment): Use dr_vec_infos. Update calls after
12447 (_vect_peel_info::dr): Replace with...
12448 (_vect_peel_info::dr_info): ...this new field.
12449 (vect_peeling_hash_get_most_frequent)
12450 (vect_peeling_hash_choose_best_peeling): Update accordingly.
12451 (vect_peeling_hash_get_lowest_cost):
12452 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
12453 routines for which the same change has been made.
12454 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
12456 * tree-vect-loop-manip.c (get_misalign_in_elems)
12457 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
12459 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
12460 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
12461 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
12462 (get_group_load_store_type, get_negative_load_store_type)
12463 (vect_get_data_ptr_increment, vectorizable_store)
12464 (vectorizable_load): Likewise.
12465 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
12466 Update calls to other routines for which the same change has been made.
12468 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12470 * tree-vectorizer.h (vec_info::move_dr): New member function.
12471 (dataref_aux): Rename to...
12472 (dr_vec_info): ...this and add "dr" and "stmt" fields.
12473 (_stmt_vec_info::dr_aux): Update accordingly.
12474 (_stmt_vec_info::data_ref_info): Delete.
12475 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
12476 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
12477 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
12478 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
12480 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
12481 (STMT_VINFO_DR_INFO): New macro.
12482 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
12483 (set_dr_misalignment): Update after rename of dataref_aux.
12484 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
12485 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
12486 initialization of STMT_VINFO_DATA_REF.
12487 * tree-vectorizer.c (vec_info::move_dr): New function.
12488 * tree-vect-patterns.c (vect_recog_bool_pattern)
12489 (vect_recog_mask_conversion_pattern)
12490 (vect_recog_gather_scatter_pattern): Use it.
12491 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
12492 the "dr" and "stmt" fields of dr_vec_info instead of
12493 STMT_VINFO_DATA_REF.
12495 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12497 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
12498 (is_pattern_stmt_p): Use it.
12499 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
12500 on pattern statements.
12502 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12504 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
12505 original stmt as a stmt_vec_info rather than a gimple stmt.
12506 (vect_pattern_recog_1): Take the statement directly as a
12507 stmt_vec_info, rather than via a gimple_stmt_iterator.
12508 Update call to vect_mark_pattern_stmts.
12509 (vect_pattern_recog): Update calls accordingly.
12511 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12513 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
12514 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
12515 a vect_def_type for the first argument.
12516 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
12517 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
12518 operand if it isn't defined by a vectorized statement.
12519 (vect_build_gather_load_calls): Remove the mask_dt argument and
12520 update calls to vect_get_vec_def_for_stmt_copy.
12521 (vectorizable_bswap): Likewise the dt argument.
12522 (vectorizable_call): Update calls to vectorizable_bswap and
12523 vect_get_vec_def_for_stmt_copy.
12524 (vectorizable_simd_clone_call, vectorizable_assignment)
12525 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
12526 (vectorizable_comparison): Update calls to
12527 vect_get_vec_def_for_stmt_copy.
12528 (vectorizable_store): Likewise. Remove now-unnecessary calls to
12529 vect_is_simple_use.
12530 (vect_get_loop_based_defs): Remove dt argument and update call
12531 to vect_get_vec_def_for_stmt_copy.
12532 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
12533 and vect_get_vec_def_for_stmt_copy.
12534 (vectorizable_load): Update calls to vect_build_gather_load_calls
12535 and vect_get_vec_def_for_stmt_copy.
12536 * tree-vect-loop.c (vect_create_epilog_for_reduction)
12537 (vectorizable_reduction, vectorizable_live_operation): Update calls
12538 to vect_get_vec_def_for_stmt_copy.
12540 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12542 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
12543 and gimple stmt arguments with a stmt_vec_info.
12544 (vect_record_base_alignments): Update calls accordingly.
12545 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
12546 and gimple stmt arguments with a stmt_vec_info.
12547 (vect_build_slp_tree_1): Remove vinfo argument and update call
12548 to vect_record_max_nunits.
12549 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
12550 and vect_record_max_nunits.
12552 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12554 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
12555 file and take a stmt_vec_info instead of a gimple stmt.
12556 (supportable_widening_operation, vect_finish_replace_stmt)
12557 (vect_finish_stmt_generation, vect_get_store_rhs)
12558 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
12559 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
12560 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
12561 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
12562 (vect_create_data_ref_ptr, bump_vector_ptr)
12563 (vect_permute_store_chain, vect_setup_realignment)
12564 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
12565 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
12566 (vectorizable_reduction, vectorizable_induction)
12567 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
12568 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
12569 than gimple stmts as arguments.
12570 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
12571 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
12572 (can_group_stmts_p, vect_check_gather_scatter)
12573 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
12574 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
12575 (vect_permute_load_chain, vect_shift_permute_load_chain)
12576 (vect_transform_grouped_load)
12577 (vect_record_grouped_load_vectors): Likewise.
12578 * tree-vect-loop.c (vect_fixup_reduc_chain)
12579 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
12580 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
12581 (vectorizable_reduction, vectorizable_induction)
12582 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
12583 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
12584 (vect_get_load_store_mask): Likewise.
12585 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
12586 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
12587 * tree-vect-stmts.c (vect_mark_relevant)
12588 (is_simple_and_all_uses_invariant)
12589 (exist_non_indexing_operands_for_use_p, process_use)
12590 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
12591 (vect_get_vec_def_for_operand, vect_get_vec_defs)
12592 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
12593 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
12594 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
12595 (get_negative_load_store_type, get_load_store_type)
12596 (vect_check_load_store_mask, vect_check_store_rhs)
12597 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
12598 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
12599 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
12600 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
12601 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
12602 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
12603 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
12604 (supportable_widening_operation): Likewise.
12606 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12608 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
12609 a stmt_vec_info instead of a gcall.
12610 (vect_check_gather_scatter): Update call accordingly.
12611 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
12613 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
12614 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
12615 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
12616 instead of a gimple stmt.
12617 (vect_transform_loop): Update calls accordingly.
12618 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
12619 stmt_vec_infos instead of gimple stmts.
12620 (vect_analyze_slp_instance): Update use accordingly.
12621 * tree-vect-stmts.c (read_vector_array, write_vector_array)
12622 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
12623 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
12624 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
12625 (vect_gen_widened_results_half, vect_get_loop_based_defs)
12626 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
12627 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
12628 down to subroutines.
12630 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12632 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
12633 of the worklist from a vector of gimple stmts to a vector of
12635 * tree-vect-stmts.c (vect_mark_relevant, process_use)
12636 (vect_mark_stmts_to_be_vectorized): Likewise
12638 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12640 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
12641 statement before passing it to vect_analyze_stmt.
12642 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
12643 the chain of phi vector definitions. Track the exit phi via its
12645 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
12646 STMT_VINFO_REDUC_DEF.
12647 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
12648 stmt_vec_infos to handle the statement chains.
12649 (vect_get_slp_defs): Record the first statement in the node
12650 using a stmt_vec_info.
12651 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
12652 statements here and pass their stmt_vec_info down to subroutines.
12653 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
12654 down to vect_finish_stmt_generation.
12655 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
12656 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
12657 stmt_vec_infos to subroutines.
12658 (vect_remove_stores): Use stmt_vec_infos to handle the statement
12661 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12663 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
12664 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
12665 (vect_permute_store_chain, vect_setup_realignment)
12666 (vect_permute_load_chain, vect_shift_permute_load_chain)
12667 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
12668 stmts internally, and when passing values to other vectorizer routines.
12669 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
12670 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
12671 (vect_analyze_loop_operations, get_initial_def_for_reduction)
12672 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
12673 (vectorizable_reduction, vectorizable_induction)
12674 (vectorizable_live_operation, vect_transform_loop_stmt)
12675 (vect_transform_loop): Likewise.
12676 * tree-vect-patterns.c (vect_reassociating_reduction_p)
12677 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
12678 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
12679 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12680 (vect_slp_analyze_node_operations_1): Likewise.
12681 * tree-vect-stmts.c (vect_mark_relevant, process_use)
12682 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
12683 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
12684 (vect_finish_stmt_generation_1, get_group_load_store_type)
12685 (get_load_store_type, vect_build_gather_load_calls)
12686 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
12687 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
12688 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
12689 (vectorizable_store, vectorizable_load, vectorizable_condition)
12690 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
12691 (supportable_widening_operation): Likewise.
12692 (vect_get_vector_types_for_stmt): Likewise.
12693 * tree-vectorizer.h (vect_dr_behavior): Likewise.
12695 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12697 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12698 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
12699 (vect_permute_store_chain, vect_permute_load_chain)
12700 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
12701 repeated stmt_vec_info lookups.
12702 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
12703 (vect_update_ivs_after_vectorizer): Likewise.
12704 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
12705 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
12706 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
12707 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12708 (vect_bb_slp_scalar_cost): Likewise.
12709 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
12711 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12713 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
12714 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
12715 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
12716 to get gassigns and gcalls, rather than operating on generc gimple
12718 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
12719 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
12720 (vectorizable_load, vect_analyze_stmt): Likewise.
12721 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
12723 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12725 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
12726 return stmt_vec_infos rather than gimple stmts. Do not accept
12728 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
12730 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
12731 Update use of get_later_stmt.
12732 (vect_get_constant_vectors): Update call accordingly.
12733 (vect_schedule_slp_instance): Likewise
12734 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
12735 (vect_slp_analyze_instance_dependence): Likewise.
12736 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
12738 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12740 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
12741 (stmt_info_for_cost::stmt_info): ...this new field.
12742 (add_stmt_costs): Update accordingly.
12743 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
12744 (vect_get_known_peeling_cost): Likewise.
12745 (vect_estimate_min_profitable_iters): Likewise.
12746 * tree-vect-stmts.c (record_stmt_cost): Likewise.
12748 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12750 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
12751 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
12752 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
12754 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
12756 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12758 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
12759 a gimple stmt to a stmt_vec_info.
12760 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
12762 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12764 * tree-vectorizer.h (vec_info::grouped_stores): Change from
12765 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
12766 (_loop_vec_info::reduction_chains): Likewise.
12767 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
12769 * tree-vect-slp.c (vect_analyze_slp): Likewise.
12771 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12773 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
12774 a gimple stmt to a stmt_vec_info.
12775 (_stmt_vec_info::next_element): Likewise.
12776 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
12777 (vect_slp_analyze_and_verify_node_alignment)
12778 (vect_analyze_group_access_1, vect_analyze_group_access)
12779 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
12780 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
12781 (vect_supportable_dr_alignment): Update accordingly.
12782 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
12783 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
12784 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
12785 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
12786 * tree-vect-slp.c (vect_build_slp_tree_1)
12787 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
12788 (vect_split_slp_store_group, vect_analyze_slp_instance)
12789 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
12790 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
12791 (get_group_load_store_type, get_load_store_type)
12792 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
12793 (vect_transform_stmt, vect_remove_stores): Likewise.
12795 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12797 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
12798 than a gimple stmt.
12799 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12800 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
12801 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
12802 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
12803 (vector_alignment_reachable_p, vect_get_data_access_cost)
12804 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
12805 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
12806 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
12807 (vect_analyze_group_access_1, vect_analyze_group_access)
12808 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
12809 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
12810 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
12811 result of vect_dr_stmt and use the stmt_vec_info instead of
12812 the associated gimple stmt.
12813 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
12814 (vect_gen_prolog_loop_niters): Likewise.
12815 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
12817 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12819 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
12820 to a vec<stmt_vec_info>.
12821 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
12822 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
12823 vec<stmt_vec_info>.
12824 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
12825 to a vec<stmt_vec_info>.
12826 (bst_traits::value_type, bst_traits::value_type): Likewise.
12827 (bst_traits::hash): Update accordingly.
12828 (vect_get_and_check_slp_defs): Change the stmts parameter from
12829 a vec<gimple *> to a vec<stmt_vec_info>.
12830 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
12831 (vect_build_slp_tree): Likewise.
12832 (vect_build_slp_tree_2): Likewise. Update uses of
12833 SLP_TREE_SCALAR_STMTS.
12834 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
12835 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
12836 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
12837 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
12838 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
12839 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
12840 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
12841 (vect_get_constant_vectors, vect_get_slp_defs)
12842 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
12843 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
12844 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
12845 instead of gimple stmts.
12846 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
12847 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
12848 (vect_slp_analyze_instance_dependence): Update uses of
12849 SLP_TREE_SCALAR_STMTS.
12850 (vect_slp_analyze_and_verify_node_alignment): Likewise.
12851 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
12852 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
12853 (get_initial_defs_for_reduction): Likewise.
12854 (vect_create_epilog_for_reduction): Likewise.
12855 (vectorize_fold_left_reduction): Likewise.
12856 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
12857 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
12858 (can_vectorize_live_stmts): Likewise.
12860 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12862 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
12863 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
12864 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
12866 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
12868 (gather_scalar_reductions): Update after above interface changes.
12869 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
12870 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
12872 (vect_force_simple_reduction): Likewise.
12873 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
12874 LOOP_VINFO_REDUCTIONS.
12875 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
12877 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12879 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
12880 a gimple stmt to a stmt_vec_info.
12881 * tree-vect-loop.c (vect_active_double_reduction_p)
12882 (vect_force_simple_reduction, vectorizable_reduction): Update
12885 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12887 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
12888 vec<gimple *> to a vec<stmt_vec_info>.
12889 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
12890 the reduction_phis argument from a vec<gimple *> to a
12891 vec<stmt_vec_info>.
12892 (vectorizable_reduction): Likewise the phis local variable that
12893 is passed to vect_create_epilog_for_reduction. Update for new type
12894 of SLP_TREE_VEC_STMTS.
12895 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
12896 (vectorizable_live_operation): Likewise.
12897 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
12898 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
12900 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12902 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
12903 a gimple stmt to a stmt_vec_info.
12904 (vectorizable_condition, vectorizable_live_operation)
12905 (vectorizable_reduction, vectorizable_induction): Pass back the
12906 vectorized statement as a stmt_vec_info.
12907 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
12908 use of STMT_VINFO_VEC_STMT.
12909 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
12910 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
12911 as stmt_vec_infos rather than gimple stmts.
12912 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
12913 to a stmt_vec_info.
12914 (vectorizable_live_operation): Likewise.
12915 (vectorizable_reduction, vectorizable_induction): Likewise,
12916 updating use of STMT_VINFO_VEC_STMT.
12917 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
12918 of STMT_VINFO_VEC_STMT.
12919 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
12920 (vectorizable_simd_clone_call, vectorizable_conversion)
12921 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
12922 (vectorizable_store, vectorizable_load, vectorizable_condition)
12923 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
12924 from a gimple stmt to a stmt_vec_info.
12925 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
12926 pointer to a stmt_vec_info to the vectorizable_* routines.
12928 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12930 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
12931 a gimple stmt to a stmt_vec_info.
12932 (is_pattern_stmt_p): Update accordingly.
12933 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
12934 (vect_record_grouped_load_vectors): Likewise.
12935 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
12936 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
12937 (vect_model_reduction_cost): Likewise.
12938 (vect_create_epilog_for_reduction): Likewise.
12939 (vectorizable_reduction, vectorizable_induction): Likewise.
12940 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
12941 Return the stmt_vec_info for the pattern statement.
12942 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
12943 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
12944 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
12945 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
12946 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
12947 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
12948 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
12949 (free_stmt_vec_info, vect_is_simple_use): Likewise.
12951 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12953 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
12954 (vect_finish_stmt_generation): Likewise.
12955 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
12956 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
12957 (vect_build_gather_load_calls): Use the return value of the above
12958 functions instead of a separate call to vinfo_for_stmt. Use narrow
12959 scopes for the input gimple stmt and wider scopes for the associated
12960 stmt_vec_info. Use vec_info::lookup_def when setting these
12961 stmt_vec_infos from an SSA_NAME definition.
12962 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
12963 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
12964 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
12965 (vectorizable_store, vectorizable_load, vectorizable_condition)
12966 (vectorizable_comparison): Likewise.
12967 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
12968 (vectorizable_reduction): Likewise.
12970 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12972 * tree-vectorizer.h (vect_is_simple_use): Add an optional
12973 stmt_vec_info * parameter before the optional gimple **.
12974 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
12975 (process_use, vect_get_vec_def_for_operand_1): Update callers.
12976 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
12977 * tree-vect-loop.c (vectorizable_reduction): Likewise.
12978 (vectorizable_live_operation): Likewise.
12979 * tree-vect-patterns.c (type_conversion_p): Likewise.
12980 (vect_look_through_possible_promotion): Likewise.
12981 (vect_recog_rotate_pattern): Likewise.
12982 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
12984 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
12986 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
12987 a typedef to a wrapper class.
12988 (NULL_STMT_VEC_INFO): New macro.
12989 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
12990 (stmt_vec_info::operator*): New function.
12991 (stmt_vec_info::operator gimple *): Likewise.
12992 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
12993 (add_stmt_costs): Likewise.
12994 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
12995 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
12996 (vect_get_known_peeling_cost): Likewise.
12997 (vect_estimate_min_profitable_iters): Likewise.
12998 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
12999 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
13000 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
13001 (vectorizable_store, free_stmt_vec_infos): Likewise.
13002 (new_stmt_vec_info): Change return type of xcalloc to
13005 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
13007 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
13008 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
13009 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
13010 a single_imm_use-based sequence.
13011 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
13013 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
13015 * tree-vectorizer.h (vec_info::lookup_def): Declare.
13016 * tree-vectorizer.c (vec_info::lookup_def): New function.
13017 * tree-vect-patterns.c (vect_get_internal_def): Use it.
13018 (vect_widened_op_tree): Likewise.
13019 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
13020 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
13021 (vectorizable_reduction): Likewise.
13022 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
13024 (vect_is_slp_reduction): Update calls accordingly. Use
13025 vec_info::lookup_def.
13026 (vect_is_simple_reduction): Likewise
13027 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
13029 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
13031 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
13032 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
13033 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
13035 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
13036 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
13037 (vect_update_vf_for_slp, vect_analyze_loop_operations)
13038 (vect_is_slp_reduction, vectorizable_induction)
13039 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
13040 * tree-vect-patterns.c (vect_init_pattern_stmt):
13041 (vect_determine_min_output_precision_1, vect_determine_precisions)
13042 (vect_pattern_recog): Likewise.
13043 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
13044 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
13045 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
13046 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
13047 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
13048 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
13049 info field from a loop to a loop_vec_info.
13051 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
13053 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
13054 (vec_info::add_stmt): Declare.
13055 * tree-vectorizer.c (vec_info::add_stmt): New function.
13056 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
13057 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
13058 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
13059 (vectorizable_induction): Likewise.
13060 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
13061 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
13062 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
13063 (vectorizable_load): Likewise.
13064 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
13065 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
13066 (vect_recog_gather_scatter_pattern): Likewise.
13067 (append_pattern_def_seq): Likewise. Remove a check that is
13068 performed by add_stmt itself.
13070 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
13072 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
13073 which make_ssa_name was called with new_stmt before new_stmt
13076 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
13078 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
13080 (vect_is_slp_reduction): ...here...
13081 (vect_is_simple_reduction): ...and here. Remove repetition of tests
13082 that are already known to be false.
13084 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
13086 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
13087 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
13088 STMT_VINFO_NUM_SLP_USES when it's true.
13089 (vect_free_slp_instance): Add a final_p parameter and pass it to
13090 vect_free_slp_tree.
13091 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
13092 (vect_analyze_slp_instance): Likewise.
13093 (vect_slp_analyze_operations): Likewise.
13094 (vect_slp_analyze_bb_1): Likewise.
13095 * tree-vectorizer.c (vec_info): Likewise.
13096 * tree-vect-loop.c (vect_transform_loop): Likewise.
13098 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
13100 * tree-vect-loop.c (vectorizable_reduction): Assert that the
13101 function is not called for second and subsequent members of
13104 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
13106 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
13107 cases for nested loops from here to ...
13108 (vect_create_epilog_for_reduction): ...here. Only call
13109 vect_is_simple_use for inner-loop reductions.
13111 2018-07-31 Martin Liska <mliska@suse.cz>
13113 PR gcov-profile/85338
13114 PR gcov-profile/85350
13115 PR gcov-profile/85372
13116 * profile.c (struct location_triplet): New.
13117 (struct location_triplet_hash): Likewise.
13118 (output_location): Do not output a BB that
13119 is already recorded for a line.
13120 (branch_prob): Use streamed_locations.
13122 2018-07-31 Martin Liska <mliska@suse.cz>
13124 PR gcov-profile/85370
13125 * coverage.c (coverage_begin_function): Do not mark target
13126 clones as artificial functions.
13128 2018-07-31 Martin Liska <mliska@suse.cz>
13130 PR gcov-profile/83813
13131 PR gcov-profile/84758
13132 PR gcov-profile/85217
13133 PR gcov-profile/85332
13134 * profile.c (branch_prob): Do not record GOTO expressions
13135 for GIMPLE statements which locations are already streamed.
13137 2018-07-31 Olivier Hainque <hainque@adacore.com>
13139 * gcc.c (handle_spec_function): Accept a soft_matched_part
13140 argument, as do_spec_1. Pass it down to ...
13141 (eval_spec_function): Accept a soft_matched_part argument,
13142 and pass it down to ...
13143 (do_spec_2): Accept a soft_matched_part argument, and pass
13144 it down to do_spec_1.
13145 (do_spec_1): Pass soft_matched_part to handle_spec_function.
13146 (handle_braces): Update call to handle_spec_function.
13147 (driver::set_up_specs): Update calls to do_spec_2.
13148 (compare_debug_dump_opt_spec_function): Likewise.
13149 (compare_debug_self_opt_spec_function): Likewise.
13151 2018-07-31 Olivier Hainque <hainque@adacore.com>
13153 * common.opt (nolibc): New option.
13154 * doc/invoke.texi (Link Options): Document it.
13155 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
13156 * config/alpha/linux.h: Likewise.
13157 * config/arc/elf.h: Likewise.
13158 * config/arm/uclinux-elf.h: Likewise.
13159 * config/arm/unknown-elf.h: Likewise.
13160 * config/avr/avrlibc.h: Likewise.
13161 * config/bfin/bfin.h: Likewise.
13162 * config/bfin/linux.h: Likewise.
13163 * config/bfin/uclinux.h: Likewise.
13164 * config/darwin.h: Likewise.
13165 * config/darwin10.h: Likewise.
13166 * config/darwin12.h: Likewise.
13167 * config/gnu-user.h: Likewise.
13168 * config/lm32/uclinux-elf.h: Likewise.
13169 * config/pa/pa-hpux11.h: Likewise.
13170 * config/pa/pa64-hpux.h: Likewise.
13171 * config/sparc/sparc.h: Likewise.
13173 2018-07-31 Olivier Hainque <hainque@adacore.com>
13175 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
13176 undefined variables.
13178 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
13181 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
13182 instead of GEN_INT.
13184 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
13186 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
13187 terminated string literal.
13189 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
13191 PR rtl-optimization/85160
13192 * combine.c (is_just_move): New function.
13193 (try_combine): Allow combining two instructions into two if neither of
13194 the original instructions was a move.
13196 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
13199 * doc/extend.texi (Global Register Variables): Discourage use of type
13201 (Local Register Variables): Likewise.
13203 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
13205 PR tree-optimization/86506
13206 * hwint.h (ceil_log2): Resync with hwint.c implementation.
13208 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
13211 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
13212 hard_regno, make sure no insn between `from` and `to` clobbers it.
13214 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
13215 Tom de Vries <tdevries@suse.de>
13217 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
13218 (PTX_DEFAULT_RUNTIME_DIM): ... this.
13219 (nvptx_goacc_validate_dims): Set default worker and gang dims to
13220 PTX_DEFAULT_RUNTIME_DIM.
13221 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
13223 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
13225 * config/pa/pa.c (pa_output_addr_vec): Align address table.
13226 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
13227 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
13229 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
13231 * config/rs6000/constraints.md (wG constraint): Delete, no longer
13233 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
13234 predicate to reflect toc fusion has been deleted.
13235 (toc_fusion_mem_raw): Delete, no longer used.
13236 (toc_fusion_mem_wrapped): Likewise.
13237 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
13239 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
13240 Delete, no longer used.
13241 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
13242 meant to be used for toc fusion.
13243 (rs6000_debug_print_mode): Delete toc fusion debugging.
13244 (rs6000_debug_reg_global): Likewise.
13245 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
13246 fusion and secondary reload support that were never used.
13247 (rs6000_option_override_internal): Delete TOC fusion, that was only
13248 partially defined, and it did not work unless you also used the
13250 (rs6000_legitimate_address_p): Delete TOC fusion support.
13251 (rs6000_opt_masks): Likewise.
13252 (fusion_wrap_memory_address): Delete function, no longer used.
13253 (fusion_split_address); Delete TOC fusion support.
13254 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
13255 longer used with toc fusion being deleted.
13256 (TARGET_TOC_FUSION_FP): Likewise.
13257 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
13259 (toc fusion spliter): Delete TOC fusion support.
13260 (toc_fusionload_<mode>): Likewise.
13261 (toc_fusionload_di): Likewise.
13262 (fusion_gpr_load_<mode>): Delete generator function, this insn no
13263 longer needs to be named. Rename predicate to delete TOC fusion.
13264 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
13265 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
13266 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
13267 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
13268 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
13270 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
13272 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
13273 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
13274 __int128 in built-in function prototypes.
13275 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
13276 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
13278 2018-07-27 Martin Sebor <msebor@redhat.com>
13280 PR tree-optimization/86696
13281 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
13282 types, including enums.
13283 (handle_char_store): Be prepared for the above function to fail.
13285 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
13287 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
13288 when optimization level is lower than 2 or optimize for size.
13290 2018-07-26 Martin Sebor <msebor@redhat.com>
13292 PR tree-optimization/86043
13293 PR tree-optimization/86042
13294 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
13295 (get_string_cst_length): Rename...
13296 (get_min_string_length): ...to this. Add argument.
13297 (handle_char_store): Extend to handle multi-character stores by
13299 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
13300 * tree.h (initializer_zerop): Add argument.
13302 2018-07-26 Jakub Jelinek <jakub@redhat.com>
13304 PR middle-end/86660
13305 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
13306 declare target to variables if they have always,{to,from,tofrom} map
13309 2018-07-26 Martin Liska <mliska@suse.cz>
13312 * lto-wrapper.c: Add linker_output as prefix
13313 for ltrans_output_file.
13315 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
13317 PR rtl-optimization/85805
13318 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
13319 value for hard registers if that was written in the same mode.
13321 2018-07-26 Martin Liska <mliska@suse.cz>
13323 PR gcov-profile/86536
13324 * gcov.c (format_gcov): Use printf format %.*f directly
13325 and do not handle special values.
13327 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
13329 * common/config/arc/arc-common.c (arc_option_optimization_table):
13330 Update default optimizations for size.
13332 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
13334 * config/arc/arc.md (movsf_insn): Add short instruction selection.
13335 * config/arc/constraints.md (CfZ): New constraint.
13336 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
13337 (subsf3_fpu): Likewise.
13338 (cmpsf_fpu): Likewise.
13339 (cmpsf_fpu_uneq): Likewise.
13341 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
13343 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
13345 (arc_is_uncached_mem_p): uncached applies to both the variable and
13348 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
13350 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
13353 2018-07-25 David Malcolm <dmalcolm@redhat.com>
13355 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
13356 field "m_scopes" from vec to auto_vec.
13358 2018-07-25 Martin Liska <mliska@suse.cz>
13360 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
13363 2018-07-25 Richard Biener <rguenther@suse.de>
13366 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
13367 special wrt context_die late.
13368 (gen_subprogram_die): Re-use DIEs in local scope.
13370 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
13372 PR tree-optimization/86644
13373 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
13375 2018-07-25 Martin Liska <mliska@suse.cz>
13377 PR middle-end/86645
13378 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
13379 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
13381 2018-07-25 Martin Liska <mliska@suse.cz>
13384 * params.def: Explain ASan abbreviation and provide
13385 a documentation link.
13387 2018-07-24 Martin Sebor <msebor@redhat.com>
13389 PR tree-optimization/86622
13390 PR tree-optimization/86532
13391 * builtins.h (string_length): Declare.
13392 * builtins.c (c_strlen): Correct handling of non-constant offsets.
13393 (check_access): Be prepared for non-constant length ranges.
13394 (string_length): Make extern.
13395 * expr.c (string_constant): Only handle the minor non-constant
13396 array index. Use string_constant to compute the length of
13397 a generic string constant.
13399 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
13401 PR tree-optimization/86618
13402 * tree-vect-stmts.c (vectorizable_call): Don't take the address
13403 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
13405 2018-07-24 David Malcolm <dmalcolm@redhat.com>
13407 PR tree-optimization/86636
13408 * json.cc (json::object::set): Fix comment. Add assertions.
13409 (json::array::append): Move here from json.h. Add comment and an
13411 (json::string::string): Likewise.
13412 * json.h (json::array::append): Move to json.cc.
13413 (json::string::string): Likewise.
13414 * optinfo-emit-json.cc
13415 (optrecord_json_writer::impl_location_to_json): Assert that we
13416 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
13417 wrapper around it. Expand the location once, rather than three
13419 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
13420 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
13422 (optrecord_json_writer::optinfo_to_json): Likewise, in four
13423 places. Fix some overlong lines.
13425 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
13427 * config/aarch64/aarch64-simd.md
13428 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
13429 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
13430 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
13431 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
13432 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
13433 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
13434 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
13435 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
13436 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
13438 2018-07-24 Jakub Jelinek <jakub@redhat.com>
13440 PR middle-end/86627
13441 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
13442 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
13443 and abs_d == d, do the power of two handling if profitable.
13445 2018-07-24 Richard Biener <rguenther@suse.de>
13447 * match.pd: Add BIT_FIELD_REF canonicalizations.
13449 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
13452 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
13454 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
13456 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
13457 terminated STRING_CST object.
13459 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
13461 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
13463 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
13465 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
13466 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
13467 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
13468 the elements into a register.
13469 (rs6000_split_v4si_init_di_reg): Delete.
13470 (rs6000_split_v4si_init): Delete.
13471 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
13472 (vsx_init_v4si): Rewrite as a define_expand.
13474 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
13476 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
13477 zero_extend argument from memory): New.
13479 2018-07-22 Martin Sebor <msebor@redhat.com>
13482 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
13483 diagnosing calls with unknown arguments unless -Walloca-larger-than
13484 is restricted to less than PTRDIFF_MAX bytes.
13486 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
13488 * doc/gcov.texi (Invoking Gcov): Editorial changes.
13490 2018-07-20 David Malcolm <dmalcolm@redhat.com>
13492 * pretty-print.c (text_info::set_location): Remove redundant
13493 "line_table" parameter from call to rich_location::set_range.
13495 2018-07-20 Martin Sebor <msebor@redhat.com>
13497 PR middle-end/82063
13498 * builtins.c (expand_builtin_alloca): Adjust.
13499 * calls.c (alloc_max_size): Simplify.
13500 * cgraphunit.c (cgraph_node::expand): Adjust.
13501 * common.opt (larger_than_size, warn_frame_larger_than): Remove
13503 (frame_larger_than_size): Same.
13504 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
13505 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
13507 * doc/invoke.texi (GCC Command Options): Document option arguments.
13508 Explain byte-size arguments and suffixes.
13509 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
13510 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
13511 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
13512 * doc/options.texi (UInteger): Expand.
13513 (Host_Wide_Int, ByteSize): Document new properties.
13514 * final.c (final_start_function_1): Include sizes in an error message.
13515 * function.c (frame_offset_overflow): Same.
13516 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
13517 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
13518 Diagnose unbounded alloca calls only for limits of less than
13520 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
13521 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
13523 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
13524 only for limits of less than PTRDIFF_MAX.
13525 * langhooks-def.h (lhd_handle_option): Change function argument
13527 * langhooks.c (lhd_handle_option): Same.
13528 * langhooks.h (handle_option): Same.
13529 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
13531 (var_type, var_type_struct): Same.
13532 (var_set): Handle ByteSize flag.
13533 * optc-gen.awk: Add comments to output to ease debugging. Make
13534 use of HOST_WIDE_INT where appropriate.
13535 * opts-gen-save.awk: Use %lx to format unsigned long.
13536 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
13537 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
13538 arguments. Parse bytes-size suffixes.
13539 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
13540 (enum_value_to_arg): Same.
13541 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
13542 (handle_option): Adjust.
13543 (generate_option): Change function argument to HOST_WIDE_INT.
13544 (cmdline_handle_error): Adjust.
13545 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
13546 (set_option): Change function argument to HOST_WIDE_INT.
13547 (option_enabled): Handle cl_host_wide_int.
13548 (get_option_state): Handle CLVC_SIZE.
13549 (control_warning_option): Same.
13550 * opts.c (common_handle_option): Change function argument to
13551 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
13552 OPT_Wvla_larger_than_.
13553 * opts.h (enum cl_var_type): Add an enumerator.
13554 * stor-layout.c (layout_decl): Print a more meaningful warning.
13555 * toplev.c (output_stack_usage): Adjust.
13557 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
13559 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
13560 call to inline_expand_builtin_string_cmp.
13561 (expand_builtin_strcmp): Likewise.
13562 (expand_builtin_strncmp): Likewise.
13563 (inline_string_cmp): Delete the last parameter, change char_type_node
13564 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
13566 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
13567 the inlining expansion on target where the type of the call has same or
13568 narrower precision than unsigned char.
13570 2018-07-20 David Malcolm <dmalcolm@redhat.com>
13572 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
13573 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
13574 * common.opt (fsave-optimization-record): New option.
13575 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
13576 * doc/invoke.texi (-fsave-optimization-record): New option.
13577 * dumpfile.c: Include "optinfo-emit-json.h".
13578 (struct kv_pair): Move to coretypes.h.
13579 (optgroup_options): Make non-static.
13580 (dump_context::end_scope): Call
13581 optimization_records_maybe_pop_dump_scope.
13582 * dumpfile.h (optgroup_options): New decl.
13583 * json.cc: New file.
13584 * json.h: New file.
13585 * optinfo-emit-json.cc: New file.
13586 * optinfo-emit-json.h: New file.
13587 * optinfo.cc: Include "optinfo-emit-json.h".
13588 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
13589 (optinfo_enabled_p): Check optimization_records_enabled_p.
13590 (optinfo_wants_inlining_info_p): Likewise.
13591 * optinfo.h: Update comment.
13592 * profile-count.c (profile_quality_as_string): New function.
13593 * profile-count.h (profile_quality_as_string): New decl.
13594 (profile_count::quality): New accessor.
13595 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
13596 and optinfo_emit_json_cc_tests.
13597 * selftest.h (selftest::json_cc_tests): New decl.
13598 (selftest::optinfo_emit_json_cc_tests): New decl.
13599 * toplev.c: Include "optinfo-emit-json.h".
13600 (compile_file): Call optimization_records_finish.
13601 (do_compile): Call optimization_records_start.
13602 * tree-ssa-live.c: Include optinfo.h.
13603 (remove_unused_scope_block_p): Retain inlining information if
13604 optinfo_wants_inlining_info_p returns true.
13606 2018-07-20 Richard Biener <rguenther@suse.de>
13609 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
13610 to cover -flto-partition=none.
13612 2018-07-20 Martin Liska <mliska@suse.cz>
13614 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
13615 (get_decl_source_range): Remove unused function.
13617 2018-07-20 Richard Biener <rguenther@suse.de>
13619 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
13620 (struct vn_phi_s): Likewise.
13621 (struct vn_reference_s): Likewise.
13622 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
13623 for searching the slot of an entry known to be in the hash itself.
13624 (vn_phi_hasher::equal): Likewise.
13625 (vn_reference_hasher::equal): Likewise.
13626 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
13628 (optimistic_info, current_info): Remove, keeping only valid_info.
13629 (vn_reference_lookup_1): Remove fallback lookup.
13630 (vn_reference_lookup_2): Likewise.
13631 (vn_nary_op_lookup_1): Likewise.
13632 (vn_phi_lookup): Likewise.
13633 (vn_nary_build_or_lookup_1): Make sure to not chain the built
13635 (vn_reference_insert): Adjust, chain the inserted hash element
13636 at last_inserted_ref.
13637 (vn_reference_insert_pieces): Likewise.
13638 (visit_reference_op_call): Likewise.
13639 (vn_nary_op_insert_into): Chain the inserted hash element at
13640 last_inserted_nary.
13641 (vn_nary_op_insert_pieces): Adjust.
13642 (vn_nary_op_insert): Likewise.
13643 (vn_nary_op_insert_stmt): Likewise.
13644 (vn_phi_insert): Adjust, chain the inserted hash element at
13646 (process_scc): Remove clearing and copying the optimistic
13647 table. Instead remove elements inserted during an optimistic
13648 iteration from the single table we maintain.
13649 (init_scc_vn): Adjust.
13650 (free_scc_vn): Likewise.
13651 (sccvn_dom_walker::record_cond): Likewise.
13652 (sccvn_dom_walker::after_dom_children): Likewise.
13654 2018-07-19 Martin Sebor <msebor@redhat.com>
13656 PR tree-optimization/84047
13657 PR tree-optimization/83776
13658 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
13659 (check_array_bounds): Call it.
13661 2018-07-19 Martin Sebor <msebor@redhat.com>
13663 * align.h (align_flags): Use member initialization.
13665 2018-07-19 David Malcolm <dmalcolm@redhat.com>
13667 * Makefile.in (OBJS): Add optinfo.o.
13668 * coretypes.h (class symtab_node): New forward decl.
13669 (struct cgraph_node): New forward decl.
13670 (class varpool_node): New forward decl.
13671 * dump-context.h: New file.
13672 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
13674 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
13675 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
13676 (set_alt_dump_file): Likewise.
13677 (dump_context::~dump_context): New dtor.
13678 (dump_gimple_stmt): Move implementation to...
13679 (dump_context::dump_gimple_stmt): ...this new member function.
13680 Add the stmt to any pending optinfo, creating one if need be.
13681 (dump_gimple_stmt_loc): Move implementation to...
13682 (dump_context::dump_gimple_stmt_loc): ...this new member function.
13683 Start a new optinfo and add the stmt to it.
13684 (dump_gimple_expr): Move implementation to...
13685 (dump_context::dump_gimple_expr): ...this new member function.
13686 Add the stmt to any pending optinfo, creating one if need be.
13687 (dump_gimple_expr_loc): Move implementation to...
13688 (dump_context::dump_gimple_expr_loc): ...this new member function.
13689 Start a new optinfo and add the stmt to it.
13690 (dump_generic_expr): Move implementation to...
13691 (dump_context::dump_generic_expr): ...this new member function.
13692 Add the tree to any pending optinfo, creating one if need be.
13693 (dump_generic_expr_loc): Move implementation to...
13694 (dump_context::dump_generic_expr_loc): ...this new member
13695 function. Add the tree to any pending optinfo, creating one if
13697 (dump_printf): Move implementation to...
13698 (dump_context::dump_printf_va): ...this new member function. Add
13699 the text to any pending optinfo, creating one if need be.
13700 (dump_printf_loc): Move implementation to...
13701 (dump_context::dump_printf_loc_va): ...this new member function.
13702 Start a new optinfo and add the stmt to it.
13703 (dump_dec): Move implementation to...
13704 (dump_context::dump_dec): ...this new member function. Add the
13705 value to any pending optinfo, creating one if need be.
13706 (dump_context::dump_symtab_node): New member function.
13707 (dump_context::get_scope_depth): New member function.
13708 (dump_context::begin_scope): New member function.
13709 (dump_context::end_scope): New member function.
13710 (dump_context::ensure_pending_optinfo): New member function.
13711 (dump_context::begin_next_optinfo): New member function.
13712 (dump_context::end_any_optinfo): New member function.
13713 (dump_context::s_current): New global.
13714 (dump_context::s_default): New global.
13715 (dump_scope_depth): Delete global.
13716 (dumpfile_ensure_any_optinfo_are_flushed): New function.
13717 (dump_symtab_node): New function.
13718 (get_dump_scope_depth): Reimplement in terms of dump_context.
13719 (dump_begin_scope): Likewise.
13720 (dump_end_scope): Likewise.
13721 (selftest::temp_dump_context::temp_dump_context): New ctor.
13722 (selftest::temp_dump_context::~temp_dump_context): New dtor.
13723 (selftest::verify_item): New function.
13724 (ASSERT_IS_TEXT): New macro.
13725 (ASSERT_IS_TREE): New macro.
13726 (ASSERT_IS_GIMPLE): New macro.
13727 (selftest::test_capture_of_dump_calls): New test.
13728 (selftest::dumpfile_c_tests): Call it.
13729 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
13730 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
13731 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
13732 descriptive comment.
13733 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
13734 (dump_node, dump_bb): Move these unrelated decls.
13735 (class dump_manager): Add leading comment.
13736 * optinfo.cc: New file.
13737 * optinfo.h: New file.
13739 2018-07-19 Michael Collison <michael.collison@arm.com>
13740 Richard Henderson <rth@redhat.com>
13742 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
13743 (subti): Handle op1 zero.
13744 (subvti4, usub4ti4): New.
13745 (*sub<GPI>3_compare1_imm): New.
13746 (sub<GPI>3_carryinCV): New.
13747 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
13748 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
13750 2018-07-19 Michael Collison <michael.collison@arm.com>
13751 Richard Henderson <rth@redhat.com>
13753 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
13754 (addti3): Create simpler code if low part is already known to be 0.
13755 (addvti4, uaddvti4): New.
13756 (*add<GPI>3_compareC_cconly_imm): New.
13757 (*add<GPI>3_compareC_cconly): New.
13758 (*add<GPI>3_compareC_imm): New.
13759 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
13760 handle constants within this pattern..
13761 (*add<GPI>3_compareV_cconly_imm): New.
13762 (*add<GPI>3_compareV_cconly): New.
13763 (*add<GPI>3_compareV_imm): New.
13764 (add<GPI>3_compareV): New.
13765 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
13766 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
13767 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
13768 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
13769 with 'comparison' operator.
13770 (*add<GPI>3_compareV_cconly_imm): Ditto.
13771 (*add<GPI>3_compareV_cconly): Ditto.
13772 (*add<GPI>3_compareV_imm): Ditto.
13773 (add<GPI>3_compareV): Ditto.
13774 (add<mode>3_carryinC): Ditto.
13775 (*add<mode>3_carryinC_zero): Ditto.
13776 (*add<mode>3_carryinC): Ditto.
13777 (add<mode>3_carryinV): Ditto.
13778 (*add<mode>3_carryinV_zero): Ditto.
13779 (*add<mode>3_carryinV): Ditto.
13781 2018-07-19 Michael Collison <michael.collison@arm.com>
13782 Richard Henderson <rth@redhat.com>
13784 * config/aarch64/aarch64-modes.def (CC_V): New.
13785 * config/aarch64/aarch64-protos.h
13786 (aarch64_addti_scratch_regs): Declare
13787 (aarch64_subvti_scratch_regs): Declare.
13788 (aarch64_expand_subvti): Declare.
13789 (aarch64_gen_unlikely_cbranch): Declare
13790 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
13791 for signed overflow using CC_Vmode.
13792 (aarch64_get_condition_code_1): Handle CC_Vmode.
13793 (aarch64_gen_unlikely_cbranch): New function.
13794 (aarch64_addti_scratch_regs): New function.
13795 (aarch64_subvti_scratch_regs): New function.
13796 (aarch64_expand_subvti): New function.
13798 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
13800 * config/aarch64/aarch64-option-extensions.def: New entry for profile
13802 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
13803 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
13806 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
13809 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13810 address check not strict.
13812 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
13814 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
13816 (store_pair_lanes<mode>): Likewise.
13817 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
13818 enum value 'ADDR_QUERY_LDP_STP_N'.
13819 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
13820 (aarch64_print_address_internal): Add declaration.
13821 (aarch64_print_ldpstp_address): Remove.
13822 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
13823 (aarch64_print_operand): Change printing of 'y'.
13824 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
13825 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
13826 'true' rather than '1'.
13827 * config/aarch64/constraints.md (Uml): Likewise.
13828 (Uml): Rename to Umn.
13831 2018-07-19 Richard Biener <rguenther@suse.de>
13833 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
13835 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
13836 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
13837 (vn_reference_hasher): Likewise.
13838 (struct vn_tables_s): Remove obstack and alloc-pool members.
13839 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
13840 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
13841 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
13842 (vn_reference_insert_pieces): Likewise.
13843 (alloc_vn_nary_op_noinit): Adjust.
13844 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
13845 (vn_phi_eq): Adjust.
13846 (shared_lookup_phiargs): Remove.
13847 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
13848 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
13849 (visit_reference_op_call): Likewise.
13850 (copy_nary, copy_phi, copy_reference): Remove.
13851 (process_scc): Rewind the obstack when iterating. Do not
13852 copy the elements to valid_info but just move them from one
13853 hashtable to the other.
13854 (allocate_vn_table): Adjust.
13855 (free_vn_table): Likewise.
13856 (init_scc_vn): Likewise.
13857 (free_scc_vn): Likewise.
13859 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
13862 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
13863 indirect_return as function type attribute.
13864 (ix86_attribute_table): Change indirect_return to function
13866 * doc/extend.texi: Update indirect_return attribute.
13868 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
13870 * wide-int.h (widest2_int): New.
13871 * gimple-fold.c (arith_overflowed_p): Use it.
13872 * tree.h (widest2_int_cst): New.
13873 * tree-vrp.c (wide_int_binop_overflow): Rename from
13874 vrp_int_const_binop.
13875 Rewrite to work on trees.
13876 (extract_range_from_multiplicative_op_1): Abstract code to...
13877 (wide_int_range_min_max): ...here.
13878 (wide_int_range_cross_product): ...and here.
13879 (extract_range_from_binary_expr_1): Abstract overflow code to...
13880 (wide_int_range_mult_wrapping): ...here.
13881 * tree-vrp.h (wide_int_range_cross_product): New.
13882 (wide_int_range_mult_wrapping): New.
13884 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
13885 Julia Koval <julia.koval@intel.com>
13887 * config/i386/x86-tune-costs.h (skylake_memcpy,
13888 skylake_memset): Replace rep_prefix with unrolling for size 512.
13890 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
13892 PR middle-end/86544
13893 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
13894 comparision with EQ_EXPR in last stmt.
13896 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
13898 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
13899 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
13900 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
13901 previously known as "PowerPC AltiVec Built-in Functions". Move
13902 some material to new subsubsections "PowerPC AltiVec Built-in
13903 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
13905 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
13906 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
13907 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
13908 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
13910 2018-07-18 Richard Biener <rguenther@suse.de>
13912 PR tree-optimization/86557
13913 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
13916 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
13918 * config/s390/s390.c (s390_function_profiler): Generate CFI.
13920 2018-07-17 Jeff Law <law@redhat.com>
13922 * config/arm/arm.c (get_label_padding): Update for recent
13923 changes to label_to_alignment.
13925 PR tree-optimization/86010
13926 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
13928 * config/mips/mips.c (vr4130_align_insns): Update for recent
13929 changes to label_to_alignment.
13931 * config/frv/frv.c (frv_label_align): Update for recent changes
13932 to label_to_alignment.
13934 * config/nios2/nios2.c (nios2_label_align): Update for recent
13935 changes which dropped ALIGN_LABELS_LOG.
13937 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
13939 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
13940 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
13942 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
13944 * config/arc/arc.c (arc_label_align): Use align_labels instead of
13945 deprecated align_labels_log.
13947 2018-07-17 Richard Biener <rguenther@suse.de>
13950 * dwarf2out.c (init_sections_and_labels): Always generate
13951 a debug_line_str_section for early LTO debug.
13952 (dwarf2out_finish): Reset debug_line_str_hash output early.
13953 Bump counter for extra dwarf5 .debug_loc labels to not conflict
13955 (dwarf2out_early_finish): Output debug_line_str.
13957 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
13959 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
13960 index register on z196 or later.
13962 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
13964 * config/s390/s390.c (s390_default_align): Set default function
13966 (s390_override_options_after_change): Call s390_default align.
13967 (s390_option_override_internal): Call s390_default_align.
13968 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
13970 2018-07-17 Jakub Jelinek <jakub@redhat.com>
13972 PR middle-end/86542
13973 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
13974 to _looptemp_ clauses, other than the first two.
13976 2018-07-17 Martin Liska <mliska@suse.cz>
13978 * opts.c: Do not enable OPT_falign_* for -Os.
13980 2018-07-17 Martin Liska <mliska@suse.cz>
13982 * align.h (MAX_CODE_ALIGN): New.
13983 (MAX_CODE_ALIGN_VALUE): New.
13984 * common/config/i386/i386-common.c (ix86_handle_option):
13985 (MAX_CODE_ALIGN): Moved to align.h.
13986 * final.c (MAX_CODE_ALIGN): Likewise.
13987 * opts.c (parse_and_check_align_values):
13988 (MAX_CODE_ALIGN): Likewise.
13989 (MAX_CODE_ALIGN_VALUE): Likewise.
13991 2018-07-17 Martin Liska <mliska@suse.cz>
13993 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
13994 in order to fulfil coding style.
13995 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
13996 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13997 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13998 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
13999 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
14000 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
14001 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
14002 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
14003 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
14005 2018-07-17 Martin Liska <mliska@suse.cz>
14007 * align.h: New file.
14008 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
14010 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
14011 align_flags of label_to_alignment.
14012 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
14014 * config/m68k/m68k.c: Do not use removed align_labels_value and
14016 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
14017 (LOOP_ALIGN): Likewise.
14018 (LABEL_ALIGN): Likewise.
14019 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
14020 Remove not used macro.
14021 (rs6000_loop_align): Change return type to align_flags.
14022 (rs6000_loop_align_max_skip): Remove.
14023 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
14024 Change return type to align_flags.
14025 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
14026 Remove not used macro.
14027 (rs6000_loop_align): Change return type to align_flags.
14028 (rs6000_loop_align_max_skip): Remove.
14029 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
14030 * config/rx/rx-protos.h (rx_align_for_label): Make it
14032 * config/rx/rx.c (rx_align_for_label): Change return type
14034 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
14036 into align_flags class.
14037 (LABEL_ALIGN): Likewise.
14038 (LOOP_ALIGN): Likewise.
14039 * config/s390/s390.c (s390_label_align): Use align_flags
14041 (s390_asm_output_function_label): Likewise.
14042 * config/sh/sh.c (sh_override_options_after_change):
14043 Use align_flags class directly without macros.
14044 (find_barrier): Likewise.
14045 (barrier_align): Likewise.
14046 (sh_loop_align): Likewise.
14047 * config/spu/spu.c (spu_option_override):
14048 Use align_flags_tuple::get_value instead of removed macros.
14049 (spu_sched_init): Likewise.
14050 * config/spu/spu.h (GTY): Likewise.
14051 * config/visium/visium.c (visium_option_override):
14052 Set "8" as default secondary alignment.
14053 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
14054 in order to guarantee secondary alignment of 8.
14055 * coretypes.h: Include align.h header file.
14056 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
14057 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
14058 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
14059 * doc/tm.texi.in: Likewise.
14060 * final.c (struct label_alignment): Remove not used structure.
14061 (LABEL_ALIGN): Change type to align_flags.
14062 (LOOP_ALIGN): Likewise.
14063 (JUMP_ALIGN): Likewise.
14064 (default_loop_align_max_skip): Remove.
14065 (default_label_align_max_skip): Likewise.
14066 (default_jump_align_max_skip): Likewise.
14067 (default_label_align_after_barrier_max_skip):
14068 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
14069 (LABEL_TO_MAX_SKIP): Remove.
14070 (label_to_alignment): Return align_flags type instead of integer.
14071 (label_to_max_skip): Remove.
14072 (align_fuzz): Use align_flags type.
14073 (compute_alignments): Use align_flags type and use align_flags::max
14074 to combine multiple alignments.
14075 (grow_label_align): Grow vec instead of C array.
14076 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
14077 (shorten_branches): Use align_flags type and use align_flags::max
14078 to combine multiple alignments.
14079 (final_scan_insn_1): Remove usage of secondary alignment that comes
14080 from label alignment, but instead use proper secondary alignment
14081 which is computed in grow_label_align.
14082 * flags.h (struct align_flags_tuple): Move to align.h.
14083 (struct align_flags): Likewise.
14084 (state_align_loops): Rename to align_loops.
14085 (state_align_jumps): Rename to align_jumps.
14086 (state_align_labels): Rename to align_labels.
14087 (state_align_functions): Rename to align_functions.
14088 (align_loops_log): Remove.
14089 (align_jumps_log): Remove.
14090 (align_labels_log): Remove.
14091 (align_functions_log): Remove.
14092 (align_loops_max_skip): Remove.
14093 (align_jumps_max_skip): Remove.
14094 (align_labels_max_skip): Remove.
14095 (align_functions_max_skip): Remove.
14096 (align_loops_value): Remove.
14097 (align_jumps_value): Remove.
14098 (align_labels_value): Remove.
14099 (align_functions_value): Remove.
14100 * output.h (label_to_alignment): Change return type to align_flags.
14101 (label_to_max_skip): Remove.
14102 * target.def: Remove loop_align_max_skip, label_align_max_skip,
14103 jump_align_max_skip macros.
14104 * targhooks.h (default_loop_align_max_skip): Remove.
14105 (default_label_align_max_skip): Likewise.
14106 (default_jump_align_max_skip): Likewise.
14107 (default_label_align_after_barrier_max_skip): Remove.
14108 * toplev.c (read_log_maxskip): Use ::normalize function.
14109 (parse_N_M): Remove not used argument and also call ::normalize.
14110 (parse_alignment_opts): Do not pass unused arguments.
14111 * varasm.c (assemble_start_function): Use directly align_functions
14112 instead of removed macros.
14113 * system.h: Do not poison removed macros.
14115 2018-07-17 Jakub Jelinek <jakub@redhat.com>
14117 PR middle-end/86539
14118 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
14119 and cond temporaries don't have reference type if iterator has
14120 pointer type. For init use &for_pre_body instead of pre_p if
14121 for_pre_body is non-empty.
14123 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
14125 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
14126 double-double modes to SFmode directly directly.
14127 (trunc<mode>sf2_fprs): Delete.
14129 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
14131 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
14132 for conversions between IFmode and the decimal floating point modes.
14133 (init_float128_ieee): Use the correct names for conversions between
14134 KFmode and the decimal floating point modes.
14136 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
14138 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
14139 for the conversions between TDmode and IFmode.
14140 (init_float128_ieee): Use more correct names for the conversions
14141 between TDmode and KFmode.
14143 2018-07-16 Jakub Jelinek <jakub@redhat.com>
14145 PR tree-optimization/86526
14146 * builtins.c (expand_builtin_memcmp): Formatting fixes.
14147 (inline_expand_builtin_string_cmp): Likewise.
14148 (inline_string_cmp): Likewise. Use c_readstr instead of
14149 builtin_memcpy_read_str. Add unit_mode temporary.
14151 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
14153 PR middle-end/86528
14154 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
14155 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
14157 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
14159 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
14160 Alphabetize prototypes of built-in functions, separating out
14161 built-in functions that are listed in this section but should be
14162 described elsewhere.
14164 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
14167 * expmed.c (emit_store_flag): Do not emit setcc followed by a
14168 conditional move when trapping comparison was split to a
14169 non-trapping one (and vice versa).
14171 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
14173 * config/s390/s390.c (s390_function_profiler): Generate nops
14174 instead of profiler call sequences.
14175 * config/s390/s390.opt: Add the new option.
14177 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
14179 * config/s390/s390.c (s390_function_profiler): Generate
14180 __mcount_loc section.
14181 * config/s390/s390.opt: Add the new option.
14183 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
14185 * common.opt: Add the new warning.
14186 * config/s390/s390.c (s390_function_profiler): Emit "brasl
14187 %r0,__fentry__" when -mfentry is specified.
14188 (s390_option_override_internal): Disallow -mfentry for 31-bit
14190 * config/s390/s390.opt: Add the new option.
14192 2018-07-16 Richard Biener <rguenther@suse.de>
14195 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
14196 for function-local FUNCTION_DECL and RESULT_DECL immediately.
14198 2018-07-16 Martin Liska <mliska@suse.cz>
14201 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
14204 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
14206 * config/arc/arcHS.md: Update ARCHS scheduling rules.
14208 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
14210 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
14212 * config/arc/arc-cpus.def (hs4x): New cpu.
14214 * config/arc/arc-tables.opt: Regenerate.
14215 * config/arc/arc.c (arc_sched_issue_rate): New function.
14216 (TARGET_SCHED_ISSUE_RATE): Define.
14217 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
14218 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
14220 (attr tune): Add ARCHS4x tune values.
14221 (attr tune_dspmpy): Define.
14222 (*tst): Correct instruction type.
14223 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
14224 * config/arc/arcHS4x.md: New file.
14225 * config/arc/fpu.md: Update instruction type attributes.
14226 * config/arc/t-multilib: Regenerate.
14228 2018-07-16 Tom de Vries <tdevries@suse.de>
14231 * var-tracking.c (vt_initialize): Fix pre_dec handling.
14233 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14235 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
14238 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
14240 PR tree-optimization/86514
14241 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
14242 conversion to a boolean type from a type with greater precision.
14244 2018-07-16 Tom de Vries <tdevries@suse.de>
14246 * var-tracking.c (vt_initialize): Print adjusted insn slim if
14247 dump_flags request TDF_SLIM.
14249 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
14251 * fold-const.c (int_const_binop_1): Abstract...
14252 (wide_int_binop): ...wide int code here.
14253 (poly_int_binop): ...poly int code here.
14254 Abstract the rest of int_const_binop_1 into int_const_binop.
14255 * fold-const.h (wide_int_binop): New.
14256 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
14257 Remove useless PLUS/MINUS_EXPR case.
14258 (zero_nonzero_bits_from_vr): Move wide int code...
14259 (zero_nonzero_bits_from_bounds): ...here.
14260 (extract_range_from_binary_expr_1): Move mask optimization code...
14261 (range_easy_mask_min_max): ...here.
14262 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
14263 (range_easy_mask_min_max): New.
14265 2018-07-15 Jeff Law <law@redhat.com>
14268 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
14271 2018-07-14 Jim Wilson <jimw@sifive.com>
14273 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
14275 2018-07-14 Paul Koning <ni1d@arrl.net>
14277 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
14279 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
14281 * lto-streamer-out.c (copy_function_or_variable): Dump info about
14284 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
14285 Steve Munroe <munroesj52@gmail.com>
14287 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
14288 (_mm_andnot_si128): Likewise.
14289 (_mm_or_si128): Likewise.
14290 (_mm_xor_si128): Likewise.
14292 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
14294 PR middle-end/78809
14295 * builtins.c (expand_builtin_memcmp): Inline the calls first
14296 when result_eq is false.
14297 (expand_builtin_strcmp): Inline the calls first.
14298 (expand_builtin_strncmp): Likewise.
14299 (inline_string_cmp): New routine. Expand a string compare
14300 call by using a sequence of char comparison.
14301 (inline_expand_builtin_string_cmp): New routine. Inline expansion
14302 a call to str(n)cmp/memcmp.
14303 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
14305 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
14307 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
14309 * config/arm/driver-arm.c: Include arm-native.h.
14310 (host_detect_local_cpu): Use auto-generated data tables.
14311 (vendors, arm_cpu_table): Delete. Move part information to ...
14312 * config/arm/arm-cpus.in: ... here.
14313 * config/arm/parsecpu.awk (gen_native): New function.
14314 (vendor, part): New CPU fields.
14315 (END): Add support for building the native CPU detection tables.
14316 * config/arm/t-arm (arm-native.h): Add build rule.
14317 (driver-arm.o): Add dependency on arm-native.h.
14319 2018-07-13 Richard Biener <rguenther@suse.de>
14321 PR middle-end/85974
14322 * match.pd (addr1 - addr2): Allow either of the operand to
14325 2018-07-13 Tom de Vries <tdevries@suse.de>
14327 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
14330 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
14332 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
14333 arrays instead of numbered variables.
14335 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
14337 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
14338 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
14340 2018-07-13 Richard Biener <rguenther@suse.de>
14343 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
14344 instead of get_context_die.
14346 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
14347 Richard Biener <rguenther@suse.de>
14349 PR middle-end/86489
14350 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
14351 that the loop latch destination where phi is defined.
14353 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
14355 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
14356 (riscv_expand_epilogue): Add assertion to check interrupt mode.
14357 (riscv_set_current_function): Extract getting interrupt type to new
14359 (riscv_get_interrupt_type): New function.
14360 (riscv_merge_decl_attributes): New function, checking interrupt type is
14362 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
14364 2018-07-12 Paul Koning <ni1d@arrl.net>
14366 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
14369 2018-07-12 Paul Koning <ni1d@arrl.net>
14371 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
14372 zero reference, add doloop_end instead.
14373 * doc/md.texi (decrement_and_branch_until_zero): Remove.
14374 (Looping patterns): Remove decrement_and_branch_until_zero. Add
14375 detail for doloop_end.
14377 2018-07-12 Martin Sebor <msebor@redhat.com>
14380 * attribs.c (decl_attributes): Reject conflicting attributes before
14381 calling attribute handlers.
14383 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
14385 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
14387 (gcc::dump_manager::get_dump_file_name): likewise.
14388 (dump_begin): Likewise.
14389 * dumpfile.h (dump_begin): Update prototype.
14390 (gcc::dump_manager::get_dump_file_name,
14391 gcc::dump_manager::get_dump_file_name): Update prototype.
14393 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
14395 * internal-fn.h (vectorizable_internal_fn_p): New function.
14396 * tree-vect-slp.c (compatible_calls_p): Likewise.
14397 (vect_build_slp_tree_1): Remove nops argument. Handle calls
14398 to internal functions.
14399 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
14401 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
14403 * fold-const.h (inverse_conditions_p): Declare.
14404 * fold-const.c (inverse_conditions_p): New function.
14405 * match.pd: Use inverse_conditions_p. Add folds of view_converts
14406 that test the inverse condition of a conditional internal function.
14407 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
14408 * internal-fn.c (internal_fn_mask_index): Handle conditional
14409 internal functions.
14410 (vectorized_internal_fn_supported_p): New function.
14411 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
14412 (any_pred_load_store): Replace with...
14413 (need_to_predicate): ...this new variable.
14414 (redundant_ssa_names): New variable.
14415 (ifcvt_can_use_mask_load_store): Move initial checks to...
14416 (ifcvt_can_predicate): ...this new function. Handle tree codes
14417 for which a conditional internal function exists.
14418 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
14419 instead of ifcvt_can_use_mask_load_store. Update after variable
14421 (predicate_load_or_store): New function, split out from
14422 predicate_mem_writes.
14423 (check_redundant_cond_expr): New function.
14424 (value_available_p): Likewise.
14425 (predicate_rhs_code): Likewise.
14426 (predicate_mem_writes): Rename to...
14427 (predicate_statements): ...this. Use predicate_load_or_store
14428 and predicate_rhs_code.
14429 (combine_blocks, tree_if_conversion): Update after above name changes.
14430 (ifcvt_local_dce): Handle redundant_ssa_names.
14431 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
14432 general conditional functions.
14433 * tree-vect-stmts.c (vectorizable_call): Likewise.
14435 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
14436 Alan Hayward <alan.hayward@arm.com>
14437 David Sherwood <david.sherwood@arm.com>
14439 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
14440 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
14441 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
14442 plus and minus and convert them into IFN_COND_FMA-based sequences.
14443 (convert_mult_to_fma): Handle conditional plus and minus.
14445 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
14447 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
14448 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
14449 (cond_fnms_optab): New optabs.
14450 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
14451 internal functions.
14452 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
14453 * internal-fn.h (get_conditional_internal_fn): Declare.
14454 (get_unconditional_internal_fn): Likewise.
14455 * internal-fn.c (cond_ternary_direct): New macro.
14456 (expand_cond_ternary_optab_fn): Likewise.
14457 (direct_cond_ternary_optab_supported_p): Likewise.
14458 (FOR_EACH_COND_FN_PAIR): Likewise.
14459 (get_conditional_internal_fn): New function.
14460 (get_unconditional_internal_fn): Likewise.
14461 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
14462 (gimple_match_op::gimple_match_op): Add a new overload for 5
14464 (gimple_match_op::set_op): Likewise.
14465 (gimple_resimplify5): Declare.
14466 * genmatch.c (decision_tree::gen): Generate simplifications for
14468 * gimple-match-head.c (gimple_simplify): Define an overload for
14469 5 operands. Handle calls with 5 arguments in the top-level overload.
14470 (convert_conditional_op): Handle conversions from unconditional
14471 internal functions to conditional ones.
14472 (gimple_resimplify5): New function.
14473 (build_call_internal): Pass a fifth operand.
14474 (maybe_push_res_to_seq): Likewise.
14475 (try_conditional_simplification): Try converting conditional
14476 internal functions to unconditional internal functions.
14477 Handle 3-operand unconditional forms.
14478 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
14479 Define ternary equivalents of the current rules for binary conditional
14480 internal functions.
14481 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
14482 ternary operations.
14483 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
14484 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
14485 (optab): Handle them.
14486 (SVE_COND_FP_TERNARY): New int iterator.
14487 (sve_fmla_op, sve_fmad_op): New int attributes.
14488 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
14489 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
14490 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
14492 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
14494 * target.def (preferred_else_value): New target hook.
14495 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
14496 * doc/tm.texi: Regenerate.
14497 * targhooks.h (default_preferred_else_value): Declare.
14498 * targhooks.c (default_preferred_else_value): New function.
14499 * internal-fn.h (conditional_internal_fn_code): Declare.
14500 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
14501 (get_conditional_internal_fn): Use it.
14502 (conditional_internal_fn_code): New function.
14503 * gimple-match.h (gimple_match_cond): New struct.
14504 (gimple_match_op): Add a cond member function.
14505 (gimple_match_op::gimple_match_op): Update all forms to take a
14507 * genmatch.c (expr::gen_transform): Use the same condition as res_op
14508 for the suboperation, but don't specify a particular else_value.
14509 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
14510 (visit_nary_op, visit_reference_op_load): Pass
14511 gimple_match_cond::UNCOND to the gimple_match_op constructor.
14512 * gimple-match-head.c: Include tree-eh.h
14513 (convert_conditional_op): New function.
14514 (maybe_resimplify_conditional_op): Likewise.
14515 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
14516 (gimple_resimplify2): Likewise.
14517 (gimple_resimplify3): Likewise.
14518 (gimple_resimplify4): Likewise.
14519 (maybe_push_res_to_seq): Return null for conditional operations.
14520 (try_conditional_simplification): New function.
14521 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
14523 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
14525 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
14527 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
14529 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
14531 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
14533 (hash_tree): Do not hash DECL_FCONTEXT
14534 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
14535 Do not stream DECL_FCONTEXT.
14536 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
14537 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
14539 2018-07-12 Richard Biener <rguenther@suse.de>
14542 * dwarf2out.c (gen_block_die): Only output blocks when they have
14543 at least one !DECL_IGNORED_P variable.
14545 2018-07-12 Richard Biener <rguenther@suse.de>
14548 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
14549 Remove -mieee-fp handling.
14551 2018-07-12 Richard Biener <rguenther@suse.de>
14553 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
14554 left-over from last patch.
14556 2018-07-12 Jakub Jelinek <jakub@redhat.com>
14558 PR tree-optimization/86492
14559 * gimple-ssa-store-merging.c
14560 (imm_store_chain_info::coalesce_immediate_stores): Call
14561 check_no_overlap even for the merge_overlapping case. Formatting fix.
14563 2018-07-12 Richard Biener <rguenther@suse.de>
14565 PR middle-end/86479
14566 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
14567 move possibly trapping operations into the conditional.
14569 2018-07-12 Richard Biener <rguenther@suse.de>
14571 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
14572 (vn_lookup_simplify_result): Remove recursion limit applied
14574 (vn_nary_build_or_lookup_1): Adjust.
14575 (try_to_simplify): Likewise.
14576 * gimple-match-head.c (gimple_resimplify1): Instead apply one
14578 (gimple_resimplify2): Likewise.
14579 (gimple_resimplify3): Likewise.
14580 (gimple_resimplify4): Likewise.
14582 2018-07-11 Jakub Jelinek <jakub@redhat.com>
14584 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
14585 Use __mmask64 type instead of __mmask8 for __M argument.
14586 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
14587 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
14589 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
14590 __mmask16 for __M argument.
14591 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
14592 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
14593 to __mmask16 instead of __mmask8.
14594 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
14595 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
14596 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
14597 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
14598 instead of __mmask16 for __U argument.
14599 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
14600 __mmask16 instead of __mmask8 for __U argument.
14601 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
14603 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
14605 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
14607 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
14609 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
14611 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
14613 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
14615 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
14617 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
14619 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
14621 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
14622 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
14623 return type as well as __M argument type and all casts from __mmask8
14625 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
14626 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
14627 return type as well as __M argument type and all casts from __mmask8
14629 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
14630 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
14631 return type as well as __M argument type and all casts from __mmask8
14633 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
14634 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
14635 return type as well as __M argument type and all casts from __mmask8
14637 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
14638 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
14641 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
14643 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
14644 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
14647 2018-07-11 Paul Koning <ni1d@arrl.net>
14649 * doc/md.texi (define_subst): Document how multiple occurrences of
14650 the same argument in the replacement pattern are handled.
14652 2018-07-11 Paul Koning <ni1d@arrl.net>
14654 * doc/extend.texi (Common Variable Attributes): Move "mode" into
14655 alphabetical order.
14656 (Common Type Attributes): Add "mode" attribute.
14658 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
14660 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
14661 stream DECL_ORIGINAL_TYPE.
14662 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
14663 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
14664 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
14665 Do not walk original type.
14666 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
14668 (write_ts_decl_non_common_tree_pointers): Do not stream
14670 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
14671 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
14673 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
14675 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
14676 thread twice from the same starting edge.
14678 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
14680 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
14681 * gimple.c (gimple_call_nonnull_result_p): ...here...
14682 (gimple_call_nonnull_arg): ...and here.
14683 * gimple.h (gimple_call_nonnull_result_p): New.
14684 (gimple_call_nonnull_arg): New.
14686 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
14688 * config/arm/arm-cpus.in: Move information from fpu field of each
14689 cpu definition to the isa field.
14690 * config/arm/parsecpu.awk (fpu): Delete match rule.
14691 (gen_comm_data): Don't add bits from the CPU's FPU entry.
14693 2018-07-11 Richard Biener <rguenther@suse.de>
14696 * dwarf2out.c (init_sections_and_labels): Use
14697 output_asm_line_debug_info consistently.
14698 (dwarf2out_early_finish): Likewise.
14699 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
14702 2018-07-11 Richard Biener <rguenther@suse.de>
14704 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
14705 Rework father_bb setting in a way to avoid propagating constants
14706 multiple times on a loop body.
14708 2018-07-10 Mark Wielaard <mark@klomp.org>
14711 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
14713 2018-07-10 Richard Biener <rguenther@suse.de>
14715 * hash-map.h (hash_map::iterator::operator*): Return
14716 references to key and value.
14718 2018-07-10 Jakub Jelinek <jakub@redhat.com>
14721 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
14722 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
14723 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
14724 (gimplify_omp_for): For composite loops, move outer
14725 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
14726 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
14727 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
14728 TREE_LIST for both the original class iterator and the "last" helper
14729 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
14730 loop, remember has_decl_expr from outer composite loops for the
14731 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
14733 2018-07-09 Martin Sebor <msebor@redhat.com>
14735 PR middle-end/77357
14736 PR middle-end/86428
14737 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
14738 accessing implicitly initialized array elements.
14739 * expr.c (string_constant): Handle string initializers of
14740 character arrays within aggregates.
14741 * gimple-fold.c (fold_array_ctor_reference): Add argument.
14742 Store element offset. As a special case, handle zero size.
14743 (fold_nonarray_ctor_reference): Same.
14744 (fold_ctor_reference): Add argument. Store subobject offset.
14745 * gimple-fold.h (fold_ctor_reference): Add argument.
14747 2018-07-09 Paul Koning <ni1d@arrl.net>
14749 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
14750 (pdp11_insn_cost): New function.
14751 (pdp11_md_asm_adjust): New function.
14752 (TARGET_INVALID_WITHIN_DOLOOP): Define.
14753 (pdp11_rtx_costs): Update to match machine better.
14754 (output_addr_const_pdp11): Correct format mismatch warnings.
14755 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
14756 * config/pdp11/pdp11.md: General change to add base_cost and/or
14757 length attributes for use by new pdp11_insn_cost function.
14758 (MIN_BRANCH): Correct definition.
14760 (doloop_end): Use standard pattern name for looping pattern.
14761 (doloop_end_nocc): New.
14762 (movsf): Add another constraint alternative.
14763 (zero_extendqihi2): Add constraint alternatives for not in place
14765 (zero_extendhisi2): Remove.
14766 (shift patterns): Add CC handling variants.
14770 (define_peephole2): New peephole to recognize mov that sets CC for
14773 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14775 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
14776 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
14779 2018-07-09 Jakub Jelinek <jakub@redhat.com>
14782 * real.c (real_nextafter): Return true if result is denormal.
14784 2018-07-09 Martin Liska <mliska@suse.cz>
14786 * common.opt: Add back wrongly removed attribute.
14788 2018-07-09 Richard Biener <rguenther@suse.de>
14791 * dwarf2out.c (gen_block_die): For an early generated DIE
14792 always output high/low PC attributes.
14794 2018-07-09 Tom de Vries <tdevries@suse.de>
14796 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
14797 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
14798 onto VAR_DECL with abstract origin.
14800 2018-07-07 Jim Wilson <jimw@sifive.com>
14802 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
14804 2018-07-07 Tom de Vries <tdevries@suse.de>
14806 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
14808 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
14810 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
14812 (combine_bound): Use wide-int overflow calculation instead of
14814 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
14816 * fold-const.c (int_const_binop_2): Same.
14817 (extract_muldiv_1): Same.
14818 (fold_div_compare): Same.
14819 (fold_abs_const): Same.
14821 * poly-int.h (add): Same.
14825 * predict.c (predict_iv_comparison): Same.
14826 * profile-count.c (slow_safe_scale_64bit): Same.
14827 * simplify-rtx.c (simplify_const_binary_operation): Same.
14828 * tree-chrec.c (tree_fold_binomial): Same.
14829 * tree-data-ref.c (split_constant_offset_1): Same.
14830 * tree-if-conv.c (idx_within_array_bound): Same.
14831 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
14832 * tree-ssa-phiopt.c (minmax_replacement): Same.
14833 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
14834 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
14835 * vr-values.c (vr_values::adjust_range_with_scev): Same.
14836 * wide-int.cc (wi::add_large): Same.
14837 (wi::mul_internal): Same.
14838 (wi::sub_large): Same.
14839 (wi::divmod_internal): Same.
14840 * wide-int.h: Change overflow type to overflow_type for neg, add,
14841 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
14842 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
14843 mul_internal, divmod_internal.
14844 (overflow_type): New enum.
14845 (accumulate_overflow): New.
14847 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
14849 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
14850 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
14852 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
14854 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
14855 argument is checked for zero before entering loop, avoid checking again.
14857 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
14859 * gimplify.h (generic_expr_could_trap_p): Set as global function.
14860 * gimplify.h (generic_expr_could_trap_p): Likwise.
14861 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
14863 2018-07-06 Jakub Jelinek <jakub@redhat.com>
14865 PR tree-optimization/86401
14866 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
14867 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
14868 (fold_bit_and_mask): ... here. New helper function for match.pd.
14869 * fold-const.h (fold_bit_and_mask): Declare.
14870 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
14872 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
14875 * target.def (translate_mode_attribute): New hook.
14876 * targhooks.h (default_translate_mode_attribute): Declare.
14877 * targhooks.c (default_translate_mode_attribute): New function.
14878 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
14879 * doc/tm.texi: Regenerate.
14880 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
14881 (rs6000_translate_mode_attribute): New function.
14883 2018-07-06 Paul Koning <ni1d@arrl.net>
14885 * doc/md.texi (define_split): Document DONE and FAIL.
14886 (define_peephole2): Ditto.
14888 2018-07-05 Jeff Law <law@redhat.com>
14890 PR tree-optimization/86010
14891 * tree-ssa-dse.c (compute_trims): More aggressively trim at
14892 both the head and tail of mem* and str* calls.
14894 2018-07-05 Jim Wilson <jimw@sifive.com>
14896 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
14898 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
14900 * config/aarch64/aarch64-simd.md: correct flags text for
14901 MIN_EXPR replacement.
14903 2018-07-05 James Clarke <jrtc27@jrtc27.com>
14905 * configure: Regenerated.
14907 2018-07-05 Carl Love <cel@us.ibm.com>
14909 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
14910 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
14911 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
14912 VSX_BUILTIN_DOUBLEL_V4SF.
14914 2018-07-05 Martin Sebor <msebor@redhat.com>
14917 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
14918 than its domain to compute its the upper bound of a char array.
14920 2018-07-05 Nathan Sidwell <nathan@acm.org>
14922 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
14923 * doc/cpp.texi: Update comment.
14924 * doc/tm.texi: Rebuilt.
14925 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
14926 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
14927 * doc/extend.texi (Backwards Compatibility): Clarify it is system
14928 headers affected by extern "C".
14929 * system.h: Poison NO_IMPLICIT_EXTERN_C.
14930 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
14931 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
14932 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
14933 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
14934 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
14935 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
14936 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
14937 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
14938 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
14939 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
14940 config/sparc/sp64-elf.h, config/spu/spu.h,
14941 config/stormy16/stormy16.h, config/v850/v850.h,
14942 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
14943 define NO_IMPLICIT_EXTERN_C.
14944 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
14946 2018-07-05 Tamar Christina <tamar.christina@arm.com>
14949 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14950 instead of GET_MODE_SIZE when comparing Units.
14952 2018-07-05 Tamar Christina <tamar.christina@arm.com>
14955 * rtlanal.c (set_noop_p): Constrain on mode change,
14956 include hard-reg-set.h
14958 2018-07-05 Tamar Christina <tamar.christina@arm.com>
14960 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
14962 2018-07-05 Jakub Jelinek <jakub@redhat.com>
14965 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
14968 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
14970 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
14972 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
14975 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
14977 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
14979 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
14981 PR middle-end/86380
14982 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
14984 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
14986 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
14987 neg_*_op* variables.
14989 2018-07-04 Martin Liska <mliska@suse.cz>
14991 * tree-switch-conversion.c: Define
14992 max_ratio_for_speed and max_ratio_for_size constants.
14994 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
14995 Martin Liska <mliska@suse.cz>
14997 PR middle-end/66240
15000 * common.opt: Rename align options with 'str_' prefix.
15001 * common/config/i386/i386-common.c (set_malign_value): New
15003 (ix86_handle_option): Use it to set -falign-* options/
15004 * config/aarch64/aarch64-protos.h (struct tune_params): Change
15005 type from int to string.
15006 * config/aarch64/aarch64.c: Update default values from int
15008 * config/alpha/alpha.c (alpha_override_options_after_change):
15010 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
15011 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15012 max skip conditionally.
15013 * config/i386/freebsd.h (SUBALIGN_LOG): New.
15014 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15015 max skip conditionally.
15016 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15017 max skip conditionally.
15018 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
15019 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15020 max skip conditionally.
15021 * config/i386/i386.c (struct ptt): Change type from int to
15023 (ix86_default_align): Set default values.
15024 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
15025 max skip conditionally.
15026 * config/i386/iamcu.h (SUBALIGN_LOG): New.
15027 (ASM_OUTPUT_MAX_SKIP_ALIGN):
15028 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
15029 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15030 max skip conditionally.
15031 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
15032 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
15033 * config/i386/x86-64.h (SUBALIGN_LOG): New.
15034 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
15035 max skip conditionally.
15036 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
15037 * config/ia64/ia64.c (ia64_option_override): Set default values
15038 for alignment options.
15039 * config/m68k/m68k.c: Handle new str_align_* options.
15040 * config/mips/mips.c (mips_set_compression_mode): Change
15042 (mips_option_override): Set default values for options.
15043 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
15045 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15047 * config/rx/rx.c (rx_option_override): Likewise.
15048 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
15049 (LABEL_ALIGN): Use align_labels_log.
15050 (LOOP_ALIGN): Use align_loops_align.
15051 * config/s390/s390.c (s390_asm_output_function_label): Use new
15053 * config/sh/sh.c (sh_override_options_after_change):
15054 Change type of constants.
15055 * config/spu/spu.c (spu_sched_init): Likewise.
15056 * config/sparc/sparc.c (sparc_option_override): Set default
15057 values for options.
15058 * config/visium/visium.c (visium_option_override): Likewise.
15059 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
15060 emit p2align format with last argument if it's not needed.
15061 * doc/invoke.texi: Document extended format of -falign-*.
15062 * final.c: Use align_labels alignment.
15063 * flags.h (struct target_flag_state): Change type to use
15065 (struct align_flags_tuple): New.
15066 (struct align_flags): Likewise.
15067 (align_loops_log): Redefine macro to use new types.
15068 (align_loops_max_skip): Redefine macro to use new types.
15069 (align_jumps_log): Redefine macro to use new types.
15070 (align_jumps_max_skip): Redefine macro to use new types.
15071 (align_labels_log): Redefine macro to use new types.
15072 (align_labels_max_skip): Redefine macro to use new types.
15073 (align_functions_log): Redefine macro to use new types.
15074 (align_loops): Redefine macro to use new types.
15075 (align_jumps): Redefine macro to use new types.
15076 (align_labels): Redefine macro to use new types.
15077 (align_functions): Redefine macro to use new types.
15078 (align_functions_max_skip): Redefine macro to use new types.
15079 (align_loops_value): New macro.
15080 (align_jumps_value): New macro.
15081 (align_labels_value): New macro.
15082 (align_functions_value): New macro.
15083 * function.c (invoke_set_current_function_hook): Propagate
15084 alignment values from flags to global variables default in
15086 * ipa-icf.c (sem_function::equals_wpa): Use
15087 cl_optimization_option_eq instead of memcmp.
15088 * lto-streamer.h (cl_optimization_stream_out): Support streaming
15090 (cl_optimization_stream_in): Likewise.
15091 * optc-save-gen.awk: Support strings in cl_optimization.
15092 * opth-gen.awk: Likewise.
15093 * opts.c (finish_options): Remove error checking of invalid
15095 (MAX_CODE_ALIGN): Remove.
15096 (MAX_CODE_ALIGN_VALUE): Likewise.
15097 (parse_and_check_align_values): New function.
15098 (check_alignment_argument): Likewise.
15099 (common_handle_option): Use check_alignment_argument.
15100 * opts.h (parse_and_check_align_values): Declare.
15101 * toplev.c (init_alignments): Remove.
15102 (read_log_maxskip): New.
15103 (parse_N_M): Likewise.
15104 (parse_alignment_opts): Likewise.
15105 (backend_init_target): Remove usage of init_alignments.
15106 * toplev.h (parse_alignment_opts): Declare.
15107 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
15109 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
15110 * tree.c (cl_option_hasher::equal): New.
15111 * varasm.c: Use new global macros.
15113 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
15115 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
15116 Use a simpler align directive also if MAXSKIP = ALIGN-1.
15117 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15118 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15119 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15120 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
15121 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
15122 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
15123 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
15124 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15125 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15126 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15127 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15129 2018-07-04 Martin Liska <mliska@suse.cz>
15130 Jonathan Wakely <jwakely@redhat.com>
15132 * coverage.c: Use correct type.
15133 * doc/invoke.texi: Language correction.
15135 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
15138 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
15139 ENDBRANCH for non-tail call which may return via indirect branch.
15140 * doc/extend.texi: Document indirect_return attribute.
15142 2018-07-03 Martin Sebor <msebor@redhat.com>
15144 PR tree-optimization/86274
15145 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
15147 (format_floating): Correct handling of infinities and NaNs.
15149 2018-07-03 Martin Sebor <msebor@redhat.com>
15151 * print-tree.c (print_real_cst): New function.
15152 (print_node_brief): Call it.
15153 (print_node): Ditto.
15155 2018-07-03 Jeff Law <law@redhat.com>
15157 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
15158 into a single pattern.
15160 * config/h8300/h8300.md (ors code_iterator): New.
15161 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
15162 a single pattern and single splitter.
15163 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
15164 (iorqi3_1, xorqi3_1): Likewise.
15165 (iorqi3, xorqi3 expanders): Similarly.
15167 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
15168 (movmd_internal) into a single pattern using the P mode iterator.
15169 (movmd splitters): Similarly.
15170 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
15171 (movsd splitters): Similarly.
15173 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
15174 ADDB, ADDW and ADDL into a single ADD attribute which selects the
15175 right table based on the size of the operand.
15176 * config/h8300/h8300.md (length_table): Corresponding changes. All
15177 references to "addb", "addw" and "addl" changed to "add".
15178 (btst patterns): Merge two variants into a single pattern.
15179 (tstqi, tsthi): Likewise.
15180 (addhi3_incdec, addsi3_incdec): Likewise.
15181 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
15182 (mulhi3, mulsi3): Likewise.
15183 (udivhi3, udivsi3): Likewise.
15184 (divhi3, divsi3): Likewise.
15185 (andorqi3, andorhi3, andorsi3): Likewise.
15187 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
15190 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
15191 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
15193 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
15195 PR tree-optimization/85694
15196 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
15197 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
15199 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
15200 (<u>avg<mode>3_ceil): New patterns.
15202 2018-07-03 David Malcolm <dmalcolm@redhat.com>
15204 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
15205 scan-tree-dump directive.
15206 * gcc.dg/vect/slp-perm-2.c: Likewise.
15207 * gcc.dg/vect/slp-perm-3.c: Likewise.
15208 * gcc.dg/vect/slp-perm-5.c: Likewise.
15209 * gcc.dg/vect/slp-perm-6.c: Likewise.
15210 * gcc.dg/vect/slp-perm-7.c: Likewise.
15211 * gcc.dg/vect/slp-perm-8.c: Likewise.
15213 2018-07-03 Marek Polacek <polacek@redhat.com>
15215 PR middle-end/86202
15216 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
15218 2018-07-03 Richard Biener <rguenther@suse.de>
15221 * tree-ssa-structalias.c (find_func_clobbers): Properly
15222 handle indirect calls.
15224 2018-07-03 Jeff Law <law@redhat.com>
15226 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
15227 (shifts): New code iterator.
15228 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
15229 expander. Fix HImode handling on H8/SX.
15230 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
15231 (subqi3, subhi3, subsi3 expanders): Likewise.
15232 (andqi3, andhi3, andsi3 expanders): Likewise.
15233 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
15234 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
15235 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
15236 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
15237 (zero_extendqihi2, zero_extendqisi2): Likewise.
15238 (extendqihi2, extendqisi2): Likewise.
15239 (rotlqi3, rotlhi3, rotlsi3): Likewise.
15240 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
15241 (rotlqi3_1, rotlhi3_1): Likewise.
15242 (logicalhi3_sn, logicalsi3_sn): Likewise.
15243 (logicalhi3, logicalsi3): Likewise.
15245 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
15247 * tree-vect-patterns.c (vect_recog_rotate_pattern)
15248 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
15249 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
15250 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
15251 type to append_pattern_def_seq instead of creating a stmt_vec_info
15253 (build_mask_conversion): Likewise. Remove vinfo argument.
15254 (vect_add_conversion_to_patterm): Likewise, renaming to...
15255 (vect_add_conversion_to_pattern): ...this.
15256 (vect_recog_mask_conversion_pattern): Update call to
15257 build_mask_conversion. Pass the vector type to
15258 append_pattern_def_seq here too.
15259 (vect_recog_gather_scatter_pattern): Update call to
15260 vect_add_conversion_to_pattern.
15262 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
15264 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
15265 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
15266 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
15267 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
15268 STMT_VINFO_PATTERN_DEF_SEQ to null here.
15269 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
15270 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
15271 append_pattern_def_seq instead of new_pattern_def_seq.
15272 (vect_recog_divmod_pattern): Do both of the above.
15273 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
15276 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
15278 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
15279 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
15280 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
15281 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
15282 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
15283 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
15284 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
15285 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
15286 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
15287 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
15288 parameter with a single stmt_vec_info.
15289 (vect_recog_func_ptr): Likewise.
15290 (vect_recog_gather_scatter_pattern): Likewise, folding in...
15291 (vect_try_gather_scatter_pattern): ...this.
15292 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
15293 the stmt_vec_info of the statement to be matched. Don't clear
15294 STMT_VINFO_RELATED_STMT.
15295 (vect_pattern_recog): Update call accordingly.
15297 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
15299 PR tree-optimization/85694
15300 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
15301 (uavgM3_ceil): Document new optabs.
15302 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
15303 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
15305 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
15306 (savg_ceil_optab): New optabs.
15307 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
15308 (vect_vect_recog_func_ptrs): Add it.
15309 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
15310 constant directly from the associated lhs.
15312 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
15314 * tree-vect-patterns.c (vect_split_statement): New function.
15315 (vect_convert_input): Use it to try to split an existing cast.
15317 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
15319 * poly-int.h (print_hex): New function.
15320 * dumpfile.h (dump_dec, dump_hex): Declare.
15321 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
15322 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
15323 min_input_precision, operation_precision and operation_sign.
15324 * tree-vect-patterns.c (vect_get_range_info): New function.
15325 (vect_same_loop_or_bb_p, vect_single_imm_use)
15326 (vect_operation_fits_smaller_type): Delete.
15327 (vect_look_through_possible_promotion): Add an optional
15328 single_use_p parameter.
15329 (vect_recog_over_widening_pattern): Rewrite to use new
15330 stmt_vec_info infomration. Handle one operation at a time.
15331 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
15332 (vect_truncatable_operation_p, vect_set_operation_type)
15333 (vect_set_min_input_precision): New functions.
15334 (vect_determine_min_output_precision_1): Likewise.
15335 (vect_determine_min_output_precision): Likewise.
15336 (vect_determine_precisions_from_range): Likewise.
15337 (vect_determine_precisions_from_users): Likewise.
15338 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
15339 (vect_vect_recog_func_ptrs): Put over_widening first.
15341 (vect_pattern_recog): Call vect_determine_precisions.
15343 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
15345 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
15346 statements that have been replaced by further pattern statements.
15347 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
15349 2018-07-03 Richard Biener <rguenther@suse.de>
15351 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
15352 always set *dt. Dump vectype in vectype overload.
15353 * dumpfile.h (dump_gimple_expr): New function.
15354 (dump_gimple_expr_loc): Likewise.
15355 * dumpfile.c (dump_gimple_expr): New function.
15356 (dump_gimple_expr_loc): Likewise.
15358 2018-07-02 Jeff Law <law@redhat.com>
15360 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
15361 the H8/300, H8/300H and H8/S variants into a single pattern.
15362 (movhi_h8300, movqi_h8300hs): Similarly.
15363 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
15364 (QHI mode iterator): New.
15366 * config/h8300/h8300.md: Remove trailing whitespace.
15368 2018-07-02 Jim Wilson <jimw@sifive.com>
15370 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
15371 instead of emit_insn for interrupt returns.
15372 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
15373 (riscv_sret, riscv_uret): Likewise.
15375 2018-07-02 David Malcolm <dmalcolm@redhat.com>
15377 * pretty-print.c (selftest::test_pp_format): Move save and restore
15378 of quotes to class auto_fix_quotes, and add an instance.
15379 * selftest.c: Include "intl.h".
15380 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
15381 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
15382 * selftest.h (selftest::auto_fix_quotes): New class.
15384 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
15386 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
15387 (aarch64_sve_prepare_conditional_op): Remove.
15388 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
15389 Allow aarch64_simd_reg_or_zero as select operand; remove
15390 the aarch64_sve_prepare_conditional_op call.
15391 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
15392 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
15393 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
15394 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
15395 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
15396 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
15397 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
15398 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
15399 and a splitters to match all of the *_any patterns.
15400 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
15402 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
15403 (SVE_COND_FP_BINARY_REV): Remove.
15404 (sve_int_op_rev, sve_fp_op_rev): New.
15405 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
15406 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
15407 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
15408 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
15409 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
15410 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
15411 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
15412 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
15413 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
15415 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
15416 Remove match_dup 1 from the inner unspec.
15417 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
15419 * config/aarch64/aarch64.md (movprfx): New attr.
15420 (length): Default movprfx to 8.
15421 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
15422 (*madd<SVE_I>, *msub<SVE_I): Likewise.
15423 (*<su>mul<SVE_I>3_highpart): Likewise.
15424 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
15425 (*v<ASHIFT><SVE_I>3): Likewise.
15426 (*<su><MAXMIN><SVE_I>3): Likewise.
15427 (*<su><MAXMIN><SVE_F>3): Likewise.
15428 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
15429 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
15430 (*div<SVE_F>4): Likewise.
15432 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
15434 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
15437 2018-07-02 Richard Biener <rguenther@suse.de>
15439 PR tree-optimization/86363
15440 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
15441 memset argument refers to a non-variable address.
15443 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
15445 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
15446 {PLUS,MINUS}_EXPR code to...
15447 (adjust_symbolic_bound): ...here,
15448 (combine_bound): ...here,
15449 (set_value_range_with_overflow): ...and here.
15451 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
15453 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
15455 (extract_range_from_abs_expr): ...here.
15457 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
15459 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
15460 -fno-omit-frame-pointer when not optimizing.
15462 2018-07-02 Martin Liska <mliska@suse.cz>
15465 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
15466 (propagate_nothrow): Likewise.
15468 2018-07-02 Martin Liska <mliska@suse.cz>
15471 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
15473 2018-07-02 David Malcolm <dmalcolm@redhat.com>
15475 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
15476 function in r262149, changing "loc" param from source_location to
15477 const dump_location_t &.
15478 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
15479 declaration, as above.
15481 2018-07-01 Paul Koning <ni1d@arrl.net>
15483 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
15484 -munit-asm, -mgnu-asm, -mdec-asm.
15485 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
15486 (pdp11_output_labelref): New.
15487 (pdp11_output_def): New.
15488 (pdp11_output_addr_vec_elt): New.
15489 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
15490 %# and %@ format codes.
15491 (pdp11_option_override): New.
15492 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
15493 (pdp11_output_ident): New.
15494 (pdp11_asm_named_section): New.
15495 (pdp11_asm_init_sections): New.
15496 (pdp11_file_start): New.
15497 (pdp11_file_end): New.
15498 (output_ascii): Use .ascii/.asciz for -mdec-asm.
15499 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
15500 %o, like %c but octal.
15501 (pdp11_option_override): New.
15502 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
15504 (DATA_SECTION_ASM_OP): Ditto.
15505 (READONLY_DATA_SECTION_ASM_OP): New.
15506 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
15507 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
15508 (ASM_OUTPUT_LABELREF): Ditto.
15509 (ASM_OUTPUT_DEF): Ditto.
15510 (ASM_OUTPUT_EXTERNAL): New.
15511 (ASM_OUTPUT_SOURCE_FILENAME): New.
15512 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
15513 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
15514 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
15515 %# and %@ format codes.
15516 * config/pdp11/pdp11.opt (mgnu-asm): New.
15517 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
15518 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
15519 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
15521 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
15523 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
15524 dereferencing path[] beyond its length.
15526 (debug_all_paths): New.
15527 (rewire_first_differing_edge): New.
15528 (adjust_paths_after_duplication): New.
15529 (duplicate_thread_path): Call adjust_paths_after_duplication.
15531 (thread_through_all_blocks): Add new argument to
15532 duplicate_thread_path.
15534 2018-06-30 Jim Wilson <jimw@sifive.com>
15536 * config/riscv/predicates.md (p2m1_shift_operand): New.
15537 (high_mask_shift_operand): New.
15538 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
15539 pattern using p2m1_shift_operand.
15540 (lshsi3_zero_extend_3+2): New combiner pattern using
15541 high_mask_shift_operand.
15543 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
15545 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
15547 (vect_recog_rotate_pattern): ...here.
15548 (vect_convert_input): Try to insert casts of invariants in the
15550 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
15551 preheader to be empty.
15553 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
15555 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
15556 vector type. If given, install it in the new statement's
15557 STMT_VINFO_VECTYPE.
15558 (vect_element_precision): New function.
15559 (vect_unpromoted_value): New struct.
15560 (vect_unpromoted_value::vect_unpromoted_value): New function.
15561 (vect_unpromoted_value::set_op): Likewise.
15562 (vect_look_through_possible_promotion): Likewise.
15563 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
15564 (vect_widened_op_tree, vect_convert_input): Likewise.
15565 (vect_convert_inputs, vect_convert_output): Likewise.
15566 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
15567 to handle the optional cast of the multiplication result and
15568 vect_widened_op_tree to detect the widened multiplication itself.
15569 Do not require the input and output of promotion casts to have
15570 the same sign, but base the signedness of the operation on the
15571 input rather than the result. If the pattern includes two
15572 promotions, check that those promotions have the same sign.
15573 Do not restrict the MULT_EXPR handling to a double-width result;
15574 handle quadruple-width results and wider. Use vect_convert_inputs
15575 to convert the inputs to the common type.
15576 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
15577 to handle the optional cast of the ABS result. Also allow a sign
15578 change or a sign extension between the ABS and MINUS.
15579 Use vect_widened_op_tree to detect the widened subtraction and use
15580 vect_convert_inputs to convert the inputs to the common type.
15581 (vect_handle_widen_op_by_const): Delete.
15582 (vect_recog_widen_op_pattern): New function.
15583 (vect_recog_widen_mult_pattern): Use it.
15584 (vect_recog_widen_shift_pattern): Likewise.
15585 (vect_recog_widen_sum_pattern): Use
15586 vect_look_through_possible_promotion to handle the promoted
15589 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
15591 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
15592 the containing gimple_seq *.
15593 * gimple-iterator.h (gsi_for_stmt): Declare it.
15594 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
15595 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
15596 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
15597 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
15598 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
15600 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
15602 (vect_mark_pattern_stmts): ...here. Handle cases in which the
15603 statement being replaced is part of an existing pattern
15604 definition sequence, inserting the new pattern statements before
15606 (vect_pattern_recog_1): Don't return a bool. If the statement
15607 is already part of a pattern, instead apply pattern matching
15608 to the pattern definition statements. Don't clear the
15609 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
15610 (vect_pattern_recog): Don't break after the first match;
15611 continue processing the pattern definition statements instead.
15612 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
15614 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
15616 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
15617 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
15618 (vect_recog_widen_sum_pattern): Use it.
15620 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
15622 * tree-vect-loop.c (vectorizable_reduction): Assert that the
15623 phi is not a pattern statement and has not been replaced by
15624 a pattern statement.
15625 * tree-vect-patterns.c (type_conversion_p): Don't check
15626 STMT_VINFO_IN_PATTERN_P.
15627 (vect_recog_vector_vector_shift_pattern): Likewise.
15628 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
15629 the pattern statement rather than the original statement; check
15630 directly for a WIDEN_MULT_EXPR here.
15631 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
15632 vect_is_simple_use to return the pattern statement rather
15633 than the original statement; use is_pattern_stmt_p to check
15634 for such a pattern statement.
15635 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
15636 to return the pattern statement rather than the original statement;
15637 don't do the same transformation here.
15638 (vect_is_simple_use): If the defining statement has been replaced
15639 by a pattern statement, return the pattern statement instead.
15640 Remove the corresponding (local) transformation from the vectype
15643 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
15645 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
15646 end and default to null.
15647 * tree-vect-loop.c (vect_create_epilog_for_reduction)
15648 (vectorizable_reduction): Update calls accordingly, dropping the
15649 gimple ** argument if the passed-back statement isn't needed.
15650 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
15651 (vect_recog_rotate_pattern): Likewise.
15652 (vect_recog_mask_conversion_pattern): Likewise.
15653 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
15654 (vect_mask_constant_operand_p): Likewise.
15655 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
15656 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
15657 (get_group_load_store_type, get_load_store_type): Likewise.
15658 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
15659 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
15660 (vectorizable_conversion, vectorizable_assignment): Likewise.
15661 (vectorizable_shift, vectorizable_operation): Likewise.
15662 (vectorizable_store, vect_is_simple_cond): Likewise.
15663 (vectorizable_condition, vectorizable_comparison): Likewise.
15664 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
15665 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
15666 and move it to the end. Cope with null def_stmt_outs.
15668 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
15670 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
15672 2018-06-29 Jeff Law <law@redhat.com>
15674 * config/v850/v850.c (v850_legitimate_address_p): Handle large
15675 displacements for TARGET_V850E2V3 and newer.
15676 (TARGET_LRA_P): Remove. Defaults to LRA now.
15677 * config/v850/v850.md (sign23byte_load): Remove.
15678 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
15679 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
15681 2018-06-29 Martin Liska <mliska@suse.cz>
15684 * coverage.c (coverage_init): Mangle full path name.
15685 * doc/invoke.texi: Document the change.
15686 * gcov-io.c (mangle_path): New.
15687 * gcov-io.h (mangle_path): Likewise.
15688 * gcov.c (mangle_name): Use mangle_path for path mangling.
15690 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15692 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
15693 if starting source register is not even.
15695 2018-06-29 Martin Liska <mliska@suse.cz>
15697 PR tree-optimization/86263
15698 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
15699 Make edge redirection.
15701 2018-06-29 David Malcolm <dmalcolm@redhat.com>
15703 * dumpfile.c (dump_loc): Add indentation based on scope depth.
15704 (dump_scope_depth): New variable.
15705 (get_dump_scope_depth): New function.
15706 (dump_begin_scope): New function.
15707 (dump_end_scope): New function.
15708 * dumpfile.h (get_dump_scope_depth): New declaration.
15709 (dump_begin_scope): New declaration.
15710 (dump_end_scope): New declaration.
15711 (class auto_dump_scope): New class.
15712 (AUTO_DUMP_SCOPE): New macro.
15713 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
15716 2018-06-29 Richard Biener <rguenther@suse.de>
15718 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
15719 compute_all_dependences succeeds.
15720 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
15721 exceed --param loop-max-datarefs-for-datadeps.
15723 2018-06-29 Jakub Jelinek <jakub@redhat.com>
15725 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
15727 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
15730 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
15731 alternative 0 in preferred_for_speed attribute.
15733 2018-06-28 Paul Koning <ni1d@arrl.net>
15735 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
15736 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
15737 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
15738 * config/pdp11/pdp11.md: Correct "length" attribute calculation
15739 for shift insn patterns.
15741 2018-06-28 David Malcolm <dmalcolm@redhat.com>
15743 * cgraph.c (cgraph_node::get_body): Replace assignments to
15744 "dump_file" with calls to set_dump_file.
15745 * dumpfile.c (alt_dump_file): Make static, and group with...
15746 (alt_flags): ...this definition.
15747 (dumps_are_enabled): New variable.
15748 (refresh_dumps_are_enabled): New function.
15749 (set_dump_file): New function.
15750 (set_alt_dump_file): New function.
15751 (gcc::dump_manager::dump_start): Replace assignments to
15752 "dump_file" and "alt_dump_file" with calls to set_dump_file and
15754 (gcc::dump_manager::dump_finish): Likewise.
15755 * dumpfile.h (alt_dump_file): Delete decl.
15756 (dumps_are_enabled): New variable decl.
15757 (set_dump_file): New function decl.
15758 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
15760 * tree-nested.c (lower_nested_functions): Replace assignments to
15761 "dump_file" with calls to set_dump_file.
15763 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
15765 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
15766 goto_locus of each outgoing edge of each basic block.
15768 2018-06-28 Richard Biener <rguenther@suse.de>
15770 * dwarf2out.c (decl_scope_table): Remove.
15771 (push_decl_scope): Likewise.
15772 (pop_decl_scope): Likewise.
15773 (gen_type_die_for_member): Do not call push/pop_decl_scope.
15774 (gen_struct_or_union_type_die): Likewise.
15775 (gen_tagged_type_die): Likewise.
15776 (dwarf2out_init): Do not initialize decl_scope_table.
15777 (dwarf2out_c_finalize): Do not free it.
15779 2018-06-28 Richard Biener <rguenther@suse.de>
15781 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
15782 deciding whether to not re-use a DIE.
15784 2018-06-28 Richard Biener <rguenther@suse.de>
15786 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
15787 DW_AT_abstract_origin attribute.
15789 2018-06-28 Martin Liska <mliska@suse.cz>
15791 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
15792 Use newly introduced constants.
15793 * tree-switch-conversion.h (struct jump_table_cluster):
15794 Define max_ratio_for_size and max_ratio_for_speed.
15796 2018-06-28 Martin Liska <mliska@suse.cz>
15798 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15799 Add new checking assert to catch invalid state.
15800 (jump_table_cluster::can_be_handled): Handle single case
15802 (jump_table_cluster::is_beneficial): Bail out for such case.
15803 (bit_test_cluster::find_bit_tests):
15804 Add new checking assert to catch invalid state.
15805 (bit_test_cluster::can_be_handled): Handle single case
15807 (bit_test_cluster::is_beneficial): Bail out for such case.
15808 (switch_decision_tree::analyze_switch_statement):
15811 2018-06-28 Martin Liska <mliska@suse.cz>
15813 * common.opt: Introduce -completion option.
15814 * gcc.c (driver_handle_option): Handle it.
15815 (driver::main): Print completions if completion
15817 * opt-suggestions.c (option_proposer::get_completions):
15819 (option_proposer::suggest_completion): Likewise.
15820 (option_proposer::find_param_completions): Likewise.
15821 (verify_autocompletions): Likewise.
15822 (test_completion_valid_options): Likewise.
15823 (test_completion_valid_params): Likewise.
15824 (in_completion_p): Likewise.
15825 (empty_completion_p): Likewise.
15826 (test_completion_partial_match): Likewise.
15827 (test_completion_garbage): Likewise.
15828 (opt_proposer_c_tests): Likewise.
15829 * opt-suggestions.h: Declare new functions.
15830 * opts.c (common_handle_option): Handle OPT__completion_.
15831 * selftest-run-tests.c (selftest::run_tests): Add
15832 opt_proposer_c_tests.
15833 * selftest.c (assert_str_startswith): New.
15834 * selftest.h (assert_str_startswith): Likewise.
15835 (opt_proposer_c_tests): New.
15836 (ASSERT_STR_STARTSWITH): Likewise.
15838 2018-06-28 Martin Liska <mliska@suse.cz>
15840 * Makefile.in: Add opt-suggestions.o.
15841 * gcc-main.c: Include opt-suggestions.h.
15842 * gcc.c (driver::driver): Likewise.
15843 (driver::~driver): Remove m_option_suggestions.
15844 (driver::build_option_suggestions): Moved to option_proposer.
15845 (driver::suggest_option): Likewise.
15846 (driver::handle_unrecognized_options): Use option_proposer.
15847 * gcc.h (class driver): Add new memver m_option_proposer.
15848 * opt-suggestions.c: New file.
15849 * opt-suggestions.h: New file.
15851 2018-06-28 Martin Liska <mliska@suse.cz>
15853 * vec.h (class auto_string_vec): New (moved from auto_argvec).
15854 (auto_string_vec::~auto_string_vec): Likewise.
15856 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
15858 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
15859 prevent_decl_creation_for_types fields up and add reset_location field.
15860 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
15861 statement if id->reset_location is true.
15862 (copy_edges_for_bb): Do not set goto_locus on the new edges if
15863 id->reset_location is true.
15864 (copy_phis_for_bb): Force input_location on the arguments if
15865 id->reset_location is true.
15866 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
15867 is set on the function to be inlined.
15869 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
15871 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
15873 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
15875 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
15876 registers for Pmode.
15877 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
15878 hard registers for the clobbered pseudo.
15880 2018-06-27 Paul Koning <ni1d@arrl.net>
15882 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
15883 mutually exclusive options.
15884 * config/pdp11/constraints.md (h): New constraint.
15885 (O): Update definition to match shift code generation.
15886 (D): New constraint.
15887 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
15889 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
15891 (output_jump): Change arguments.
15892 (pdp11_fixed_cc_regs): New function.
15893 (pdp11_cc_mode): Ditto.
15894 (pdp11_expand_shift): Ditto.
15895 (pdp11_assemble_shift): Ditto.
15896 (pdp11_small_shift): Ditto.
15897 (pdp11_branch_cost): Remove.
15898 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
15900 (pdp11_register_move_cost): Update for CC registers.
15901 (pdp11_rtx_costs): Add case for LSHIFTRT.
15902 (pdp11_output_jump): Add CCNZ mode conditional branches.
15903 (notice_update_cc_on_set): Remove.
15904 (pdp11_cc_mode): New function.
15905 (simple_memory_operand): Correct pre/post decrement case.
15906 (no_side_effect_operand): New function.
15907 (pdp11_regno_reg_class): Add CC_REGS class.
15908 (pdp11_fixed_cc_regs): New function.
15909 (pdp11_small_shift): New function.
15910 (pdp11_expand_shift): New function to expand shift insns.
15911 (pdp11_assemble_shift): New function to output shifts.
15912 (pdp11_branch_cost): Remove.
15913 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
15914 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
15915 (WCHAR_TYPE): Ditto.
15916 (PTRDIFF_TYPE): Ditto.
15917 (ADJUST_INSN_LENGTH): New macro.
15918 (FIXED_REGISTERS): Add CC registers.
15919 (CALL_USED_REGISTERS): Ditto.
15920 (reg_class): Ditto.
15921 (REG_CLASS_NAMES): Ditto.
15922 (REG_CLASS_CONTENTS): Ditto.
15923 (SELECT_CC_MODE): Use new function.
15924 (TARGET_FLAGS_REGNUM): New macro.
15925 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
15926 (cc0_reg_rtx): Remove.
15927 (CC_STATUS_MDEP): Remove.
15928 (CC_STATUS_MDEFP_INIT): Remove.
15929 (CC_IN_FPU): Remove.
15930 (NOTICE_UPDATE_CC): Remove.
15931 (REGISTER_NAMES): Add CC registers.
15932 (BRANCH_COST): Change to constant 1.
15933 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
15935 * config/pdp11/pdp11.opt (mbcopy): Remove.
15936 (mbcopy-builtin): Remove.
15937 (mbranch-cheap): Remove.
15938 (mbranch-expensive): Remove.
15939 * config/pdp11/predicates.md (expand_shift_operand): Update to
15940 match shift code generation.
15941 (ccnz_operator): New predicate.
15942 * doc/invoke.texi (PDP-11 Options): Remove deleted options
15943 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
15944 Remove non-existent option -mabshi, -mno-abshi. Document mutually
15946 * doc/md.texi (PDP-11): Document new D and h constraints. Update
15947 description of O constraint.
15949 2018-06-27 Jeff Law <law@redhat.com>
15950 Austin Law <austinklaw@gmail.com>
15952 * config/v850/v850.md (addsi3_set_flags): New pattern.
15953 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
15954 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
15955 (zero_extendhisi2_v850_set_flags): Likewise.
15956 (zero_extendqisi2_v850_set_flags): Likewise.
15957 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
15958 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
15959 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
15961 * config/v850/v850-protos.h (notice_update_cc): Remove.
15962 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
15963 (v850_print_operand): Handle 'D' and "d".
15964 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
15965 Add handling of arithmetic/logical operations compared against zero.
15966 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
15967 Do not look at v850_compare_op, instead get mode from last argument.
15968 (v850_gen_compare): Remove
15969 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
15970 after reload for prologue insns.
15971 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
15973 (construct_save_jarl): Likewise.
15974 (TARGET_FLAGS_REGNUM): Define.
15975 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
15976 (NOTICE_UPDATE_CC): Remove.
15977 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
15978 than cc0. Conditionalize on reload_completed.
15979 (cmpsi_insn, setfcc_insn): Likewise.
15980 (tst1 splitter): Turn into define_and_split which sets the flags
15982 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
15983 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
15984 (cstoresf4, cstoredf4): Clobber the flags.
15985 (cmpsi, cmpsf, cmpdf): Remove expanders.
15986 (setf_insn): Remove pattern.
15987 (addsi3): Turn into define_and_split which clobbers the flags after
15988 reload and a suitable pattern (addsi3_clobber_flags) for use after
15990 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
15991 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
15992 (ashrsi3, ashrsi3_v850e2): Likewise.
15993 (bins): Clobber the flags.
15994 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
15995 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
15996 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
15997 (call_value_internal_short, call_value_internal_long): Likewise.
15998 (callt_save_interrupt, callt_return_interrupt): Likewise.
15999 (save_interrupt, return_interrupt): Likewise.
16000 (callt_save_all_interrupt, save_all_interrupt): Likewise.
16001 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
16002 (restore_all_interrupt, _restore_all_interrupt): Likewise.
16003 (All FP comparisons): Only allow after reload has completed.
16005 (divh, divhu): Tweak output template.
16006 (branch_z_normal, branch_z_invert): Remove
16007 (branch_nz_normal, branch_nz_invert): Likewise.
16008 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
16010 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
16011 * config/v850/v850.c (notice_update_cc): Remove.
16012 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
16013 (CC_NO_CARRY): Likewise.
16014 (NOTICE_UPDATE_CC): Define to nothing.
16015 * config/v850/v850.md: Remove block comment on cc0 handling
16016 Remove "cc" attribute from all patterns. Remove cc_status handling
16017 from all patterns. Minor formatting fixes.
16019 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16021 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
16022 (cortex-a76.cortex-a55): Likewise.
16023 * config/aarch64/aarch64-tune.md: Regenerate.
16024 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
16025 cortex-a76.cortex-a55.
16027 2018-06-27 Jeff Law <law@redhat.com>
16029 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
16030 (MULTILIB_DIRNAMES): Similarly.
16032 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
16034 * gimple.h (gimple_return_retbnd): Delete.
16035 (gimple_return_set_retbnd): Likewise.
16036 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
16037 gimple_return_set_retbnd.
16038 * gimple-pretty-print.c (dump_gimple_return): Remove call to
16039 gimple_return_retbnd and adjust.
16040 * tree-inline.h (struct copy_body_data): Remove retbnd field.
16041 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
16042 Explicitly return NULL in a couple more cases. Move assertion
16043 on debug statements and remove unreachable code.
16044 (reset_debug_binding): Do not test id->retbnd.
16045 (expand_call_inline): Do not set it.
16047 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
16049 * configure.ac: Add --disable-gcov option.
16050 * configure: Regenerate.
16051 * Makefile.in: Honour @enable_gcov@.
16052 * doc/install.texi: Document --disable-gcov.
16054 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16056 * config/arm/arm-cpus.in (cortex-a76): New entry.
16057 (cortex-a76.cortex-a55): Likewise.
16058 * config/arm/arm-tables.opt: Regenerate.
16059 * config/arm/arm-tune.md: Likewise.
16060 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
16061 * doc/invoke.texi (ARM Options): Document cortex-a76 and
16062 cortex-a76.cortex-a55.
16064 2018-06-27 Tamar Christina <tamar.christina@arm.com>
16067 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
16069 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
16071 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
16073 (EPILOGUE_USES): Likewise.
16075 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
16077 * tree-inline.c (remap_location): New function extracted from...
16078 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
16079 (copy_phis_for_bb): ...here. Call remap_location.
16080 (copy_cfg_body): Adjust call to copy_edges_for_bb.
16082 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
16084 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
16085 unaligned vsx for 16B memset.
16087 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
16090 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
16091 ieee128_float_type_node to long_double_type_node unless
16092 TARGET_LONG_DOUBLE_128 is set.
16094 2018-06-26 David Malcolm <dmalcolm@redhat.com>
16096 * cfgloop.c (get_loop_location): Convert return type from
16097 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
16098 by implicit construction from rtx_insn *, and using
16099 dump_user_location_t::from_function_decl for the fallback case.
16100 * cfgloop.h (get_loop_location): Convert return type from
16101 location_t to dump_user_location_t.
16102 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
16103 dump_printf_loc to pass in a dump_location_t rather than a
16104 location_t, via the gimple stmt.
16105 * coverage.c (get_coverage_counts): Update calls to
16106 dump_printf_loc to pass in dump_location_t rather than a
16108 * doc/optinfo.texi (Dump types): Convert example of
16109 dump_printf_loc from taking "locus" to taking "insn". Update
16110 description of the "_loc" calls to cover dump_location_t.
16111 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
16113 (dump_user_location_t::dump_user_location_t): New constructors,
16114 from gimple *stmt and rtx_insn *.
16115 (dump_user_location_t::from_function_decl): New function.
16116 (dump_loc): Make static.
16117 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
16118 const dump_location_t &.
16119 (dump_generic_expr_loc): Delete.
16120 (dump_printf_loc): Convert param "loc" from location_t to
16121 const dump_location_t &.
16122 (selftest::test_impl_location): New function.
16123 (selftest::dumpfile_c_tests): New function.
16124 * dumpfile.h: Include "profile-count.h".
16125 (class dump_user_location_t): New class.
16126 (struct dump_impl_location_t): New struct.
16127 (class dump_location_t): New class.
16128 (dump_printf_loc): Convert 2nd param from source_location to
16129 const dump_location_t &.
16130 (dump_generic_expr_loc): Delete.
16131 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
16132 const dump_location_t &.
16133 * gimple-fold.c (fold_gimple_assign): Update call to
16134 dump_printf_loc to pass in a dump_location_t rather than a
16135 location_t, via the gimple stmt.
16136 (gimple_fold_call): Likewise.
16137 * gimple-loop-interchange.cc
16138 (loop_cand::analyze_iloop_reduction_var): Update for change to
16139 check_reduction_path.
16140 (tree_loop_interchange::interchange): Update for change to
16141 find_loop_location.
16142 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
16143 change in return-type of find_loop_location.
16144 (graphite_regenerate_ast_isl): Likewise.
16145 * graphite-optimize-isl.c (optimize_isl): Likewise.
16146 * graphite.c (graphite_transform_loops): Likewise.
16147 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
16148 pass in a dump_location_t rather than a location_t, via the
16150 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16151 * ipa.c (walk_polymorphic_call_targets): Likewise.
16152 * loop-unroll.c (report_unroll): Convert "locus" param from
16153 location_t to dump_location_t.
16154 (decide_unrolling): Update for change to get_loop_location's
16156 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
16157 location_t to dump_user_location_t.
16158 (grid_find_single_omp_among_assignments_1): Updates calls to
16159 dump_printf_loc to pass in a dump_location_t rather than a
16160 location_t, via the gimple stmt.
16161 (grid_parallel_clauses_gridifiable): Convert "tloc" from
16162 location_t to dump_location_t. Updates calls to dump_printf_loc
16163 to pass in a dump_location_t rather than a location_t, via the
16165 (grid_inner_loop_gridifiable_p): Likewise.
16166 (grid_dist_follows_simple_pattern): Likewise.
16167 (grid_gfor_follows_tiling_pattern): Likewise.
16168 (grid_target_follows_gridifiable_pattern): Likewise.
16169 (grid_attempt_target_gridification): Convert initialization
16170 of local "grid" from memset to zero-initialization; FIXME: does
16171 this require C++11? Update call to dump_printf_loc to pass in a
16172 optinfo_location rather than a location_t, via the gimple stmt.
16173 * profile.c (read_profile_edge_counts): Updates call to
16174 dump_printf_loc to pass in a dump_location_t rather than a
16176 (compute_branch_probabilities): Likewise.
16177 * selftest-run-tests.c (selftest::run_tests): Call
16179 * selftest.h (dumpfile_c_tests): New decl.
16180 * tree-loop-distribution.c (pass_loop_distribution::execute):
16181 Update for change in return type of find_loop_location.
16182 * tree-parloops.c (parallelize_loops): Likewise.
16183 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
16184 "locus" from location_t to dump_user_location_t.
16185 (canonicalize_loop_induction_variables): Likewise.
16186 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
16187 for change in return type of find_loop_location.
16188 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
16189 to dump_printf_loc to pass in a dump_location_t rather than a
16190 location_t, via the stmt.
16191 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
16193 * tree-vect-loop-manip.c (find_loop_location): Convert return
16194 type from source_location to dump_user_location_t.
16195 (vect_do_peeling): Update for above change.
16196 (vect_loop_versioning): Update for change in type of
16198 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
16199 from location_t to dump_user_location_t.
16200 (vect_estimate_min_profitable_iters): Update for change in type
16202 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
16203 location_t to dump_location_t.
16204 (vect_slp_bb): Update for change in type of vect_location.
16205 * tree-vectorizer.c (vect_location): Convert from source_location
16206 to dump_user_location_t.
16207 (try_vectorize_loop_1): Update for change in vect_location's type.
16208 (vectorize_loops): Likewise.
16209 (increase_alignment): Likewise.
16210 * tree-vectorizer.h (vect_location): Convert from source_location
16211 to dump_user_location_t.
16212 (find_loop_location): Convert return type from source_location to
16213 dump_user_location_t.
16214 (check_reduction_path): Convert 1st param from location_t to
16215 dump_user_location_t.
16216 * value-prof.c (check_counter): Update call to dump_printf_loc to
16217 pass in a dump_user_location_t rather than a location_t; update
16218 call to error_at for change in type of "locus".
16219 (check_ic_target): Update call to dump_printf_loc to
16220 pass in a dump_user_location_t rather than a location_t, via the
16223 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
16225 * config/s390/s390.h (enum processor_flags): Do not use
16226 default tune parameter when -march was specified.
16228 2018-06-26 Jakub Jelinek <jakub@redhat.com>
16231 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
16232 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
16235 2018-06-26 Richard Biener <rguenther@suse.de>
16237 PR tree-optimization/86287
16239 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
16240 (vect_analyze_loop): Initialize n_stmts.
16242 2018-06-26 Richard Biener <rguenther@suse.de>
16244 PR middle-end/86271
16245 * fold-const.c (fold_convertible_p): Pointer extension
16248 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
16251 * dwarf2out.c (loc_list_has_views): Adjust comments.
16252 (dw_loc_list): Split single cross-partition range with
16255 2018-06-25 Jeff Law <law@redhat.com>
16257 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
16258 on -mbig-switch by default.
16260 * config/v850/predicates.md (const_float_1_operand): Fix match_code
16262 (const_float_0_operand): Remove unused predicate.
16263 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
16264 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
16265 (recipsf2): New expander. Original pattern now called
16267 (recipdf2, recipdf2_insn): Similarly.
16268 (rsqrtsf2, rsqrtsf2_insn): Similarly
16269 (rsqrtdf2, rsqrtdf2_insn): Similarly
16271 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
16273 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
16274 Simplify logic for FreeBSD (twice).
16276 2018-06-25 Martin Sebor <msebor@redhat.com>
16278 PR tree-optimization/86204
16279 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
16280 a strnlen result if it's less than the length of the string.
16282 2018-06-25 Martin Sebor <msebor@redhat.com>
16284 PR tree-optimization/85700
16285 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
16286 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
16287 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
16289 2018-06-25 Martin Sebor <msebor@redhat.com>
16291 * doc/extend.texi (Zero-length arrays): Update and clarify.
16293 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
16295 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
16296 added IEEE/IBM long double multilib support on PowerPC little
16297 endian Linux systems.
16298 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
16299 (MULTILIB_DEFAULTS): Likewise.
16300 * config/rs6000/rs6000.c (rs6000_option_override_internal):
16302 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
16303 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
16304 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
16306 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
16308 PR middle-end/86311
16309 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
16310 (REORDER_45): Likewise.
16312 2018-06-25 Jeff Law <law@redhat.com>
16314 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
16315 dividend to 32 bits. Adjust length.
16316 (udivmodhi4): Cleanup output template. Fix length.
16318 2018-06-25 Carl Love <cel@us.ibm.com>
16320 * config/rs6000/vsx.md: Change word selector to prefered location.
16322 2018-06-25 Richard Biener <rguenther@suse.de>
16324 PR tree-optimization/86304
16325 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
16326 epilogue-if-converted loops as well.
16328 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
16330 * lto-section-out.c (lto_begin_section): Do not print section
16331 name for noaddr and unnumbered dumps.
16333 2018-06-25 Richard Biener <rguenther@suse.de>
16335 * tree-vectorizer.h (struct vec_info_shared): New structure
16336 with parts split out from struct vec_info and loop_nest from
16337 struct _loop_vec_info.
16338 (struct vec_info): Adjust accordingly.
16339 (struct _loop_vec_info): Likewise.
16340 (LOOP_VINFO_LOOP_NEST): Adjust.
16341 (LOOP_VINFO_DATAREFS): Likewise.
16342 (LOOP_VINFO_DDRS): Likewise.
16343 (struct _bb_vec_info): Likewise.
16344 (BB_VINFO_DATAREFS): Likewise.
16345 (BB_VINFO_DDRS): Likewise.
16346 (struct _stmt_vec_info): Add dr_aux member.
16347 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
16348 (DR_MISALIGNMENT_UNINITIALIZED): New.
16349 (set_dr_misalignment): Adjust.
16350 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
16351 (vect_analyze_loop): Adjust prototype.
16352 (vect_analyze_loop_form): Likewise.
16353 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
16354 Compute dependences lazily.
16355 (vect_record_base_alignments): Use shared datarefs/ddrs.
16356 (vect_verify_datarefs_alignment): Likewise.
16357 (vect_analyze_data_refs_alignment): Likewise.
16358 (vect_analyze_data_ref_accesses): Likewise.
16359 (vect_analyze_data_refs): Likewise.
16360 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
16361 constructor parameter for shared part.
16362 (vect_analyze_loop_form): Pass in shared part and adjust.
16363 (vect_analyze_loop_2): Pass in storage for the number of
16364 stmts. Move loop nest finding to the caller. Compute
16366 (vect_analyze_loop): Pass in shared part.
16367 (vect_transform_loop): Verify shared datarefs are unchanged.
16368 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
16369 constructor parameter for shared part.
16370 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
16371 (vect_slp_bb): Verify shared datarefs are unchanged before
16373 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
16375 (new_stmt_vec_info): Initialize DR_AUX misalignment to
16376 DR_MISALIGNMENT_UNINITIALIZED.
16377 * tree-vectorizer.c (vec_info::vec_info): Add constructor
16378 parameter for shared part.
16379 (vec_info::~vec_info): Adjust.
16380 (vec_info_shared::vec_info_shared): New.
16381 (vec_info_shared::~vec_info_shared): Likewise.
16382 (vec_info_shared::save_datarefs): Likewise.
16383 (vec_info_shared::check_datarefs): Likewise.
16384 (try_vectorize_loop_1): Construct shared part live for analyses
16385 of a single loop for multiple vector sizes.
16386 * tree-parloops.c (gather_scalar_reductions): Adjust.
16388 2018-06-25 Richard Biener <rguenther@suse.de>
16390 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
16391 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
16392 (vect_analyze_data_refs): Remove similar code from here and
16393 simplify accordingly.
16395 2018-06-25 Richard Biener <rguenther@suse.de>
16397 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
16398 for reverse storage order accesses rather than asserting
16399 they cannot happen here.
16401 2018-06-25 Tom de Vries <tdevries@suse.de>
16404 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
16405 Use data16 instead of .byte for insn prefix.
16407 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
16410 * parser.c (make_char_string_pack): Pass this literal chars
16411 through cpp_interpret_string.
16412 (cp_parser_userdef_numeric_literal): Check the result of
16413 make_char_string_pack.
16415 2018-06-24 Maya Rashish <coypu@sdf.org>
16417 * ginclude/stddef.h: Simplify conditions around avoiding
16418 re-definition of __size_t.
16420 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
16422 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
16423 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
16425 2018-06-22 Maya Rashish <coypu@sdf.org>
16427 * doc/invoke.texi (mno-fancy-math-387): Update for changes
16428 made to OpenBSD and NetBSD through the years.
16430 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16432 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
16433 behavior of vec_pack (vector double, vector double) to match
16434 behavior of vec_float2 (vector double, vector double).
16436 2018-06-22 Olivier Hainque <hainque@adacore.com>
16438 * gimplify.c (gimplify_function_tree): Prevent creation
16439 of a trampoline for the address of the current function
16440 passed to entry/exit instrumentation hooks.
16442 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
16445 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
16448 2018-06-22 Martin Liska <mliska@suse.cz>
16450 PR tree-optimization/86263
16451 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
16452 Bail out if is_enabled is false.
16453 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
16455 (jump_table_cluster::is_enabled): New function.
16457 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
16459 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
16460 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
16461 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
16462 (lto_input_ts_binfo_tree_pointers): Likewise.
16463 * tree-streamer-out.c (streamer_write_tree_bitfields,
16464 write_ts_binfo_tree_pointers): Likewise.
16465 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
16467 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
16469 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
16471 2018-06-22 Martin Liska <mliska@suse.cz>
16473 * symbol-summary.h (get): Make it pure and inline move
16474 functionality from ::get function.
16475 (get): Remove and inline into ::get and ::get_create.
16476 (get_create): Move code from ::get function.
16478 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16481 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
16482 -x assembler-with-cpp.
16484 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16486 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
16487 _FILE_OFFSET_BITS=64 for C++.
16489 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
16491 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
16492 conversion insn that shows up when pr85657-3.c is compiled using
16493 IEEE 128-bit long double.
16494 (neg<mode>2_internal): Use the correct mode to check whether the
16495 mode is IBM extended.
16496 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
16497 multiply and divide external functions from being created more
16500 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
16502 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
16504 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
16505 the edge can be forwarded.
16506 (cfg_layout_merge_blocks): Likewise.
16508 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
16510 * except.c (finish_eh_generation): Commit edge insertions only after
16511 the EH edges have been redirected from post-landing to landing pads.
16513 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
16515 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
16516 create_tmp_var_for to create the FRAME decl.
16517 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
16519 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
16521 * tree-inline.c (copy_edges_for_bb): Minor tweak.
16522 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
16523 debug statement when resetting its value.
16524 (expand_call_inline): Copy the locus of the call onto the assignment
16525 of the return value, if any. Use local variable in more cases.
16527 2018-06-21 Martin Liska <mliska@suse.cz>
16529 * ipa-pure-const.c (propagate_nothrow): Use
16530 funct_state_summaries->get.
16531 (dump_malloc_lattice): Likewise.
16532 (propagate_malloc): Likewise.
16534 2018-06-21 Richard Biener <rguenther@suse.de>
16536 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
16537 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
16538 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
16540 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
16541 BLOCK_ABSTRACT_ORIGIN unconditionally.
16543 2018-06-21 David Malcolm <dmalcolm@redhat.com>
16545 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
16547 * ipa-reference.c (ipa_reference_c_finalize): Delete
16548 ipa_ref_opt_sum_summaries and set it to NULL.
16550 2018-06-21 Tom de Vries <tdevries@suse.de>
16552 PR tree-optimization/85859
16553 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
16554 test with comment from bb_no_side_effects_p.
16556 2018-06-21 Richard Biener <rguenther@suse.de>
16558 PR tree-optimization/86232
16559 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
16560 max for constant niter.
16562 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
16564 * config/aarch64/aarch64-simd.md
16565 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
16567 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
16569 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
16570 Make opernads of the unspec commutative.
16572 2018-06-21 Richard Biener <rguenther@suse.de>
16574 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
16575 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
16576 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
16577 (vect_analyze_data_ref_dependence): Re-order checks to deal with
16579 (vect_record_base_alignments): Do not record base alignment
16580 for gathers or scatters.
16581 (vect_compute_data_ref_alignment): Drop return value that is always
16582 true. Bail out early for gathers or scatters.
16583 (vect_enhance_data_refs_alignment): Bail out early for gathers
16585 (vect_find_same_alignment_drs): Likewise.
16586 (vect_analyze_data_refs_alignment): Remove dead code.
16587 (vect_slp_analyze_and_verify_node_alignment): Likewise.
16588 (vect_analyze_data_refs): For possible gathers or scatters do
16589 not create an alternate DR, just check their possible validity
16590 and mark them. Adjust DECL_NONALIASED handling to not rely
16591 on DR_BASE_ADDRESS.
16592 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
16593 update inits of gathers or scatters.
16594 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
16595 Also copy gather/scatter flag to pattern vinfo.
16597 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
16599 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
16600 behavior of vec_packsu (vector unsigned long long, vector unsigned
16601 long long) to match behavior of vec_packs with same signature.
16603 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
16604 Thomas Schwinge <thomas@codesourcery.com>
16605 Cesar Philippidis <cesar@codesourcery.com>
16607 * gimplify.c (gimplify_scan_omp_clauses): Add support for
16608 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
16609 (gimplify_adjust_omp_clauses): Likewise.
16610 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
16611 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
16612 (gimplify_omp_target_update): Update handling of acc update and
16614 * omp-low.c (install_var_field): Remove unused parameter
16615 base_pointers_restrict.
16616 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
16617 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
16619 (omp_target_base_pointers_restrict_p): Delete.
16620 (scan_omp_target): Update call to scan_sharing_clauses.
16621 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
16623 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
16624 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
16625 (convert_local_omp_clauses): Likewise.
16626 * tree-pretty-print.c (dump_omp_clause): Likewise.
16627 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
16629 (omp_clause_code_name): Likewise.
16631 2018-06-20 Jakub Jelinek <jakub@redhat.com>
16634 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
16637 PR tree-optimization/86231
16638 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
16639 anti-range don't overwrite *vr0min before using it to compute *vr0max.
16641 2018-06-20 Tom de Vries <tdevries@suse.de>
16643 PR tree-optimization/86097
16644 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
16645 iv type if signedness of iv type is not the same as that of *nit.
16647 2018-06-20 Jakub Jelinek <jakub@redhat.com>
16649 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
16650 EDGE_EH edges, verify they are all EDGE_EH.
16652 2018-06-20 Maya Rashish <coypu@sdf.org>
16654 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
16656 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16658 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
16659 * config/aarch64/aarch64.c (xgene1_tunings): Add
16660 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
16661 (aarch64_mode_valid_for_sched_fusion_p):
16662 Allow 16-byte modes.
16663 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
16664 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
16666 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
16668 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
16669 * config/aarch64/iterators.md (VQ2): New mode iterator.
16671 2018-06-20 Martin Liska <mliska@suse.cz>
16673 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
16674 Change default ratio from 10 to 8.
16676 2018-06-20 Martin Liska <mliska@suse.cz>
16678 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
16680 (bit_test_cluster::find_bit_tests): Likewise.
16681 (switch_decision_tree::analyze_switch_statement): Find clusters.
16682 * tree-switch-conversion.h (struct jump_table_cluster): Document
16685 2018-06-20 Martin Liska <mliska@suse.cz>
16687 * tree-switch-conversion.c (switch_conversion::collect):
16688 Record m_uniq property.
16689 (switch_conversion::expand): Bail out for special conditions.
16690 (group_cluster::~group_cluster): New.
16691 (group_cluster::group_cluster): Likewise.
16692 (group_cluster::dump): Likewise.
16693 (jump_table_cluster::emit): New.
16694 (switch_decision_tree::fix_phi_operands_for_edges): New.
16695 (struct case_node): Remove struct.
16696 (jump_table_cluster::can_be_handled): New.
16697 (case_values_threshold): Moved to header.
16698 (reset_out_edges_aux): Likewise.
16699 (jump_table_cluster::is_beneficial): New.
16700 (bit_test_cluster::can_be_handled): Likewise.
16701 (add_case_node): Remove.
16702 (bit_test_cluster::is_beneficial): New.
16703 (case_bit_test::cmp): New.
16704 (bit_test_cluster::emit): New.
16705 (expand_switch_as_decision_tree_p): Remove.
16706 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
16707 (fix_phi_operands_for_edge): Likewise.
16708 (switch_decision_tree::analyze_switch_statement): New.
16709 (compute_cases_per_edge): Move ...
16710 (switch_decision_tree::compute_cases_per_edge): ... here.
16711 (try_switch_expansion): Likewise.
16712 (switch_decision_tree::try_switch_expansion): Likewise.
16713 (record_phi_operand_mapping): Likewise.
16714 (switch_decision_tree::record_phi_operand_mapping): Likewise.
16715 (emit_case_decision_tree): Likewise.
16716 (switch_decision_tree::emit): Likewise.
16717 (balance_case_nodes): Likewise.
16718 (switch_decision_tree::balance_case_nodes): Likewise.
16719 (dump_case_nodes): Likewise.
16720 (switch_decision_tree::dump_case_nodes): Likewise.
16721 (emit_jump): Likewise.
16722 (switch_decision_tree::emit_jump): Likewise.
16723 (emit_cmp_and_jump_insns): Likewise.
16724 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
16725 (emit_case_nodes): Likewise.
16726 (switch_decision_tree::emit_case_nodes): Likewise.
16727 (conditional_probability): Remove.
16728 * tree-switch-conversion.h (enum cluster_type): New.
16730 (struct cluster): Likewise.
16731 (cluster::cluster): Likewise.
16732 (struct simple_cluster): Likewise.
16733 (simple_cluster::simple_cluster): Likewise.
16734 (struct group_cluster): Likewise.
16735 (struct jump_table_cluster): Likewise.
16736 (struct bit_test_cluster): Likewise.
16737 (struct min_cluster_item): Likewise.
16738 (struct case_tree_node): Likewise.
16739 (case_tree_node::case_tree_node): Likewise.
16740 (jump_table_cluster::case_values_threshold): Likewise.
16741 (struct case_bit_test): Likewise.
16742 (struct switch_decision_tree): Likewise.
16743 (struct switch_conversion): Likewise.
16744 (switch_decision_tree::reset_out_edges_aux): Likewise.
16746 2018-06-20 Martin Liska <mliska@suse.cz>
16748 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
16749 (hoist_edge_and_branch_if_true): Likewise.
16750 (expand_switch_using_bit_tests_p): Likewise.
16751 (struct case_bit_test): Likewise.
16752 (case_bit_test_cmp): Likewise.
16753 (emit_case_bit_tests): Likewise.
16754 (switch_conversion::switch_conversion): New class.
16755 (struct switch_conv_info): Remove old struct.
16756 (collect_switch_conv_info): More to ...
16757 (switch_conversion::collect): ... this.
16758 (check_range): Likewise.
16759 (switch_conversion::check_range): Likewise.
16760 (check_all_empty_except_final): Likewise.
16761 (switch_conversion::check_all_empty_except_final): Likewise.
16762 (check_final_bb): Likewise.
16763 (switch_conversion::check_final_bb): Likewise.
16764 (create_temp_arrays): Likewise.
16765 (switch_conversion::create_temp_arrays): Likewise.
16766 (free_temp_arrays): Likewise.
16767 (gather_default_values): Likewise.
16768 (switch_conversion::gather_default_values): Likewise.
16769 (build_constructors): Likewise.
16770 (switch_conversion::build_constructors): Likewise.
16771 (constructor_contains_same_values_p): Likewise.
16772 (switch_conversion::contains_same_values_p): Likewise.
16773 (array_value_type): Likewise.
16774 (switch_conversion::array_value_type): Likewise.
16775 (build_one_array): Likewise.
16776 (switch_conversion::build_one_array): Likewise.
16777 (build_arrays): Likewise.
16778 (switch_conversion::build_arrays): Likewise.
16779 (gen_def_assigns): Likewise.
16780 (switch_conversion::gen_def_assigns): Likewise.
16781 (prune_bbs): Likewise.
16782 (switch_conversion::prune_bbs): Likewise.
16783 (fix_phi_nodes): Likewise.
16784 (switch_conversion::fix_phi_nodes): Likewise.
16785 (gen_inbound_check): Likewise.
16786 (switch_conversion::gen_inbound_check): Likewise.
16787 (process_switch): Use the newly created class.
16788 (switch_conversion::expand): New.
16789 (switch_conversion::~switch_conversion): New.
16790 * tree-switch-conversion.h: New file.
16792 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
16794 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
16795 tree-vect-patterns.c.
16796 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
16797 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
16798 (vect_recog_sad_pattern): Likewise.
16799 (vect_recog_widen_sum_pattern): Likewise.
16800 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
16801 (vect_recog_widen_shift_pattern): Remove the type_in argument.
16802 (vect_recog_rotate_pattern): Likewise.
16803 (vect_recog_mult_pattern): Likewise.
16804 (vect_recog_vector_vector_shift_pattern): Likewise.
16805 (vect_recog_divmod_pattern): Likewise.
16806 (vect_recog_mixed_size_cond_pattern): Likewise.
16807 (vect_recog_bool_pattern): Likewise.
16808 (vect_recog_mask_conversion_pattern): Likewise.
16809 (vect_try_gather_scatter_pattern): Likewise.
16810 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
16811 (vect_recog_over_widening_pattern): Likewise.
16812 (vect_recog_gather_scatter_pattern): Likewise.
16813 (vect_recog_func_ptr): Move from tree-vectorizer.h
16814 (vect_vect_recog_func_ptrs): Move further down the file.
16815 (vect_recog_func): Likewise. Remove the third argument.
16816 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
16817 (vect_pattern_recog_1): Expect the pattern function to do any
16818 necessary target tests. Also expect it to provide a vector type.
16819 Remove the type_in handling.
16821 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
16823 * tree-vect-patterns.c (vect_pattern_detected): New function.
16824 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
16825 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
16826 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
16827 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
16828 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
16829 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
16830 (vect_recog_mask_conversion_pattern)
16831 (vect_try_gather_scatter_pattern): Likewise.
16833 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
16835 * tree-vect-patterns.c (vect_get_internal_def): New function.
16836 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
16837 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
16838 (search_type_for_mask_1): Use it.
16840 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
16842 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
16843 redundant WIDEN_SUM_EXPR handling.
16844 (vect_recog_sad_pattern): Likewise.
16846 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
16848 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
16849 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
16850 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
16851 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
16852 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
16854 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
16856 * tree-vect-stmts.c (vectorizable_call): Make sure that we
16857 use the stmt_vec_info of the original bb statement for the
16858 new zero assignment, even if the call is part of a pattern.
16860 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
16862 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
16863 that the sequence is attached to the original statement rather
16864 than the pattern statement.
16865 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
16866 PATTERN_DEF_SEQ from the original statement rather than
16867 the main pattern statement.
16868 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
16869 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
16870 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
16872 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
16874 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
16875 definition statements before the early exit for statements that aren't
16877 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
16879 (vect_transform_loop): ...here. Process pattern definition
16880 statements without first checking whether the main pattern
16881 statement is live or relevant.
16883 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
16885 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
16886 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
16888 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
16890 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
16891 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
16892 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
16893 (expand_block_compare): Change select_block_compare_mode call.
16894 (expand_strncmp_align_check): Use new functions, fix comment.
16895 (emit_final_str_compare_gpr): New function.
16896 (expand_strn_compare): Refactor and clean up code.
16897 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
16899 2018-06-19 Tony Reix <tony.reix@atos.com>
16900 Damien Bergamini <damien.bergamini@atos.com>
16901 David Edelsohn <dje.gcc@gmail.com>
16903 * collect2.c (static_obj): New variable.
16904 (static_libs): New variable.
16905 (is_in_list): Uncomment declaration.
16906 (main): Track AIX libraries linked statically.
16907 (is_in_list): Uncomment definition.
16908 (scan_prog_file): Don't add AIX shared libraries initializer
16909 to constructor list if linking statically.
16911 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
16913 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
16915 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
16917 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
16919 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
16922 2018-06-19 Martin Liska <mliska@suse.cz>
16924 * config/i386/i386.c (ix86_can_inline_p): Do not use
16925 ipa_fn_summaries::get_create.
16926 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
16928 (devirtualization_time_bonus): Likewise.
16929 (ipcp_propagate_stage): Likewise.
16930 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16931 (edge_set_predicate): Likewise.
16932 (evaluate_conditions_for_known_args): Likewise.
16933 (evaluate_properties_for_edge): Likewise.
16934 (ipa_call_summary::reset): Tranform to ...
16935 (ipa_call_summary::~ipa_call_summary): ... this.
16936 (ipa_fn_summary::reset): Transform to ...
16937 (ipa_fn_summary::~ipa_fn_summary): ... this.
16938 (ipa_fn_summary_t::remove): Rename to ...
16939 (ipa_fn_summary_t::remove_callees): ... this.
16940 (ipa_fn_summary_t::duplicate): Use placement new
16941 instead of memory copy.
16942 (ipa_call_summary_t::duplicate): Likewise.
16943 (ipa_call_summary_t::remove): Remove.
16944 (dump_ipa_call_summary): Change get_create to get.
16945 (ipa_dump_fn_summary): Dump only when summary exists.
16946 (analyze_function_body): Use symbol_summary::get instead
16948 (compute_fn_summary): Likewise.
16949 (estimate_edge_devirt_benefit): Likewise.
16950 (estimate_edge_size_and_time): Likewise.
16951 (inline_update_callee_summaries): Likewise.
16952 (remap_edge_change_prob): Likewise.
16953 (remap_edge_summaries): Likewise.
16954 (ipa_merge_fn_summary_after_inlining): Likewise.
16955 (write_ipa_call_summary): Likewise.
16956 (ipa_fn_summary_write): Likewise.
16957 (ipa_free_fn_summary): Likewise.
16958 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
16959 (struct ipa_call_summary): Likewise.
16960 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
16962 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
16963 (estimate_size_after_inlining): Likewise.
16964 (estimate_growth): Likewise.
16965 (growth_likely_positive): Likewise.
16966 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16967 (inline_call): Likewise.
16968 * ipa-inline.c (caller_growth_limits): Likewise.
16969 (can_inline_edge_p): Likewise.
16970 (can_inline_edge_by_limits_p): Likewise.
16971 (compute_uninlined_call_time): Likewise.
16972 (compute_inlined_call_time): Likewise.
16973 (want_inline_small_function_p): Likewise.
16974 (edge_badness): Likewise.
16975 (update_caller_keys): Likewise.
16976 (update_callee_keys): Likewise.
16977 (inline_small_functions): Likewise.
16978 (inline_to_all_callers_1): Likewise.
16979 (dump_overall_stats): Likewise.
16980 (early_inline_small_functions): Likewise.
16981 (early_inliner): Likewise.
16982 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16983 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16984 * ipa-pure-const.c (malloc_candidate_p): Likewise.
16985 * ipa-split.c (execute_split_functions): Likewise.
16986 * symbol-summary.h: Likewise.
16987 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16989 2018-06-19 Richard Biener <rguenther@suse.de>
16991 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
16992 (vectorize_loops): ... here. Fix dbgcnt handling.
16993 (try_vectorize_loop): Wrap try_vectorize_loop_1.
16995 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
16998 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
16999 ieee128 argument takes up only one (vector) register, not two (floating
17002 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
17004 * gimplify.c (gimplify_init_constructor): Really never clear for an
17005 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
17007 2018-06-19 Richard Biener <rguenther@suse.de>
17009 PR tree-optimization/86179
17010 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
17011 after failed recognition.
17013 2018-06-18 Martin Sebor <msebor@redhat.com>
17015 PR middle-end/85602
17016 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
17017 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
17018 Handle integer subtraction.
17019 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
17020 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
17022 2018-06-18 David Malcolm <dmalcolm@redhat.com>
17024 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
17025 param from rtx to rtx_insn *.
17026 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
17028 (frv_ifcvt_modify_insn): Likwise.
17029 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
17030 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
17031 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
17032 as_a <rtx_insn *> cast to local "unprotected_region" once
17033 it's been established that it's not NULL or pc_rtx.
17034 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
17035 param "sethi" from rtx to rtx_insn *.
17036 (nds32_group_float_insns): Likewise for param "insn".
17037 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
17039 (vax_output_int_subtract): Likewise.
17040 * config/vax/vax.c (vax_output_int_add): Likewise for param
17042 (vax_output_int_subtract): Likewise.
17043 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
17044 (emit_pattern_after): Likewise for param "after".
17045 (emit_insn_after): Likewise.
17046 (emit_jump_insn_after): Likewise.
17047 (emit_call_insn_after): Likewise.
17048 (emit_debug_insn_after): Likewise.
17049 (emit_pattern_before): Likewise for param "before".
17050 (emit_insn_before): Likewise.
17051 (emit_jump_insn_before): Likewise.
17052 * final.c (get_insn_template): Likewise for param "insn", removing
17054 * output.h (get_insn_template): Likewise for 2nd param.
17055 * rtl.h (emit_insn_before): Likewise.
17056 (emit_jump_insn_before): Likewise.
17057 (emit_debug_insn_before_noloc): Likewise.
17058 (emit_insn_after): Likewise.
17059 (emit_jump_insn_after): Likewise.
17060 (emit_call_insn_after): Likewise.
17061 (emit_debug_insn_after): Likewise.
17062 (set_insn_deleted): Likewise for param.
17064 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
17067 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
17068 floating point modes, so that IFmode is numerically greater than
17069 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
17070 to declare the ordering. This prevents IFmode from being
17071 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
17072 machine. Include rs6000-modes.h to share the fractional values
17073 between genmodes* and the rest of the compiler.
17074 (IFmode): Likewise.
17075 (KFmode): Likewise.
17076 (TFmode): Likewise.
17077 * config/rs6000/rs6000-modes.h: New file.
17078 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
17079 meaning of rs6000_long_double_size so that 126..128 selects an
17080 appropriate 128-bit floating point type.
17081 (rs6000_option_override_internal): Likewise.
17082 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
17083 (TARGET_LONG_DOUBLE_128): Change the meaning of
17084 rs6000_long_double_size so that 126..128 selects an appropriate
17085 128-bit floating point type.
17086 (LONG_DOUBLE_TYPE_SIZE): Update comment.
17087 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
17088 source and destination to match the standard usage.
17089 (truncifkf2): Likewise.
17090 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
17091 ISA 2.07 to use an explicit clobber, instead of passing in a
17093 (copysign<mode>3_soft): Likewise.
17095 2018-06-18 David Malcolm <dmalcolm@redhat.com>
17097 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
17098 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
17099 (vect_slp_analyze_instance_dependence): Likewise.
17100 (vect_enhance_data_refs_alignment): Likewise.
17101 (vect_analyze_data_refs_alignment): Likewise.
17102 (vect_slp_analyze_and_verify_instance_alignment
17103 (vect_analyze_data_ref_accesses): Likewise.
17104 (vect_prune_runtime_alias_test_list): Likewise.
17105 (vect_analyze_data_refs): Likewise.
17106 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
17107 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
17108 (vect_analyze_scalar_cycles_1): Likewise.
17109 (vect_get_loop_niters): Likewise.
17110 (vect_analyze_loop_form_1): Likewise.
17111 (vect_update_vf_for_slp): Likewise.
17112 (vect_analyze_loop_operations): Likewise.
17113 (vect_analyze_loop): Likewise.
17114 (vectorizable_induction): Likewise.
17115 (vect_transform_loop): Likewise.
17116 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
17117 * tree-vect-slp.c (vect_analyze_slp): Likewise.
17118 (vect_make_slp_decision): Likewise.
17119 (vect_detect_hybrid_slp): Likewise.
17120 (vect_slp_analyze_operations): Likewise.
17121 (vect_slp_bb): Likewise.
17122 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
17123 (vectorizable_bswap): Likewise.
17124 (vectorizable_call): Likewise.
17125 (vectorizable_simd_clone_call): Likewise.
17126 (vectorizable_conversion): Likewise.
17127 (vectorizable_assignment): Likewise.
17128 (vectorizable_shift): Likewise.
17129 (vectorizable_operation): Likewise.
17130 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
17132 2018-06-18 Martin Sebor <msebor@redhat.com>
17134 PR tree-optimization/81384
17135 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
17136 * builtins.c (expand_builtin_strnlen): New function.
17137 (expand_builtin): Call it.
17138 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
17139 * builtins.def (BUILT_IN_STRNLEN): New.
17140 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
17141 Warn for bounds in excess of maximum object size.
17142 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
17143 single-value ranges. Handle strnlen.
17144 (handle_builtin_strlen): Handle strnlen.
17145 (strlen_check_and_optimize_stmt): Same.
17146 * doc/extend.texi (Other Builtins): Document strnlen.
17148 2018-06-18 Maya Rashish <coypu@sdf.org>
17150 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
17151 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
17152 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
17154 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
17156 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
17158 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17160 * tree.c (escaped_string::escape): Replace cast to char * by
17161 const_cast<char *> (unescaped).
17163 2018-06-18 Nick Clifton <nickc@redhat.com>
17166 * tree.c (escaped_string): New class. Converts an unescaped
17167 string into its escaped equivalent.
17168 (warn_deprecated_use): Use the new class to convert the
17169 deprecation message, if present.
17170 (test_escaped_strings): New self test.
17171 (test_c_tests): Add test_escaped_strings.
17172 * doc/extend.texi (deprecated): Add a note that the
17173 deprecation message is affected by the -fmessage-length
17174 option, and that control characters will be escaped.
17175 (#pragma GCC error): Document this pragma.
17176 (#pragma GCC warning): Likewise.
17177 * doc/invoke.texi (-fmessage-length): Document this option's
17178 effect on the #warning and #error preprocessor directives and
17179 the deprecated attribute.
17181 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
17183 * tree.c (decl_value_expr_lookup): Revert latest change.
17184 (decl_value_expr_insert): Likewise.
17186 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
17188 * gimplify.c (nonlocal_vlas): Delete.
17189 (nonlocal_vla_vars): Likewise.
17190 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
17192 (gimplify_body): Do not create and destroy nonlocal_vlas.
17193 * tree-nested.c: Include diagnostic.h.
17194 (use_pointer_in_frame): Tweak.
17195 (lookup_field_for_decl): Add assertion and declare the transformation.
17196 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
17197 internal error when the reference is in a wrong context. Do not
17198 create a debug decl by default.
17199 (note_nonlocal_block_vlas): Delete.
17200 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
17201 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
17202 create a debug decl by default.
17203 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
17204 call is in a wrong context.
17205 (fixup_vla_decls): New function.
17206 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
17207 debug variables were created.
17208 * tree.c (decl_value_expr_lookup): Add checking assertion.
17209 (decl_value_expr_insert): Likewise.
17211 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
17213 PR middle-end/82479
17214 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
17215 * tree-scalar-evolution.c (interpret_expr): Likewise.
17216 (expression_expensive_p): Likewise.
17217 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
17218 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
17219 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
17220 (ssa_defined_by_minus_one_stmt_p): New.
17222 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
17224 PR middle-end/64946
17225 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
17226 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
17227 * dojump.c (do_jump): Likewise.
17228 * expr.c (expand_expr_real_2): Check operand type's sign.
17229 * fold-const.c (const_unop): Handle ABSU_EXPR.
17230 (fold_abs_const): Likewise.
17231 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
17232 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
17233 (strip_sign_op_1): Likesise.
17234 * match.pd: Add new pattern to generate ABSU_EXPR.
17235 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
17236 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
17237 * tree-eh.c (operation_could_trap_helper_p): Likewise.
17238 * tree-inline.c (estimate_operator_cost): Likewise.
17239 * tree-pretty-print.c (dump_generic_node): Likewise.
17240 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
17241 * tree.def (ABSU_EXPR): New.
17243 2018-06-16 Jakub Jelinek <jakub@redhat.com>
17245 PR middle-end/86095
17246 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
17247 documented as preserved for backward compatibility only.
17248 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
17250 PR rtl-optimization/86108
17251 * bb-reorder.c (create_forwarder_block): Renamed to ...
17252 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
17253 jump from new landing pad to the second part.
17254 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
17257 2018-06-15 Jakub Jelinek <jakub@redhat.com>
17259 PR middle-end/85878
17260 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
17261 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
17262 Only call store_expr for halves if the mode is the same.
17264 PR middle-end/86123
17265 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
17266 Fix up comment formatting.
17268 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
17270 * typed-splay-tree.h (typed_splay_tree::remove): New function.
17271 (typed_splay_tree::closure,
17272 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
17273 (typed_splay_tree::typed_splay_tree,
17274 typed_splay_tree::operator =): Declared private.
17275 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
17276 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
17277 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
17278 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
17279 typed_splay_tree::splay_tree_splay,
17280 typed_splay_tree::splay_tree_foreach_helper,
17281 typed_splay_tree::splay_tree_insert,
17282 typed_splay_tree::splay_tree_remove,
17283 typed_splay_tree::splay_tree_lookup,
17284 typed_splay_tree::splay_tree_predecessor,
17285 typed_splay_tree::splay_tree_successor,
17286 typed_splay_tree::splay_tree_min,
17287 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
17288 (typed_splay_tree::root, typed_splay_tree::comp,
17289 typed_splay_tree::delete_key,
17290 typed_splay_tree::delete_value): New data members.
17291 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
17292 typed_splay_tree::remove.
17294 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
17296 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
17297 -mginv and -mno-ginv to the assembler.
17298 * config/mips/mips.opt (-mcrc): New option.
17299 (-mginv): Likewise.
17300 * doc/invoke.text (-mcrc): Document.
17301 (-mginv): Likewise.
17303 2018-06-15 Nick Clifton <nickc@redhat.com>
17306 * tree.c (escaped_string): New class. Converts an unescaped
17307 string into its escaped equivalent.
17308 (warn_deprecated_use): Use the new class to convert the
17309 deprecation message, if present.
17310 (test_escaped_strings): New self test.
17311 (test_c_tests): Add test_escaped_strings.
17312 * doc/extend.texi (deprecated): Add a note that the
17313 deprecation message is affected by the -fmessage-length
17314 option, and that control characters will be escaped.
17315 (#pragma GCC error): Document this pragma.
17316 (#pragma GCC warning): Likewise.
17317 * doc/invoke.texi (-fmessage-length): Document this option's
17318 effect on the #warning and #error preprocessor directives and
17319 the deprecated attribute.
17321 2018-06-15 Richard Biener <rguenther@suse.de>
17323 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
17324 here, also noting vector size used.
17325 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
17326 size used in MSG_OPTIMIZED_LOCATIONS dump.
17327 (pass_slp_vectorize::execute): Adjust.
17329 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
17332 * config/arc/arc.c (arc_return_address_register): Fix
17335 2018-06-15 Richard Biener <rguenther@suse.de>
17337 PR middle-end/86159
17338 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
17339 leave useless conversion stripping to force_gimple_operand_gsi.
17340 (gimplify_build2): Likewise.
17341 (gimplify_build1): Likewise.
17343 2018-06-15 Richard Biener <rguenther@suse.de>
17345 PR middle-end/86076
17346 * tree-cfg.c (move_stmt_op): unshare invariant addresses
17347 before adjusting their block.
17349 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
17351 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
17352 multilibs for *-*-rtems*.
17353 * config/riscv/t-rtems: New file.
17355 2018-06-14 Jakub Jelinek <jakub@redhat.com>
17357 PR middle-end/86122
17358 * match.pd ((A +- CST1) +- CST2): Punt if last resort
17359 unsigned_type_for returns NULL.
17362 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
17363 subregs of multi-word pseudos unless the float mode has word size.
17365 2018-06-14 Richard Biener <rguenther@suse.de>
17367 PR middle-end/86139
17368 * tree-vect-generic.c (build_word_mode_vector_type): Remove
17369 duplicate and harmful type_hash_canon.
17370 * tree.c (type_hash_canon): Assert we didn't find ourselves.
17372 2018-06-14 Richard Biener <rguenther@suse.de>
17375 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
17378 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
17380 * config/rtems.h (STDINT_LONG32): Define.
17382 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
17383 Prachi Godbole <prachi.godbole@imgtec.com>
17385 * config/mips/mips-cpus.def: Define P6600.
17386 * config/mips/mips-tables.opt: Regenerate.
17387 * config/mips/mips.c (mips_ucbranch_type): New enum.
17388 (mips_rtx_cost_data): Add support for P6600.
17389 (mips_issue_rate): Likewise.
17390 (mips_multipass_dfa_lookahead): Likewise.
17391 (mips_avoid_hazard): Likewise.
17392 (mips_reorg_process_insns): Likewise.
17393 (mips_classify_branch_p6600): New function.
17394 * config/mips/mips.h (TUNE_P6600): New define.
17395 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
17396 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
17397 * config/mips/mips.md: Include p6600.md.
17398 (processor): Add p6600.
17399 * config/mips/p6600.md: New file.
17400 * doc/invoke.texi: Add p6600 to supported architectures.
17402 2018-06-13 Martin Sebor <msebor@redhat.com>
17404 PR tree-optimization/86114
17405 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
17407 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
17409 2018-06-13 Richard Biener <rguenther@suse.de>
17411 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
17412 Properly set vector type of the intermediate stmt.
17413 * tree-vect-stmts.c (vectorizable_operation): The destination
17414 var always has vectype_out type.
17416 2018-06-13 Jeff Law <law@redhat.com>
17418 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
17419 integer 0 for argument to print_rtl_with_bb.
17420 (rl78_reorg): Likewise.
17422 2018-06-13 David Malcolm <dmalcolm@redhat.com>
17424 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
17425 from rtx to rtx_insn *.
17426 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
17428 (add_sched_insns_for_speculation): Likewise for local "target",
17429 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
17430 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
17431 from rtx_insn ** to rtx_code_label **.
17432 (reorg_emit_nops): Likewise.
17433 (c6x_reorg): Likewise for local "call_labels".
17434 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
17436 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
17437 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
17438 the loops over LABEL_REFS.
17439 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
17441 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
17442 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
17443 (split_branches): Strengthen local "olabel" from rtx to
17444 rtx_insn *, adding a safe_as_a cast.
17445 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
17447 (add_insn_after): Likewise for first two params.
17448 (add_insn_before): Likewise.
17449 (remove_insn): Likewise for param.
17450 (emit_pattern_before_noloc): Likewise for second and third params.
17451 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
17452 (emit_call_insn_before_noloc): Likewise.
17453 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
17455 (emit_barrier_before): Likewise.
17456 (emit_label_before): Strengthen "label" param from "rtx" to
17457 "rtx_code_label *". Strengthen "before" param from "rtx" to
17459 (emit_insn_after_1): Strengthen "after" param from "rtx" to
17461 (emit_pattern_after_noloc): Likewise.
17462 (emit_insn_after_noloc): Likewise.
17463 (emit_jump_insn_after_noloc): Likewise.
17464 (emit_call_insn_after_noloc): Likewise.
17465 (emit_debug_insn_after_noloc): Likewise.
17466 (emit_barrier_after): Likewise.
17467 (emit_label_after): Likewise for both params.
17468 (emit_pattern_after_setloc): Likewise for "after" param. Convert
17469 "loc" param from "int" to "location_t".
17470 (emit_insn_after_setloc): Likewise.
17471 (emit_jump_insn_after_setloc): Likewise.
17472 (emit_call_insn_after_setloc): Likewise.
17473 (emit_debug_insn_after_setloc): Likewise.
17474 (emit_pattern_before_setloc): Likewise for "before" param. Convert
17475 "loc" param from "int" to "location_t".
17476 (emit_pattern_before): Convert NULL_RTX to NULL.
17477 (emit_insn_before_setloc): Convert "loc" param from "int" to
17479 (emit_jump_insn_before_setloc): Likewise.
17480 (emit_call_insn_before_setloc): Likewise.
17481 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
17482 rtx_insn *. Convert "loc" param from "int" to "location_t".
17483 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
17484 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
17485 Convert 3rd param from "int" to "location_t".
17486 (emit_barrier_before, emit_barrier_after, next_real_insn):
17487 Strengthen param from rtx to rtx_insn *.
17488 (emit_label_before): Strengthen 1st param from "rtx" to
17489 "rtx_code_label *". Strengthen 2nd param from "rtx" to
17491 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
17492 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
17493 Strengthen 2nd param from "rtx" to "rtx_insn *".
17494 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
17495 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
17496 Likewise. Convert 3rd param from "int" to "location_t".
17497 (emit_label_after): Strengthen 1st param from "rtx" to
17498 "rtx_code_label *".
17499 (next_real_insn, remove_insn): Strengthen param from "rtx" to
17501 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
17502 from "rtx" to "rtx_insn *".
17504 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
17506 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
17507 bodies streamed in with -Q.
17508 * dumpfile.c (dump_files): Add lto-stream-out dump file.
17509 * dumpfile.h (tree_dump_index): Add lto_stream_out.
17510 * gimple-streamer-out.c: Include gimple-pretty-print.h
17511 (output_bb): Dump stmts streamed.
17512 * lto-section-out.c: Include print-tree.h
17513 (lto_begin_section): Dump sections created.
17514 (lto_output_decl_index): Dump decl encoded.
17515 * lto-streamer-out.c: Include print-tree.h
17516 (create_output_block): Dump output block created.
17517 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
17518 (output_function): Dump function output.
17519 (output_constructor): Dump constructor streamed.
17520 (write_global_stream): Output indexes encoded.
17521 (produce_asm_for_decls): Dump streams encoded.
17522 * lto-streamer.c (streamer_dump_file): New global var.
17523 * lto-streamer.h (streamer_dump_file): Declare.
17524 * passes.c (ipa_write_summaries): Initialize streamer dump.
17525 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
17528 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
17531 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
17532 offsets for register save directives. Emit a second batch of save
17533 directives, if need be, when the function accesses prior frames.
17535 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
17537 * config/arc/fpu.md (fmasf4): Force operand to register.
17538 (fnmasf4): Likewise.
17540 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
17542 * config/arc/arc-protos.h (arc_pad_return): Remove.
17543 * config/arc/arc.c (machine_function): Remove force_short_suffix
17545 (arc_print_operand): Adjust printing of '&'.
17546 (arc_verify_short): Remove conditional printing of short suffix.
17547 (arc_final_prescan_insn): Remove reference to size_reason.
17548 (pad_return): New function.
17549 (arc_reorg): Call pad_return.
17550 (arc_pad_return): Remove.
17551 (arc_init_machine_status): Remove reference to force_short_suffix.
17552 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
17553 (attr length): When attribute iscompact is true force to 2
17554 regardless; in the case of maybe check if we want to force the
17555 instruction to have 4 bytes length.
17556 (nopv): Change it to generate 4 byte long nop as well.
17557 (blockage): New pattern.
17558 (simple_return): Remove call to arc_pad_return.
17559 (p_return_i): Likewise.
17561 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
17563 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
17565 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
17567 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
17570 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
17572 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
17573 for ARC700 and ARCv2.
17575 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
17578 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
17579 operands[2] instead of operands[1].
17582 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
17584 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
17585 case, check whether the outer register overlaps an unallocatable
17586 register, not just whether it fits the required class.
17588 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
17590 * poly-int.h (can_div_trunc_p): Add new overload in which all values
17592 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
17593 (memrefs_conflict_p): Likewise.
17594 (init_alias_analysis): Likewise.
17595 * cfgexpand.c (expand_debug_expr): Likewise.
17596 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
17597 * cse.c (fold_rtx): Likewise.
17598 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
17599 * expr.c (emit_block_move_hints): Likewise.
17600 (clear_storage_hints, push_block, emit_push_insn): Likewise.
17601 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
17602 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
17603 (emit_group_store): Likewise.
17604 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
17605 to read the PRE/POST_MODIFY increment.
17606 * calls.c (store_one_arg): Use strip_offset.
17607 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
17609 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
17611 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
17612 (simplify_binary_operation_1): Extend CONST_INT handling to
17614 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
17615 than a HOST_WIDE_INT.
17616 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
17618 (adjust_mems, add_stores): Update accodingly.
17619 (vt_canonicalize_addr): Track polynomial offsets.
17620 (emit_note_insn_var_location): Likewise.
17621 (vt_add_function_parameter): Likewise.
17622 (vt_initialize): Likewise.
17624 2018-06-12 Jeff Law <law@redhat.com>
17626 * config.gcc (alpha*-*-freebsd*): Remove.
17627 * config/alpha/freebsd.h: Remove.
17629 2018-06-12 David Malcolm <dmalcolm@redhat.com>
17632 * spellcheck-tree.c (levenshtein_distance): Rename to...
17633 (get_edit_distance): ...this, and update for underlying renaming.
17634 * spellcheck-tree.h (levenshtein_distance): Rename to...
17635 (get_edit_distance): ...this.
17636 * spellcheck.c (levenshtein_distance): Rename to...
17637 (get_edit_distance): ...this. Convert from Levenshtein distance
17638 to Damerau-Levenshtein distance by supporting transpositions of
17639 adjacent characters. Rename "v1" to "v_next" and "v0" to
17641 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
17642 (selftest::test_edit_distance_unit_test_oneway): ...this, and
17643 update for underlying renaming.
17644 (selftest::levenshtein_distance_unit_test): Rename to...
17645 (selftest::test_get_edit_distance_unit): ...this, and update for
17646 underlying renaming.
17647 (selftest::test_find_closest_string): Add example from PR 69968
17648 where transposition helps
17649 (selftest::test_metric_conditions): Update for renaming.
17650 (selftest::test_metric_conditions): Likewise.
17651 (selftest::spellcheck_c_tests): Likewise.
17652 * spellcheck.h (levenshtein_distance): Rename both overloads to...
17653 (get_edit_distance): ...this.
17654 (best_match::consider): Update for renaming.
17656 2018-06-12 Martin Sebor <msebor@redhat.com>
17658 PR tree-optimization/85259
17659 * builtins.c (compute_objsize): Handle constant offsets.
17660 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
17661 true iff a warning has been issued.
17662 * gimple.h (gimple_nonartificial_location): New function.
17663 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
17664 gimple_nonartificial_location and handle -Wno-system-headers.
17665 (handle_builtin_stxncpy): Same.
17667 2018-06-12 Martin Sebor <msebor@redhat.com>
17670 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
17672 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
17674 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
17675 BUILTIN_VEC_XST entries for pointer to double and long long.
17677 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
17680 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
17682 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
17685 2018-06-12 Martin Liska <mliska@suse.cz>
17687 * doc/options.texi: Document IntegerRange.
17689 2018-06-12 Martin Liska <mliska@suse.cz>
17691 * config/i386/i386.opt: Make MPX-related options as Deprecated.
17692 * opt-functions.awk: Handle Deprecated flag.
17693 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
17695 (read_cmdline_option): Report warning for a deprecated option.
17696 * opts.h (struct cl_option): Add new field cl_deprecated.
17697 (CL_ERR_DEPRECATED): New.
17699 2018-06-12 Martin Liska <mliska@suse.cz>
17701 * doc/options.texi: Document Deprecated option flag.
17703 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
17705 * config/arc/arc-arch.h (arc_extras): New enum.
17706 (arc_cpu_t):Add field extra.
17707 (arc_cpu_types): Consider the extras.
17708 * config/arc/arc-cpus.def: Add extras info.
17709 * config/arc/arc-opts.h (processor_type): Consider extra field.
17710 * config/arc/arc.c (arc_override_options): Handle extra field.
17712 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
17714 * config/arc/arc-arch.h: Update ARC_OPTX macro.
17715 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
17717 * config/arc/arc.c (arc_init): Update pic warning.
17718 (irq_range): Update irq range parsing warnings.
17719 (arc_override_options): Update various warning messages.
17720 (arc_handle_aux_attribute): Likewise.
17722 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
17724 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
17726 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
17728 * doc/sourcebuild.texi: Document usage of line number 0 in verify
17729 compiler messages directives.
17731 2018-06-12 Matthew Fortune <mfortune@gmail.com>
17733 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
17734 * config/mips/mips-tables.opt: Regenerate.
17735 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
17737 * doc/invoke.texi: Document -march=i6500.
17739 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
17741 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
17742 (i6400_gpmul): Add cpu_unit.
17743 (i6400_gpdiv): Likewise.
17744 (i6400_msa_add_d): Update reservations.
17745 (i6400_msa_int_add) Likewise.
17746 (i6400_msa_short_logic3) Likewise.
17747 (i6400_msa_short_logic2) Likewise.
17748 (i6400_msa_short_logic) Likewise.
17749 (i6400_msa_move) Likewise.
17750 (i6400_msa_cmp) Likewise.
17751 (i6400_msa_short_float2) Likewise.
17752 (i6400_msa_div_d) Likewise.
17753 (i6400_msa_long_logic1) Likewise.
17754 (i6400_msa_long_logic2) Likewise.
17755 (i6400_msa_mult) Likewise.
17756 (i6400_msa_long_float2) Likewise.
17757 (i6400_msa_long_float4) Likewise.
17758 (i6400_msa_long_float5) Likewise.
17759 (i6400_msa_long_float8) Likewise.
17760 (i6400_fpu_fadd): Include frint type.
17761 (i6400_fpu_store): New define_insn_reservation.
17762 (i6400_fpu_load): Likewise.
17763 (i6400_fpu_move): Likewise.
17764 (i6400_fpu_fcmp): Likewise.
17765 (i6400_fpu_fmadd): Likewise.
17766 (i6400_int_mult): Include imul3nc type and update reservation.
17767 (i6400_int_div): Include idiv3 type and update reservation.
17768 (i6400_int_load): Update to check type not move_type.
17769 (i6400_int_store): Likewise.
17770 (i6400_int_prefetch): Set zero latency.
17772 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
17774 * gcc.c: Document new %@{...} sequence.
17775 (LINK_COMMAND_SPEC): Use it for the -L switches.
17776 (cpp_unique_options): Use it for the -I switches.
17777 (at_file_argbuf): New global variable.
17778 (in_at_file): Likewise.
17779 (alloc_args): Create at_file_argbuf.
17780 (clear_args): Truncate at_file_argbuf.
17781 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
17782 (open_at_file): New function.
17783 (close_at_file): Likewise.
17784 (create_at_file): Delete.
17785 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
17788 (validate_switches_from_spec): Deal with %@{...} sequence.
17789 (validate_switches): Likewise.
17790 (driver::finalize): Call clear_args.
17792 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
17794 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
17796 2018-06-11 Martin Sebor <msebor@redhat.com>
17798 * doc/invoke.texi (-Wall): List -Wc++17-compat.
17799 (Wno-class-memaccess): Add @opindex.
17800 (Wno-templates, Wno-multiple-inheritance): Same.
17801 (Wno-virtual-inheritance, Wno-namespaces): Same.
17802 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
17803 (Wno-format-overflow, Wno-format-truncation): Same.
17804 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
17805 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
17806 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
17807 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
17808 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
17809 (Wno-misspelled-isr): Same.
17811 2018-06-11 Martin Sebor <msebor@redhat.com>
17813 * PR tree-optimization/86083
17814 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
17816 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
17818 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
17820 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
17823 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
17824 on the correct operand.
17825 (*movdi_internal64): Ditto.
17827 2018-06-11 Martin Liska <mliska@suse.cz>
17829 PR tree-optimization/86089
17830 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
17832 2018-06-11 Julia Koval <julia.koval@intel.com>
17834 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
17835 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
17836 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
17838 2018-06-11 Olivier Hainque <hainque@adacore.com>
17840 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
17841 for Ada with strict dwarf2.
17843 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
17846 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
17849 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
17851 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
17853 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
17855 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
17858 2018-06-08 Martin Liska <mliska@suse.cz>
17860 * tree-cfg.h (debug_function): Fix argument type to match
17863 2018-06-08 Martin Liska <mliska@suse.cz>
17865 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
17866 Remove usage of MPX-related (and removed) fields.
17867 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
17869 2018-06-08 David Malcolm <dmalcolm@redhat.com>
17871 * cfg.c (debug): Use TDF_NONE rather than 0.
17872 * cfghooks.c (debug): Likewise.
17873 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
17874 (struct dump_option_value_info): Convert to...
17875 (struct kv_pair): ...this template type.
17876 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
17878 (optinfo_verbosity_options): Likewise.
17879 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
17881 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
17882 than int for "optgroup_flags" param.
17883 (dump_generic_expr_loc): Use dump_flags_t rather than int for
17885 (dump_dec): Likewise.
17886 (dump_finish): Use TDF_NONE rather than 0.
17887 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
17888 rather than int for "optgroup_flags" param. Use TDF_NONE rather
17889 than 0. Update for change to option_ptr.
17890 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
17891 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
17892 0. Update for changes to optinfo_verbosity_options and
17894 (opt_info_switch_p): Convert optgroup_flags from int to
17896 (dump_basic_block): Use dump_flags_t rather than int
17897 for "dump_kind" param.
17898 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
17899 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
17900 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
17901 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
17902 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
17903 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
17904 TDF_NONE): Convert from macros to...
17905 (enum dump_flag): ...this new enum.
17906 (dump_flags_t): Update to use enum.
17907 (operator|, operator&, operator~, operator|=, operator&=):
17908 Implement for dump_flags_t.
17909 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
17910 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
17911 Convert from macros to...
17912 (enum optgroup_flag): ...this new enum.
17913 (optgroup_flags_t): New typedef.
17914 (operator|, operator|=): Implement for optgroup_flags_t.
17915 (struct dump_file_info): Convert field "alt_flags" to
17916 dump_flags_t. Convert field "optgroup_flags" to
17918 (dump_basic_block): Use dump_flags_t rather than int for param.
17919 (dump_generic_expr_loc): Likewise.
17920 (dump_dec): Likewise.
17921 (dump_register): Convert param "optgroup_flags" to
17923 (opt_info_enable_passes): Likewise.
17924 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
17926 * gimple-pretty-print.c (debug): Likewise.
17927 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
17928 (merged_store_group::apply_stores): Likewise.
17929 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
17930 * gimple.c (verify_gimple_pp): Likewise.
17931 * graphite-poly.c (print_pbb_body): Likewise.
17932 * passes.c (pass_manager::register_one_dump_file): Convert
17933 local "optgroup_flags" to optgroup_flags_t.
17934 * print-tree.c (print_node): Use TDF_NONE rather than 0.
17936 (debug_body): Likewise.
17937 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
17938 to optgroup_flags_t.
17939 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
17941 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
17942 (convert_mult_to_fma): Likewise.
17943 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
17944 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
17945 * tree-vect-data-refs.c (dump_lower_bound): Convert param
17946 "dump_kind" to dump_flags_t.
17948 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
17950 * config/rs6000/rs6000.c (min, max): Delete.
17952 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
17954 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
17955 -mabi=spe and -mabi=no-spe.
17957 2018-06-08 Martin Liska <mliska@suse.cz>
17959 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
17960 where we expect an existing summary.
17962 2018-06-08 Martin Liska <mliska@suse.cz>
17964 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
17965 * ipa-inline.h (estimate_edge_growth): Likewise.
17967 2018-06-08 Martin Liska <mliska@suse.cz>
17969 * cgraph.c (function_version_hasher::hash): Use
17970 cgraph_node::get_uid ().
17971 (function_version_hasher::equal):
17972 * cgraph.h (cgraph_node::get_uid): New method.
17973 * ipa-inline.c (update_caller_keys): Use
17974 cgraph_node::get_uid ().
17975 (update_callee_keys): Likewise.
17976 * ipa-utils.c (searchc): Likewise.
17977 (ipa_reduced_postorder): Likewise.
17978 * lto-cgraph.c (input_node): Likewise.
17979 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
17980 * symbol-summary.h (symtab_insertion): Likewise.
17981 (symtab_removal): Likewise.
17982 (symtab_duplication): Likewise.
17983 * tree-pretty-print.c (dump_function_header): Likewise.
17984 * tree-sra.c (convert_callers_for_node): Likewise.
17986 2018-06-08 Martin Liska <mliska@suse.cz>
17988 * cgraph.c (symbol_table::create_edge): Always assign a new
17990 (symbol_table::free_edge): Do not recycle numbers.
17991 * cgraph.h (cgraph_edge::get): New method.
17992 * symbol-summary.h (symtab_removal): Use it.
17993 (symtab_duplication): Likewise.
17994 (call_summary::hashable_uid): Remove.
17996 2018-06-08 Martin Liska <mliska@suse.cz>
17998 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
17999 (initialize_growth_caches): Remove.
18000 (free_growth_caches): Likewise.
18001 (do_estimate_edge_time): Use edge_growth_cache.
18002 (do_estimate_edge_size): Likewise.
18003 (do_estimate_edge_hints): Likewise.
18004 * ipa-inline.c (reset_edge_caches): Likewise.
18005 (recursive_inlining): Likewise.
18006 (inline_small_functions): Likewise.
18007 * ipa-inline.h (initialize_growth_caches): Remove.
18008 (estimate_edge_size): Likewise.
18009 (estimate_edge_time): Likewise.
18010 (estimate_edge_hints): Likewise.
18011 (reset_edge_growth_cache): Likewise.
18012 * symbol-summary.h (call_summary::remove): New method.
18014 2018-06-08 Martin Liska <mliska@suse.cz>
18016 * ipa-cp.c (class edge_clone_summary): New summary.
18017 (grow_edge_clone_vectors): Remove.
18018 (ipcp_edge_duplication_hook): Remove.
18019 (class edge_clone_summary_t): New call_summary class.
18020 (ipcp_edge_removal_hook): Remove.
18021 (edge_clone_summary_t::duplicate): New function.
18022 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
18023 (create_specialized_node): Likewise.
18024 (ipcp_driver): Initialize edge_clone_summaries and do not
18027 2018-06-08 Martin Liska <mliska@suse.cz>
18029 * symbol-summary.h (get): New function.
18030 (call_summary::m_initialize_when_cloning): New class member.
18032 2018-06-08 Martin Liska <mliska@suse.cz>
18034 * cgraph.c (cgraph_node::remove): Do not recycle uid.
18035 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
18036 (symbol_table::allocate_cgraph_symbol): Do not set uid.
18037 * passes.c (uid_hash_t): Record removed_nodes by their uids.
18038 (remove_cgraph_node_from_order): Use the removed_nodes set.
18039 (do_per_function_toporder): Likwise.
18040 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
18041 instead of summary_uid.
18042 (symtab_removal): Likewise.
18043 (symtab_duplication): Likewise.
18045 2018-06-08 Martin Liska <mliska@suse.cz>
18047 * ipa-cp.c (ipcp_store_bits_results): Use
18048 ipcp_transformation_sum.
18049 (ipcp_store_vr_results): Likewise.
18050 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
18052 (ipcp_transformation_initialize): ... this.
18053 (ipa_set_node_agg_value_chain):
18054 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
18055 (write_ipcp_transformation_info): Likewise.
18056 (read_ipcp_transformation_info): Likewise.
18057 (ipcp_update_bits): Likewise.
18058 (ipcp_update_vr): Likewise.
18059 (ipcp_transform_function): Likewise.
18060 * ipa-prop.h: Rename ipcp_transformation_summary to
18061 ipcp_transformation.
18062 (class ipcp_transformation_t): New function summary.
18063 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
18064 (ipa_get_agg_replacements_for_node): Likewise.
18066 2018-06-08 Martin Liska <mliska@suse.cz>
18068 * ipa-pure-const.c (struct funct_state_d): Do it class instead
18070 (class funct_state_summary_t): New function_summary class.
18071 (has_function_state): Remove.
18072 (get_function_state): Likewise.
18073 (set_function_state): Likewise.
18074 (add_new_function): Likewise.
18075 (funct_state_summary_t::insert): New function.
18076 (duplicate_node_data): Remove.
18077 (remove_node_data): Remove.
18078 (funct_state_summary_t::duplicate): New function.
18079 (register_hooks): Create new funct_state_summaries.
18080 (pure_const_generate_summary): Use it.
18081 (pure_const_write_summary): Likewise.
18082 (pure_const_read_summary): Likewise.
18083 (propagate_pure_const): Likewise.
18084 (propagate_nothrow): Likewise.
18085 (dump_malloc_lattice): Likewise.
18086 (propagate_malloc): Likewise.
18087 (execute): Do not register hooks, just remove summary
18089 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
18092 2018-06-08 Martin Liska <mliska@suse.cz>
18094 * ipa-reference.c (remove_node_data): Remove.
18095 (duplicate_node_data): Likewise.
18096 (class ipa_ref_var_info_summary_t): New class.
18097 (class ipa_ref_opt_summary_t): Likewise.
18098 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
18099 (get_reference_optimization_summary): Use
18100 ipa_ref_opt_sum_summaries.
18101 (set_reference_vars_info): Remove.
18102 (set_reference_optimization_summary): Likewise.
18103 (ipa_init): Create summaries.
18104 (init_function_info): Use function summary.
18105 (ipa_ref_opt_summary_t::duplicate): New function.
18106 (ipa_ref_opt_summary_t::remove): New function.
18107 (get_read_write_all_from_node): Fix GNU coding style.
18108 (propagate): Use function summary.
18109 (write_node_summary_p): Fix GNU coding style.
18110 (stream_out_bitmap): Likewise.
18111 (ipa_reference_read_optimization_summary): Use function summary.
18112 (ipa_reference_c_finalize): Do not release hooks.
18114 2018-06-08 Martin Liska <mliska@suse.cz>
18116 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
18117 (analyze_function_body): Extract multiple calls of get_create.
18118 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
18119 * ipa-inline.c (recursive_inlining): Use ::get method.
18120 * ipa-inline.h (estimate_edge_growth): Likewise.
18122 2018-06-08 Martin Liska <mliska@suse.cz>
18124 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
18126 (hsa_function_summary::hsa_function_summary): Use the new enum
18128 (hsa_gpu_implementation_p): Use hsa_summaries::get.
18129 * hsa-gen.c (hsa_get_host_function): Likewise.
18130 (get_brig_function_name): Likewise.
18131 * ipa-hsa.c (process_hsa_functions): Likewise.
18132 (ipa_hsa_write_summary): Likewise.
18133 * symbol-summary.h (symtab_duplication): Use ::get function/
18134 (get): New function.
18136 2018-06-08 Martin Liska <mliska@suse.cz>
18138 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
18140 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
18141 (hsa_register_kernel): Likewise.
18142 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
18143 * hsa-gen.c (hsa_get_host_function): Likewise.
18144 (get_brig_function_name): Likewise.
18145 (generate_hsa): Likewise.
18146 (pass_gen_hsail::execute): Likewise.
18147 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
18148 (devirtualization_time_bonus): Likewise.
18149 (ipcp_propagate_stage): Likewise.
18150 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
18151 (edge_set_predicate): Likewise.
18152 (evaluate_conditions_for_known_args): Likewise.
18153 (evaluate_properties_for_edge): Likewise.
18154 (ipa_fn_summary::reset): Likewise.
18155 (ipa_fn_summary_t::duplicate): Likewise.
18156 (dump_ipa_call_summary): Likewise.
18157 (ipa_dump_fn_summary): Likewise.
18158 (analyze_function_body): Likewise.
18159 (compute_fn_summary): Likewise.
18160 (estimate_edge_devirt_benefit): Likewise.
18161 (estimate_edge_size_and_time): Likewise.
18162 (estimate_calls_size_and_time): Likewise.
18163 (estimate_node_size_and_time): Likewise.
18164 (inline_update_callee_summaries): Likewise.
18165 (remap_edge_change_prob): Likewise.
18166 (remap_edge_summaries): Likewise.
18167 (ipa_merge_fn_summary_after_inlining): Likewise.
18168 (ipa_update_overall_fn_summary): Likewise.
18169 (read_ipa_call_summary): Likewise.
18170 (inline_read_section): Likewise.
18171 (write_ipa_call_summary): Likewise.
18172 (ipa_fn_summary_write): Likewise.
18173 (ipa_free_fn_summary): Likewise.
18174 * ipa-hsa.c (process_hsa_functions): Likewise.
18175 (ipa_hsa_write_summary): Likewise.
18176 (ipa_hsa_read_section): Likewise.
18177 * ipa-icf.c (sem_function::merge): Likewise.
18178 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
18179 (do_estimate_edge_time): Likewise.
18180 (estimate_size_after_inlining): Likewise.
18181 (estimate_growth): Likewise.
18182 (growth_likely_positive): Likewise.
18183 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
18184 (inline_call): Likewise.
18185 * ipa-inline.c (caller_growth_limits): Likewise.
18186 (can_inline_edge_p): Likewise.
18187 (can_inline_edge_by_limits_p): Likewise.
18188 (compute_uninlined_call_time): Likewise.
18189 (compute_inlined_call_time): Likewise.
18190 (want_inline_small_function_p): Likewise.
18191 (edge_badness): Likewise.
18192 (update_caller_keys): Likewise.
18193 (update_callee_keys): Likewise.
18194 (recursive_inlining): Likewise.
18195 (inline_small_functions): Likewise.
18196 (inline_to_all_callers_1): Likewise.
18197 (dump_overall_stats): Likewise.
18198 (early_inline_small_functions): Likewise.
18199 (early_inliner): Likewise.
18200 * ipa-inline.h (estimate_edge_growth): Likewise.
18201 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
18202 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
18203 * ipa-prop.h (IPA_NODE_REF): Likewise.
18204 (IPA_EDGE_REF): Likewise.
18205 * ipa-pure-const.c (malloc_candidate_p): Likewise.
18206 (propagate_malloc): Likewise.
18207 * ipa-split.c (execute_split_functions): Likewise.
18208 * symbol-summary.h: Rename get to get_create.
18210 (get_create): Likewise.
18211 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
18213 2018-06-08 Martin Liska <mliska@suse.cz>
18215 * symbol-summary.h (release): Move definition out of class
18217 (symtab_removal): Likewise.
18218 (symtab_duplication): Likewise.
18220 2018-06-08 Martin Liska <mliska@suse.cz>
18222 * symbol-summary.h (function_summary): Move constructor
18223 implementation out of class declaration.
18224 (release): Likewise.
18225 (symtab_insertion): Likewise.
18226 (symtab_removal): Likewise.
18227 (symtab_duplication): Likewise.
18230 2018-06-08 Martin Liska <mliska@suse.cz>
18232 * Makefile.in: Remove support for MPX (macros, related functions,
18233 fields in cgraph_node, ...).
18234 * builtin-types.def (BT_BND): Likewise.
18235 (BT_FN_BND_CONST_PTR): Likewise.
18236 (BT_FN_CONST_PTR_BND): Likewise.
18237 (BT_FN_VOID_PTR_BND): Likewise.
18238 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
18239 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
18240 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
18241 (expand_builtin_mempcpy_with_bounds): Likewise.
18242 (expand_builtin_memset_with_bounds): Likewise.
18243 (expand_builtin_memset_args): Likewise.
18244 (std_expand_builtin_va_start): Likewise.
18245 (expand_builtin): Likewise.
18246 (expand_builtin_with_bounds): Likewise.
18247 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
18248 (DEF_LIB_BUILTIN_CHKP): Likewise.
18249 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
18250 (DEF_CHKP_BUILTIN): Likewise.
18251 (BUILT_IN_MEMCPY): Likewise.
18252 (BUILT_IN_MEMMOVE): Likewise.
18253 (BUILT_IN_MEMPCPY): Likewise.
18254 (BUILT_IN_MEMSET): Likewise.
18255 (BUILT_IN_STPCPY): Likewise.
18256 (BUILT_IN_STRCAT): Likewise.
18257 (BUILT_IN_STRCHR): Likewise.
18258 (BUILT_IN_STRCPY): Likewise.
18259 (BUILT_IN_STRLEN): Likewise.
18260 (BUILT_IN_MEMCPY_CHK): Likewise.
18261 (BUILT_IN_MEMMOVE_CHK): Likewise.
18262 (BUILT_IN_MEMPCPY_CHK): Likewise.
18263 (BUILT_IN_MEMSET_CHK): Likewise.
18264 (BUILT_IN_STPCPY_CHK): Likewise.
18265 (BUILT_IN_STRCAT_CHK): Likewise.
18266 (BUILT_IN_STRCPY_CHK): Likewise.
18267 * calls.c (store_bounds): Likewise.
18268 (emit_call_1): Likewise.
18269 (special_function_p): Likewise.
18270 (maybe_warn_nonstring_arg): Likewise.
18271 (initialize_argument_information): Likewise.
18272 (finalize_must_preallocate): Likewise.
18273 (compute_argument_addresses): Likewise.
18274 (expand_call): Likewise.
18275 * cfgexpand.c (expand_call_stmt): Likewise.
18276 (expand_return): Likewise.
18277 (expand_gimple_stmt_1): Likewise.
18278 (pass_expand::execute): Likewise.
18279 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
18280 (cgraph_node::remove): Likewise.
18281 (cgraph_node::dump): Likewise.
18282 (cgraph_node::verify_node): Likewise.
18283 * cgraph.h (chkp_function_instrumented_p): Likewise.
18284 (symtab_node::get_alias_target): Likewise.
18285 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
18286 (cgraph_local_p): Likewise.
18287 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
18288 (cgraph_edge::rebuild_references): Likewise.
18289 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
18290 (walk_polymorphic_call_targets): Likewise.
18291 (cgraph_node::expand_thunk): Likewise.
18292 (symbol_table::output_weakrefs): Likewise.
18293 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
18294 (ix86_handle_option): Likewise.
18295 * config/i386/constraints.md: Likewise.
18296 * config/i386/i386-builtin-types.def (BND): Likewise.
18300 * config/i386/i386-builtin.def (BDESC_END): Likewise.
18301 (BDESC_FIRST): Likewise.
18303 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
18304 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
18305 * config/i386/i386.c (enum reg_class): Likewise.
18306 (ix86_target_string): Likewise.
18307 (ix86_option_override_internal): Likewise.
18308 (ix86_conditional_register_usage): Likewise.
18309 (ix86_valid_target_attribute_inner_p): Likewise.
18310 (ix86_set_indirect_branch_type): Likewise.
18311 (ix86_set_current_function): Likewise.
18312 (ix86_function_arg_regno_p): Likewise.
18313 (init_cumulative_args): Likewise.
18314 (ix86_function_arg_advance): Likewise.
18315 (ix86_function_arg): Likewise.
18316 (ix86_pass_by_reference): Likewise.
18317 (ix86_function_value_regno_p): Likewise.
18318 (ix86_function_value_1): Likewise.
18319 (ix86_function_value_bounds): Likewise.
18320 (ix86_return_in_memory): Likewise.
18321 (ix86_setup_incoming_vararg_bounds): Likewise.
18322 (ix86_va_start): Likewise.
18323 (indirect_thunk_need_prefix): Likewise.
18324 (print_reg): Likewise.
18325 (ix86_print_operand): Likewise.
18326 (ix86_expand_call): Likewise.
18327 (ix86_output_function_return): Likewise.
18328 (reg_encoded_number): Likewise.
18329 (BDESC_VERIFYS): Likewise.
18330 (ix86_init_mpx_builtins): Likewise.
18331 (ix86_init_builtins): Likewise.
18332 (ix86_emit_cmove): Likewise.
18333 (ix86_emit_move_max): Likewise.
18334 (ix86_expand_builtin): Likewise.
18335 (ix86_builtin_mpx_function): Likewise.
18336 (ix86_get_arg_address_for_bt): Likewise.
18337 (ix86_load_bounds): Likewise.
18338 (ix86_store_bounds): Likewise.
18339 (ix86_load_returned_bounds): Likewise.
18340 (ix86_store_returned_bounds): Likewise.
18341 (ix86_class_likely_spilled_p): Likewise.
18342 (ix86_hard_regno_mode_ok): Likewise.
18343 (x86_order_regs_for_local_alloc): Likewise.
18344 (ix86_mitigate_rop): Likewise.
18345 (ix86_bnd_prefixed_insn_p): Likewise.
18346 (ix86_mpx_bound_mode): Likewise.
18347 (ix86_make_bounds_constant): Likewise.
18348 (ix86_initialize_bounds): Likewise.
18349 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
18350 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
18351 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
18352 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
18353 (TARGET_CHKP_BOUND_MODE): Likewise.
18354 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
18355 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
18356 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
18357 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
18358 * config/i386/i386.h (TARGET_MPX): Likewise.
18359 (TARGET_MPX_P): Likewise.
18360 (VALID_BND_REG_MODE): Likewise.
18361 (FIRST_BND_REG): Likewise.
18362 (LAST_BND_REG): Likewise.
18363 (enum reg_class): Likewise.
18364 (BND_REG_P): Likewise.
18365 (BND_REGNO_P): Likewise.
18366 (BNDmode): Likewise.
18367 (ADJUST_INSN_LENGTH): Likewise.
18368 * config/i386/i386.md: Likewise.
18369 * config/i386/i386.opt: Likewise.
18370 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
18371 (defined): Likewise.
18372 (LINK_MPX): Likewise.
18373 (MPX_SPEC): Likewise.
18374 (LIBMPX_SPEC): Likewise.
18375 (LIBMPXWRAPPERS_SPEC): Likewise.
18376 (CHKP_SPEC): Likewise.
18377 * config/i386/predicates.md: Likewise.
18378 * dbxout.c (dbxout_type): Likewise.
18379 * doc/extend.texi: Likewise.
18380 * doc/invoke.texi: Likewise.
18381 * doc/md.texi: Likewise.
18382 * doc/tm.texi: Likewise.
18383 * doc/tm.texi.in: Likewise.
18384 * dwarf2out.c (is_base_type): Likewise.
18385 (gen_formal_types_die): Likewise.
18386 (gen_subprogram_die): Likewise.
18387 (gen_type_die_with_usage): Likewise.
18388 (gen_decl_die): Likewise.
18389 (dwarf2out_late_global_decl): Likewise.
18390 * expr.c (expand_assignment): Likewise.
18391 (emit_storent_insn): Likewise.
18392 (store_expr_with_bounds): Likewise.
18393 (store_expr): Likewise.
18394 (expand_expr_real_1): Likewise.
18395 * expr.h (store_expr_with_bounds): Likewise.
18396 * function.c (use_register_for_decl): Likewise.
18397 (struct bounds_parm_data): Likewise.
18398 (assign_parms_augmented_arg_list): Likewise.
18399 (assign_parm_find_entry_rtl): Likewise.
18400 (assign_parm_is_stack_parm): Likewise.
18401 (assign_parm_load_bounds): Likewise.
18402 (assign_bounds): Likewise.
18403 (assign_parms): Likewise.
18404 (expand_function_start): Likewise.
18405 * gcc.c (CHKP_SPEC): Likewise.
18406 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
18407 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
18408 (wrestrict_dom_walker::check_call): Likewise.
18409 * gimple.c (gimple_build_call_from_tree): Likewise.
18410 * gimple.h (enum gf_mask): Likewise.
18411 (gimple_call_with_bounds_p): Likewise.
18412 (gimple_call_set_with_bounds): Likewise.
18413 * gimplify.c (gimplify_init_constructor): Likewise.
18414 * ipa-cp.c (initialize_node_lattices): Likewise.
18415 (propagate_constants_across_call): Likewise.
18416 (find_more_scalar_values_for_callers_subset): Likewise.
18417 * ipa-hsa.c (process_hsa_functions): Likewise.
18418 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
18419 * ipa-icf.c (sem_function::merge): Likewise.
18420 * ipa-inline.c (early_inliner): Likewise.
18421 * ipa-pure-const.c (warn_function_noreturn): Likewise.
18422 (warn_function_cold): Likewise.
18423 (propagate_pure_const): Likewise.
18424 * ipa-ref.h (enum GTY): Likewise.
18425 * ipa-split.c (find_retbnd): Likewise.
18426 (consider_split): Likewise.
18427 (split_function): Likewise.
18428 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
18429 * ipa.c (walk_polymorphic_call_targets): Likewise.
18430 (symbol_table::remove_unreachable_nodes): Likewise.
18431 (process_references): Likewise.
18432 (cgraph_build_static_cdtor_1): Likewise.
18433 * lto-cgraph.c (lto_output_node): Likewise.
18434 (output_refs): Likewise.
18435 (compute_ltrans_boundary): Likewise.
18436 (input_overwrite_node): Likewise.
18437 (input_node): Likewise.
18438 (input_cgraph_1): Likewise.
18439 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
18440 * passes.c (pass_manager::execute_early_local_passes): Likewise.
18441 (class pass_chkp_instrumentation_passes): Likewise.
18442 (make_pass_chkp_instrumentation_passes): Likewise.
18443 * passes.def: Likewise.
18444 * rtl.h (struct GTY): Likewise.
18445 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
18446 * stor-layout.c (layout_type): Likewise.
18447 * symtab.c: Likewise.
18448 * target.def: Likewise.
18449 * targhooks.c (default_chkp_bound_type): Likewise.
18450 (default_chkp_bound_mode): Likewise.
18451 (default_builtin_chkp_function): Likewise.
18452 (default_chkp_function_value_bounds): Likewise.
18453 (default_chkp_make_bounds_constant): Likewise.
18454 (default_chkp_initialize_bounds): Likewise.
18455 * targhooks.h (default_chkp_bound_type): Likewise.
18456 (default_chkp_bound_mode): Likewise.
18457 (default_builtin_chkp_function): Likewise.
18458 (default_chkp_function_value_bounds): Likewise.
18459 (default_chkp_make_bounds_constant): Likewise.
18460 (default_chkp_initialize_bounds): Likewise.
18461 * toplev.c (compile_file): Likewise.
18462 (process_options): Likewise.
18463 * tree-core.h (DEF_BUILTIN): Likewise.
18464 (DEF_BUILTIN_CHKP): Likewise.
18465 * tree-inline.c (declare_return_variable): Likewise.
18466 (remap_gimple_stmt): Likewise.
18467 (copy_bb): Likewise.
18468 (initialize_inlined_parameters): Likewise.
18469 (expand_call_inline): Likewise.
18470 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
18471 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
18472 (make_pass_ipa_chkp_produce_thunks): Likewise.
18473 (make_pass_chkp): Likewise.
18474 (make_pass_chkp_opt): Likewise.
18475 (make_pass_chkp_instrumentation_passes): Likewise.
18476 * tree-pretty-print.c (dump_generic_node): Likewise.
18477 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
18478 * tree-ssa-dce.c (propagate_necessity): Likewise.
18479 (eliminate_unnecessary_stmts): Likewise.
18480 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
18481 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
18482 * tree-ssa-sccvn.h: Likewise.
18483 * tree-ssa-strlen.c (get_string_length): Likewise.
18484 (valid_builtin_call): Likewise.
18485 (adjust_last_stmt): Likewise.
18486 (handle_builtin_strchr): Likewise.
18487 (handle_builtin_strcpy): Likewise.
18488 (handle_builtin_stxncpy): Likewise.
18489 (handle_builtin_memcpy): Likewise.
18490 (handle_builtin_strcat): Likewise.
18491 (strlen_check_and_optimize_stmt): Likewise.
18492 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
18493 * tree-streamer-in.c: Likewise.
18494 * tree-streamer.c (record_common_node): Likewise.
18495 * tree.c (tree_code_size): Likewise.
18496 (wide_int_to_tree_1): Likewise.
18497 (type_contains_placeholder_1): Likewise.
18498 (build_common_tree_nodes): Likewise.
18499 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
18500 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
18501 (POINTER_BOUNDS_P): Likewise.
18502 (BOUNDED_TYPE_P): Likewise.
18503 (BOUNDED_P): Likewise.
18504 (CALL_WITH_BOUNDS_P): Likewise.
18505 (pointer_bounds_type_node): Likewise.
18506 * value-prof.c (gimple_ic): Likewise.
18507 * var-tracking.c (vt_add_function_parameters): Likewise.
18508 * varasm.c (make_decl_rtl): Likewise.
18509 (assemble_start_function): Likewise.
18510 (output_constant): Likewise.
18511 (maybe_assemble_visibility): Likewise.
18512 * varpool.c (ctor_for_folding): Likewise.
18513 * chkp-builtins.def: Remove.
18514 * ipa-chkp.c: Remove.
18515 * ipa-chkp.h: Remove.
18516 * rtl-chkp.c: Remove.
18517 * rtl-chkp.h: Remove.
18518 * tree-chkp-opt.c: Remove.
18519 * tree-chkp.c: Remove.
18520 * tree-chkp.h: Remove.
18522 2018-06-07 Carl Love <cel@us.ibm.com>
18524 * config/rs6000/vsx.md (vextract_fp_from_shorth,
18525 vextract_fp_from_shortl): Add BE support.
18527 2018-06-07 Paul Koning <ni1d@arrl.net>
18529 * compare-elim.c (try_merge_compare): Don't merge compare if
18530 address contains a side effect.
18531 (try_eliminate_compare): Likewise.
18533 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
18535 * config.gcc: Support "tremont".
18536 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
18537 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18539 * config/i386/i386.c (m_TREMONT): Define.
18540 (processor_target_table): Add "tremont".
18541 (PTA_TREMONT): Define.
18542 (ix86_lea_outperforms): Add TARGET_TREMONT.
18543 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
18544 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
18545 and M_INTEL_GOLDMONT_PLUS.
18546 (fold_builtin_cpu): Add "tremont".
18547 (ix86_add_stmt_cost): Add TARGET_TREMONT.
18548 (ix86_option_override_internal): Add "tremont".
18549 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
18550 (processor_type): Add PROCESSOR_TREMONT.
18551 * config/i386/x86-tune.def: Add m_TREMONT.
18552 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
18554 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
18556 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
18557 symbol defined for msp430i* devices to be lower case.
18559 2018-06-07 Richard Biener <rguenther@suse.de>
18561 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
18562 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
18563 Properly wrap signed arithmetic if overflow wraps.
18565 2018-06-07 Jakub Jelinek <jakub@redhat.com>
18567 PR tree-optimization/69615
18568 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
18569 of a cast from a same precision integral SSA_NAME in a bb dominated
18570 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
18571 cast to utype if rhs2 has already a compatible type.
18573 2018-06-07 Richard Biener <rguenther@suse.de>
18575 PR tree-optimization/85935
18576 * graphite-scop-detection.c (find_params_in_bb): Analyze
18577 condition operands with respect to the correct loop. Assert
18578 the analysis doesn't fail.
18580 2018-06-04 Carl Love <cel@us.ibm.com>
18582 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
18583 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
18584 as it is slightly cheaper.
18585 (first_match_or_eos_index_<mode>):
18586 Calculate index using natural element order.
18587 (first_match_index_<mode>):
18588 Calculate index using natural element order.
18589 (first_match_or_eos_index_<mode>):
18590 Calculate index using natural order.
18591 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
18592 for BE and LE modes.
18593 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
18594 P9V_BUILTIN_VCLZLSBB_V16QI.
18595 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
18598 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
18600 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
18601 indentation and line wrap for many prototypes. Add missing
18602 @smallexample directives around block of prototypes for vec_xl and
18605 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
18607 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
18608 track if we pass or return IEEE 128-bit floating point.
18609 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
18610 C++ mangling that is compatible with GCC 8.1.
18611 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
18612 (init_cumulative_args): Note if we pass or return IEEE 128-bit
18613 floating point types.
18614 (rs6000_function_arg_advance_1): Likewise.
18615 (rs6000_mangle_type): Optionally generate mangled names that match
18616 what GCC 8.1 generated for IEEE 128-bit floating point types.
18617 (rs6000_globalize_decl_name): If we have an external function that
18618 passes or returns IEEE 128-bit types, generate a weak reference
18619 from the mangled name used in GCC 8.1 to the current mangled
18621 (rs6000_init_builtins): Make __ibm128 use the long double type if
18622 long double is IBM extended double. Make __float128 use the long
18623 double type if long double is IEEE 128-bit.
18626 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
18627 macro for __ibm128 built-in functions.
18628 (PACK_IF): Add __ibm128 pack/unpack functions.
18629 (UNPACK_IF): Likewise.
18630 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
18631 enable long double built-in functions if long double is IEEE
18632 128-bit floating point.
18633 (rs6000_invalid_builtin): Update long double built-in function
18635 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
18636 functions, adjust the built-in function to use the long double
18637 built-in function if __ibm128 and long double are the same type.
18638 * doc/extend.texi (PowerPC builtins): Update documention for
18639 __builtin_{,un}pack_longdouble. Add documentation for
18640 __builtin_{,un}pack_ibm128.
18642 2018-06-06 Jim Wilson <jimw@sifive.com>
18644 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
18645 (struct machine_function): New field interrupt_mode.
18646 (riscv_handle_type_attribute): New function. Add forward declaration.
18647 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
18648 (riscv_expand_epilogue): Check interrupt_mode field.
18649 (riscv_set_current_function): Check interrupt attribute args and
18650 set interrupt_mode field.
18651 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
18652 (riscv_sret, riscv_uret): New.
18653 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
18654 new arguments to interrupt attribute.
18656 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
18659 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
18660 Don't handle -mcpu=power8 if -mpower9-vector is also used.
18662 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
18664 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
18665 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
18666 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
18667 several redundant entries.
18669 2018-06-06 David Malcolm <dmalcolm@redhat.com>
18671 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
18672 type from "rtx" to "rtx_insn *".
18673 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
18674 for local "call_insn", removing cast.
18675 (ix86_expand_call): Likewise, introducing a "call_insn" local.
18677 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
18679 PR tree-optimization/86066
18680 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
18681 for BIT_INSERT_EXPR stores.
18683 2018-06-06 Richard Biener <rguenther@suse.de>
18685 PR tree-optimization/86062
18686 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
18687 component refs ontop
18688 of to be offsetted base.
18690 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
18692 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
18693 to be static and remove check on interrupt attribute name.
18695 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
18697 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
18698 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
18700 2018-06-05 Steve Ellcey <sellcey@cavium.com>
18703 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
18705 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
18706 Remove second argument, change how error is called.
18707 (aarch64_layout_arg): Remove second argument from
18708 aarch64_err_no_fpadvsimd call.
18709 (aarch64_init_cumulative_args): Ditto.
18710 (aarch64_gimplify_va_arg_expr): Ditto.
18711 * config/aarch64/aarch64.md (mov<mode>): Ditto.
18713 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
18715 * config/i386/i386.md (simple_return_indirect_internal): New expander.
18716 (*simple_return_indirect_internal<mode>): Rename from
18717 simple_return_indirect_internal. Use W mode iterator.
18718 (rstorssp): New expander.
18719 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
18720 (clrssbsy): New expander.
18721 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
18723 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
18725 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
18727 (cmse_check_pointed_object): Likewise.
18729 2018-06-05 Martin Liska <mliska@suse.cz>
18731 PR gcov-profile/47618
18732 * doc/invoke.texi: Document how -fprofile-dir format
18735 2018-06-05 Richard Biener <rguenther@suse.de>
18737 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
18738 removal pretend DOM info isn't available so we do not update
18739 it and only remove edges, not dominated blocks. Actually free
18740 DOM info in case we removed something. Remove unreachable blocks.
18741 (mfb_keep_latches): Work with either DOM info or marked backedges.
18742 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
18743 first. Mark backedges if DOM info isn't available.
18744 (Re-)compute DOM info after cleanup_control_flow_pre.
18746 2018-06-05 Richard Biener <rguenther@suse.de>
18748 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
18749 (locus_discrim_hasher::hash): Adjust.
18750 (locus_discrim_hasher::equal): Likewise.
18751 (next_discriminator_for_locus): Work on line directly.
18752 (same_line_p): Pass in expanded locus1 as well.
18753 (assign_discriminators): Avoid redundant location expansions.
18755 2018-06-05 Richard Biener <rguenther@suse.de>
18757 PR tree-optimization/86046
18758 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
18759 if required after clearing TREE_ADDRESSABLE.
18761 2018-06-05 Richard Biener <rguenther@suse.de>
18763 PR tree-optimization/86047
18764 * tree-ssa-loop.c (for_each_index): Glob handling of all
18765 decls and constants and really handle all of them.
18767 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18770 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
18771 qualifier_void_pointer and qualifier_const_void_pointer.
18772 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
18773 (arm_init_builtins): Handle the above.
18774 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
18775 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
18776 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
18779 2018-06-05 Martin Liska <mliska@suse.cz>
18781 * auto-profile.c (read_autofdo_file): Do not use
18782 gcov_ctr_summary struct.
18783 (afdo_callsite_hot_enough_for_early_inline): Likewise.
18784 * coverage.c (struct counts_entry): Likewise.
18785 (read_counts_file): Read just single summary entry.
18786 (get_coverage_counts): Use gcov_summary struct.
18787 * coverage.h (get_coverage_counts): Likewise.
18788 * gcov-dump.c (dump_working_sets): Likewise.
18789 (tag_summary): Dump just single summary.
18790 * gcov-io.c (gcov_write_summary): Write just histogram
18792 (gcov_read_summary): Read just single summary.
18793 (compute_working_sets): Use gcov_summary struct.
18794 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
18795 of GCOV_COUNTERS_SUMMABLE.
18796 (GCOV_COUNTERS_SUMMABLE): Remove.
18797 (GCOV_FIRST_VALUE_COUNTER): Replace with
18798 GCOV_COUNTER_V_INTERVAL.
18799 (struct gcov_ctr_summary): Remove.
18800 (struct gcov_summary): Directly use fields of former
18802 (compute_working_sets): Use gcov_summary struct.
18803 * gcov.c (read_count_file): Do not use ctrs fields.
18804 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
18806 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
18808 * profile.c: Likewise.
18809 * profile.h: Likewise.
18811 2018-06-05 Martin Liska <mliska@suse.cz>
18813 PR gcov-profile/84846
18814 * gcov.c (output_lines): Print working directory only
18815 in intermediate format.
18817 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
18819 * config/s390/s390-builtin-types.def: Add void function type.
18820 * config/s390/s390-builtins.def: Use the function type for the
18823 2018-06-04 Jim Wilson <jimw@sifive.com>
18825 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
18827 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
18828 and maybe_eh_return. Change regno to unsigned int. Use new args to
18829 handle EH_RETURN_DATA_REGNO registers properly.
18830 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
18831 (riscv_expand_epilogue): Update comment. Change argument name and
18832 type. Update code to use new name and type. Pass new args to
18833 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
18835 * config/riscv/riscv.md (NORMAL_RETURN): New.
18836 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
18837 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
18838 (eh_return): Call gen_eh_return_internal and emit barrier.
18839 (eh_return_internal): Call riscv_expand_epilogue.
18841 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
18843 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
18844 bit_insertion field and declare can_be_merged_into method.
18845 (merged_store_group::can_be_merged_into): New method.
18846 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
18847 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
18848 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
18850 2018-06-04 Richard Biener <rguenther@suse.de>
18852 PR tree-optimization/85955
18853 * builtins.c (fold_builtin_sincos): Convert pointers to
18854 destination to appropriate type before dereferencing.
18856 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
18858 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
18860 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
18862 * expr.c (expand_expr_real_1): Force the operand into memory if
18863 its TYPE_MODE is BLKmode and if there is no integer mode for
18864 the number of bits being extracted.
18866 2018-06-04 Jakub Jelinek <jakub@redhat.com>
18870 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
18871 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
18873 2018-06-04 Richard Biener <rguenther@suse.de>
18875 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
18876 (cleanup_tree_cfg_noloop): ... single caller. Do
18877 start_recording_case_labels later.
18879 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
18881 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
18882 to _IMMINTRIN_H_INCLUDED.
18883 * config/i386/pconfigintrin.h: Ditto.
18884 * config/i386/waitpkgintrin.h: Ditto.
18885 * config/i386/immintrin.h: Add includes for sgxintrin.h,
18886 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
18887 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
18888 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
18889 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
18890 waitpkgintrin.h and cldemoteintrin.h.
18892 2018-06-04 Richard Biener <rguenther@suse.de>
18894 PR tree-optimization/86038
18895 * tracer.c (find_best_successor): Check probability for
18896 being initialized, bail out if not.
18898 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
18901 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
18902 of bits to ignore when comparing architectures.
18904 2018-06-04 Jakub Jelinek <jakub@redhat.com>
18906 PR tree-optimization/69615
18907 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
18908 maximum or minimum of the type, try to merge it also as if
18909 range1 is + [-, x - 1] or + [x + 1, -].
18912 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
18914 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
18916 PR tree-optimization/86034
18917 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
18918 the unsigned bitfield type in a bit insertion sequence if it does not
18919 have a larger precision than the bitfield size.
18920 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
18922 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
18924 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
18926 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
18928 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
18929 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
18930 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
18931 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
18933 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
18935 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
18936 Disable -fdelete-null-pointer-checks for ELF toolchain.
18938 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
18939 Kito Cheng <kito.cheng@gmail.com>
18941 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
18942 (nds32le-*-*, nds32be-*-*): Integrate checking process.
18943 (nds32*-*-*): Add glibc and uclibc conditions.
18944 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
18945 (TARGET_EXCEPT_UNWIND_INFO): Define.
18946 * config/nds32/elf.h: New file.
18947 * config/nds32/linux.h: New file.
18948 * config/nds32/nds32-elf.opt: New file.
18949 * config/nds32/nds32-linux.opt: New file.
18950 * config/nds32/nds32-fp-as-gp.c
18951 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
18952 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
18954 (nds32_asm_file_end): Ditto.
18955 (nds32_print_operand): Ditto.
18956 (nds32_insert_attributes): Ditto.
18957 (nds32_init_libfuncs): New function.
18958 (TARGET_HAVE_TLS): Define.
18959 (TARGET_INIT_LIBFUNCS): Define.
18960 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
18962 (TARGET_ELF): Apply different mcmodel setting.
18963 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
18964 been migrated into elf.h and linux.h files.
18965 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
18966 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
18967 (mcmodel): The content has been migrated into nds32-elf.opt and
18968 nds32-linux.opt files.
18969 * config/nds32/t-elf: New file.
18970 * config/nds32/t-linux: New file.
18972 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
18973 Shiva Chen <shiva0217@gmail.com>
18975 * config/nds32/constants.md (unspec_volatile_element): Add
18976 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
18977 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
18979 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
18980 (make_pass_nds32_fp_as_gp): Declare.
18981 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
18983 (nds32_asm_function_end_prologue): Remove unused asm output.
18984 (nds32_asm_function_begin_epilogue): Remove unused asm output.
18985 (nds32_asm_file_start): Output necessary fp_as_gp information.
18986 (nds32_option_override): Adjust register usage.
18987 (nds32_expand_prologue): Consider fp_as_gp situation.
18988 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
18989 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
18992 (simple_return): Ditto.
18993 (omit_fp_begin): Output special directive for fp_as_gp.
18994 (omit_fp_end): Output special directive for fp_as_gp.
18995 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
18996 mforbid-fp-as-gp): New options.
18998 2018-06-01 Mark Wielaard <mark@klomp.org>
19000 * dwarf2out.c (dwarf2out_finish): Remove generation of
19001 DW_AT_loclists_base.
19003 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
19005 * gimple-ssa-store-merging.c: Include gimple-fold.h.
19006 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
19007 (struct merged_store_group): Add bit_insertion field.
19008 (dump_char_array): Use standard hexadecimal format.
19009 (merged_store_group::merged_store_group): Set bit_insertion to false.
19010 (merged_store_group::apply_stores): Use optimal buffer size. Deal
19011 with BIT_INSERT_EXPR stores. Move up code updating the mask and
19012 also print the mask in the dump file.
19013 (pass_store_merging::gate): Minor tweak.
19014 (imm_store_chain_info::coalesce_immediate): Fix wrong association
19015 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
19016 stores with INTEGER_CST stores.
19017 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
19018 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
19019 and use it throughout. Generate bit insertion sequences if need be.
19020 (pass_store_merging::process_store): Remove redundant condition.
19021 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
19023 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
19025 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
19026 the 128-bit floating point types. Fix function comment.
19028 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19030 * config/aarch64/aarch64-simd.md
19031 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
19033 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
19036 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
19038 PR tree-optimization/85989
19039 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
19041 (backprop::intersect_uses): Check it when deciding whether this
19042 is a backedge reference.
19043 (backprop::process_block): Add each phi to m_visited_phis
19044 after visiting it, then clear it at the end.
19046 2018-06-01 Richard Biener <rguenther@suse.de>
19048 * tree-vectorizer.h (vect_dr_stmt): New function.
19049 (vect_get_load_cost): Adjust.
19050 (vect_get_store_cost): Likewise.
19051 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
19052 Use vect_dr_stmt instead of DR_SMTT.
19053 (vect_record_base_alignments): Likewise.
19054 (vect_calculate_target_alignment): Likewise.
19055 (vect_compute_data_ref_alignment): Likewise and make static.
19056 (vect_update_misalignment_for_peel): Likewise.
19057 (vect_verify_datarefs_alignment): Likewise.
19058 (vector_alignment_reachable_p): Likewise.
19059 (vect_get_data_access_cost): Likewise. Pass down
19060 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
19061 (vect_get_peeling_costs_all_drs): Likewise.
19062 (vect_peeling_hash_get_lowest_cost): Likewise.
19063 (vect_enhance_data_refs_alignment): Likewise.
19064 (vect_find_same_alignment_drs): Likewise.
19065 (vect_analyze_data_refs_alignment): Likewise.
19066 (vect_analyze_group_access_1): Likewise.
19067 (vect_analyze_group_access): Likewise.
19068 (vect_analyze_data_ref_access): Likewise.
19069 (vect_analyze_data_ref_accesses): Likewise.
19070 (vect_vfa_segment_size): Likewise.
19071 (vect_small_gap_p): Likewise.
19072 (vectorizable_with_step_bound_p): Likewise.
19073 (vect_prune_runtime_alias_test_list): Likewise.
19074 (vect_analyze_data_refs): Likewise.
19075 (vect_supportable_dr_alignment): Likewise.
19076 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
19077 (vect_gen_prolog_loop_niters): Likewise.
19078 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
19079 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
19081 (vect_recog_mask_conversion_pattern): Likewise.
19082 (vect_try_gather_scatter_pattern): Likewise.
19083 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
19084 to vect_get_store_cost.
19085 (vect_get_store_cost): Get stmt_info instead of DR.
19086 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
19087 (vect_get_load_cost): Get stmt_info instead of DR.
19089 2018-06-01 Richard Biener <rguenther@suse.de>
19091 PR middle-end/86017
19092 * gimple-fold.c (var_decl_component_p): Also allow offsetted
19093 vars wrapped in MEM_REFs.
19095 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
19097 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
19098 Fix subreg tests so that we only return a choice between
19099 GENERAL_REGS and FP_REGS if the original classes included both.
19101 2018-06-01 Richard Biener <rguenther@suse.de>
19104 * tree-ssa-structalias.c (get_function_part_constraint):
19105 Handle NULL fi->decl.
19106 (find_func_aliases_for_call): Properly handle indirect
19107 fi from direct call.
19108 (find_func_clobbers): Likewise.
19109 (ipa_pta_execute): Likewise.
19110 (create_variable_info_for): For functions that are ifunc_resolver
19111 resolve to a varinfo that contains the result of the resolver call.
19112 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
19115 2018-05-31 Michael Collison <michael.collison@arm.com>
19117 * config/aarch64/aarch64.md:
19118 (*fix_to_zero_extenddfdi2): New pattern.
19119 * gcc.target/aarch64/fix_extend1.c: New testcase.
19121 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
19123 PR middle-end/78809
19124 PR middle-end/83026
19125 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
19126 and BUILT_IN_STRNCMP_EQ.
19127 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
19128 BUILT_IN_STRNCMP_EQ.
19129 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
19130 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
19131 (gimple_fold_builtin): Likewise.
19132 * tree-ssa-strlen.c (compute_string_length): New function.
19133 (determine_min_obsize): New function.
19134 (handle_builtin_string_cmp): New function to handle calls to
19135 string compare functions.
19136 (strlen_optimize_stmt): Add handling to builtin string compare
19138 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
19139 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
19140 * tree.c (build_common_builtin_nodes): Add new defines of
19141 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
19143 2018-05-31 Jakub Jelinek <jakub@redhat.com>
19146 * bb-reorder.c (pass_partition_blocks::gate): Return false for
19147 functions with naked attribute.
19149 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
19151 * config/i386/sse.md (avx_vec_concat<mode>):
19152 Substitute concat_tg_mode mode attribute with xtg_mode.
19153 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
19154 (concat_tg_mode): Remove mode attribute.
19156 2018-05-31 Martin Sebor <msebor@redhat.com>
19159 * calls.c (alloc_max_size): Correct a logic error/typo.
19160 Treat excessive arguments as infinite. Warn for invalid arguments.
19161 * doc/invoke.texi (-Walloc-size-larger-than): Update.
19163 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
19166 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
19167 and movx for Haswell.
19169 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
19170 Cesar Philippidis <cesar@codesourcery.com>
19172 PR middle-end/85879
19173 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
19174 when emitting error on private/firstprivate reductions.
19175 * omp-low.c (lower_omp_target): Avoid reference-type processing
19176 on pointers for firstprivate clause.
19178 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
19180 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
19183 * config/aarch64/aarch64-simd.md
19184 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
19185 (aarch64_ld1_x3_<mode>): Likewise
19186 (aarch64_st1x2<VALLDIF:mode>): Likewise
19187 (aarch64_st1_x2_<mode>): Likewise
19188 (aarch64_st1x3<VALLDIF:mode>): Likewise
19189 (aarch64_st1_x3_<mode>): Likewise
19190 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
19191 (vld1_s8_x3): Likewise.
19192 (vld1_u16_x3): Likewise.
19193 (vld1_s16_x3): Likewise.
19194 (vld1_u32_x3): Likewise.
19195 (vld1_s32_x3): Likewise.
19196 (vld1_u64_x3): Likewise.
19197 (vld1_s64_x3): Likewise.
19198 (vld1_f16_x3): Likewise.
19199 (vld1_f32_x3): Likewise.
19200 (vld1_f64_x3): Likewise.
19201 (vld1_p8_x3): Likewise.
19202 (vld1_p16_x3): Likewise.
19203 (vld1_p64_x3): Likewise.
19204 (vld1q_u8_x3): Likewise.
19205 (vld1q_s8_x3): Likewise.
19206 (vld1q_u16_x3): Likewise.
19207 (vld1q_s16_x3): Likewise.
19208 (vld1q_u32_x3): Likewise.
19209 (vld1q_s32_x3): Likewise.
19210 (vld1q_u64_x3): Likewise.
19211 (vld1q_s64_x3): Likewise.
19212 (vld1q_f16_x3): Likewise.
19213 (vld1q_f32_x3): Likewise.
19214 (vld1q_f64_x3): Likewise.
19215 (vld1q_p8_x3): Likewise.
19216 (vld1q_p16_x3): Likewise.
19217 (vld1q_p64_x3): Likewise.
19218 (vst1_s64_x2): Likewise.
19219 (vst1_u64_x2): Likewise.
19220 (vst1_f64_x2): Likewise.
19221 (vst1_s8_x2): Likewise.
19222 (vst1_p8_x2): Likewise.
19223 (vst1_s16_x2): Likewise.
19224 (vst1_p16_x2): Likewise.
19225 (vst1_s32_x2): Likewise.
19226 (vst1_u8_x2): Likewise.
19227 (vst1_u16_x2): Likewise.
19228 (vst1_u32_x2): Likewise.
19229 (vst1_f16_x2): Likewise.
19230 (vst1_f32_x2): Likewise.
19231 (vst1_p64_x2): Likewise.
19232 (vst1q_s8_x2): Likewise.
19233 (vst1q_p8_x2): Likewise.
19234 (vst1q_s16_x2): Likewise.
19235 (vst1q_p16_x2): Likewise.
19236 (vst1q_s32_x2): Likewise.
19237 (vst1q_s64_x2): Likewise.
19238 (vst1q_u8_x2): Likewise.
19239 (vst1q_u16_x2): Likewise.
19240 (vst1q_u32_x2): Likewise.
19241 (vst1q_u64_x2): Likewise.
19242 (vst1q_f16_x2): Likewise.
19243 (vst1q_f32_x2): Likewise.
19244 (vst1q_f64_x2): Likewise.
19245 (vst1q_p64_x2): Likewise.
19246 (vst1_s64_x3): Likewise.
19247 (vst1_u64_x3): Likewise.
19248 (vst1_f64_x3): Likewise.
19249 (vst1_s8_x3): Likewise.
19250 (vst1_p8_x3): Likewise.
19251 (vst1_s16_x3): Likewise.
19252 (vst1_p16_x3): Likewise.
19253 (vst1_s32_x3): Likewise.
19254 (vst1_u8_x3): Likewise.
19255 (vst1_u16_x3): Likewise.
19256 (vst1_u32_x3): Likewise.
19257 (vst1_f16_x3): Likewise.
19258 (vst1_f32_x3): Likewise.
19259 (vst1_p64_x3): Likewise.
19260 (vst1q_s8_x3): Likewise.
19261 (vst1q_p8_x3): Likewise.
19262 (vst1q_s16_x3): Likewise.
19263 (vst1q_p16_x3): Likewise.
19264 (vst1q_s32_x3): Likewise.
19265 (vst1q_s64_x3): Likewise.
19266 (vst1q_u8_x3): Likewise.
19267 (vst1q_u16_x3): Likewise.
19268 (vst1q_u32_x3): Likewise.
19269 (vst1q_u64_x3): Likewise.
19270 (vst1q_f16_x3): Likewise.
19271 (vst1q_f32_x3): Likewise.
19272 (vst1q_f64_x3): Likewise.
19273 (vst1q_p64_x3): Likewise.
19275 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
19277 * config/msp430/msp430.c (msp430_output_labelref): Prepend
19278 user_label_prefix to name.
19280 * tree-core.h: Update comment about the format of NAME string
19281 passed to handler in attribute_spec.
19283 * config/msp430/msp430.md: Remove erroneous subreg expression from
19284 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
19285 zero_extend{q,h}isi2.
19287 2018-05-30 Borislav Petkov <bp@suse.de>
19289 * doc/extend.texi: Document some architecture specific
19290 constraints and sort entries.
19292 2018-05-30 Martin Sebor <msebor@redhat.com>
19294 PR middle-end/85369
19295 * builtins.c (expand_builtin_stpcpy_1): New function.
19296 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
19297 only if the former succeeds.
19299 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
19301 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
19304 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
19306 * doc/invoke.texi (-flinker-output): Document
19308 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
19310 * passes.c (ipa_write_summaries): Only modify statements if body
19312 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
19313 incrementally linking.
19314 (ipa_passes): Likewise.
19315 * lto-cgraph.c (lto_output_node): When incrementally linking do not
19316 pass down resolution info.
19317 * common.opt (flag_incremental_link): Update info.
19318 * gcc.c (plugin specs): Turn flinker-output=* to
19319 -plugin-opt=-linker-output-known
19320 * toplev.c (compile_file): Also cut compilation when doing incremental
19322 * flag-types. (enum lto_partition_model): Add
19323 LTO_LINKER_OUTPUT_NOLTOREL.
19324 (invoke.texi): Add -flinker-output docs.
19325 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
19326 link same way as WPA; do not stream in dead initializers.
19328 * dwarf2out.c (dwarf2out_die_ref_for_decl,
19329 darf2out_register_external_decl): Support incremental link.
19331 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
19333 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
19335 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
19337 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
19338 it down to simple_object_copy_lto_debug_sections.
19339 (run_gcc): Determine incremental LTO link time and configure
19340 lto1 into non-wpa mode, disable renaming of debug sections.
19342 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
19344 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
19345 descriptions of various incorrectly documented functions.
19347 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
19350 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
19351 address check not strict.
19353 2018-05-30 Richard Biener <rguenther@suse.de>
19355 PR tree-optimization/85964
19356 * tracer.c (better_p): Drop initialized count check, we only
19357 call the function with initialized counts now.
19358 (find_best_successor): Do find a best edge if one
19359 has uninitialized count.
19360 (find_best_predecessor): Likewise. Do BB frequency check only
19361 if count is initialized.
19363 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
19365 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
19366 (aarch64_ldrstr_offset_compare): New.
19367 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
19368 load/store orderings.
19369 (aarch64_gen_adjusted_ldpstp): Likewise.
19371 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
19373 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
19374 Check for subset of GENERAL_REGS and FP_REGS.
19375 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
19378 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
19380 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
19381 and wi::to_poly_offset. Add the current offset and then check
19382 whether the sum fits, rather than using an unchecked addition of
19383 a checked term. Check for a shwi rather than a uhwi.
19384 * expr.c (get_bit_range): Use tree_to_poly_uint64.
19385 (store_constructor): Use poly_int_tree_p.
19386 (expand_expr_real_1): Likewise.
19387 * function.c (assign_temp): Likewise.
19388 * fold-const.c (const_binop): Use poly_int_tree_p and
19389 wi::to_poly_offset.
19390 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
19392 * ipa-icf-gimple.c (func_checker::compare_operand): Use
19393 to_poly_offset for MEM offsets.
19394 * ipa-icf.c (sem_variable::equals): Likewise.
19395 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
19396 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
19397 wi::to_poly_offset for BIT_FIELD_REF offsets.
19398 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
19399 wi::to_poly_offset.
19400 * var-tracking.c (emit_note_insn_var_location): Use
19401 tree_to_poly_uint64.
19403 2018-05-29 Jim Wilson <jimw@sifive.com>
19405 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
19407 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
19410 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
19411 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
19413 (sse4_1_round<mode>2): Use nonimmediate_operand
19414 for operand 1 predicate.
19416 2018-05-29 Martin Sebor <msebor@redhat.com>
19417 Richard Biener <rguenther@suse.de>
19420 * calls.c (get_size_range): Call determine_value_range instead
19421 of get_value_range..
19422 * tree-vrp.h (determine_value_range): Declared new function.
19423 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
19425 2018-05-29 Richard Biener <rguenther@suse.de>
19427 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
19428 sure to use non-pattern stmts for get_earlier_stmt arguments.
19429 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
19430 called on pattern stmts.
19431 (get_later_stmt): Likewise.
19433 2018-05-29 Martin Liska <mliska@suse.cz>
19435 PR gcov-profile/85759
19436 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
19439 2018-05-29 Jakub Jelinek <jakub@redhat.com>
19441 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
19443 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
19447 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
19448 VEC_PACK_FLOAT_EXPR): New tree codes.
19449 * tree-pretty-print.c (op_code_prio): Handle
19450 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
19451 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
19452 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
19453 * tree-inline.c (estimate_operator_cost): Likewise.
19454 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
19455 * fold-const.c (const_binop): Likewise.
19456 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
19457 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
19458 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
19459 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
19460 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
19461 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
19462 * expr.c (expand_expr_real_2): Likewise.
19463 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
19464 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
19465 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
19467 * optabs.c (expand_widen_pattern_expr): For
19468 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
19469 sign from result type rather than operand's type.
19470 (expand_binop_directly): For vec_packu_float_optab and
19471 vec_packs_float_optab allow result type to be different from operand's
19473 * optabs-tree.c (optab_for_tree_code): Handle
19474 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
19475 VEC_PACK_FLOAT_EXPR. Formatting fixes.
19476 * tree-vect-generic.c (expand_vector_operations_1): Handle
19477 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
19478 VEC_PACK_FLOAT_EXPR.
19479 * tree-vect-stmts.c (supportable_widening_operation): Handle
19481 (supportable_narrowing_operation): Handle FLOAT_EXPR.
19482 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
19483 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
19484 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
19485 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
19487 (vec_pack<floatprefix>_float_<mode>): New expander.
19488 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
19490 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
19491 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
19492 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
19493 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
19494 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
19496 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
19497 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
19498 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
19499 VEC_PACK_FLOAT_EXPR): Document.
19501 2018-05-29 Richard Biener <rguenther@suse.de>
19503 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
19505 (stmt_vec_info_vec): Make pointer.
19506 (init_stmt_vec_info_vec): Remove.
19507 (free_stmt_vec_info_vec): Likewise.
19508 (set_stmt_vec_info_vec): New function.
19509 (free_stmt_vec_infos): Likewise.
19510 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
19511 (set_vinfo_for_stmt): Likewise.
19512 (get_earlier_stmt): Likewise.
19513 (get_later_stmt): Likewise.
19514 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
19515 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
19516 (vec_info::~vec_info): Free stmt_vec_infos.
19517 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
19518 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
19519 (pass_slp_vectorize::execute): Likewise.
19520 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
19521 (free_stmt_vec_info_vec): Likewise.
19522 (set_stmt_vec_info_vec): New function.
19523 (free_stmt_vec_infos): Likewise.
19524 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
19525 the global stmt_vec_info_vec.
19526 * tree-parloops.c (gather_scalar_reductions): Use
19527 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
19530 2018-05-29 Richard Biener <rguenther@suse.de>
19532 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
19534 2018-05-29 Martin Liska <mliska@suse.cz>
19535 David Malcolm <dmalcolm@redhat.com>
19537 * vec.c (test_reverse): New.
19538 (vec_c_tests): Add new test.
19539 * vec.h (vl_ptr>::reverse): New function.
19541 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
19543 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
19545 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
19548 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
19550 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
19552 2018-05-28 Richard Biener <rguenther@suse.de>
19554 PR tree-optimization/85933
19555 * tree-vect-data-refs.c (vect_record_base_alignments): Only
19556 look at stmts marked as vectorizable.
19558 2018-05-28 Richard Biener <rguenther@suse.de>
19560 PR tree-optimization/85934
19561 * tree-vect-generic.c (expand_vector_operations_1): Hoist
19562 vector boolean check before scalar optimization.
19564 2018-05-28 Jakub Jelinek <jakub@redhat.com>
19566 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
19569 2018-05-28 Mark Wielaard <mark@klomp.org>
19571 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
19572 if it is an expression containing a minus sign.
19574 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
19576 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
19578 2018-05-27 Paul Koning <ni1d@arrl.net>
19580 * config/pdp11/pdp11.md (truncsihi2): Remove.
19582 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
19583 Chung-Ju Wu <jasonwucj@gmail.com>
19585 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
19587 (unaligned_store_dw): Ditto.
19588 * config/nds32/nds32-memory-manipulation.c
19589 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
19590 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
19591 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
19592 (emit_setmem_word_loop): Rename to ...
19593 (emit_setmem_doubleword_loop): ... this.
19594 (nds32_gen_dup_4_byte_to_word_value): New function.
19595 (nds32_gen_dup_8_byte_to_double_word_value): New function.
19596 (nds32_expand_setmem_loop): Refine implementation.
19597 (nds32_expand_setmem_loop_v3m): Ditto.
19598 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
19601 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
19603 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
19605 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
19607 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
19608 (nds32_init_machine_status): Initialize machine->attr_naked_p and
19609 machine->attr_no_prologue_p.
19610 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
19611 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
19612 (nds32_expand_epilogue): Consider attr_naked_p.
19613 (nds32_expand_epilogue_v3pop): Likewise.
19614 (nds32_can_use_return_insn): Likewise.
19615 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
19616 attr_no_prologue_p fields.
19617 * config/nds32/nds32.opt (mret-in-naked-func): New option.
19619 2018-05-27 Jakub Jelinek <jakub@redhat.com>
19622 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
19624 * config/i386/sse.md
19625 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
19627 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
19629 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
19631 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
19633 (*<floatsuffix>floatv2div2sf2): Rename to ...
19634 (*float<floatunssuffix>v2div2sf2): ... this.
19635 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
19636 (float<floatunssuffix>v2div2sf2_mask): ... this.
19637 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
19638 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
19639 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
19641 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
19643 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
19645 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
19647 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
19649 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
19651 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
19652 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
19653 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
19654 gen_ufix_truncv8dfv8si2.
19655 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
19656 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
19657 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
19658 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
19659 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
19660 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
19661 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
19662 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
19664 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
19668 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
19670 2018-05-25 Jim Wilson <jimw@sifive.com>
19672 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
19673 * config/riscv/riscv.c (struct machine_function): Add
19674 interrupt_handler_p and attribute_checked_p fields.
19675 (riscv_attribute_table): Add interrupt.
19676 (riscv_interrupt_type_p): New.
19677 (riscv_save_reg_p): Save extra regs for interrupt handler.
19678 (riscv_use_save_libcall): Return false for interrupt handler.
19679 (riscv_first_stack_step): Add forward declaration.
19680 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
19681 for interrupt handler with large frame. Use it for saved reg list.
19682 (riscv_expand_prologue): Move flag_stack_usage_info support to
19683 eliminate duplication.
19684 (riscv_expand_epilogue): Generate mret for interrupt handler.
19685 (riscv_epilogue_uses): New.
19686 (riscv_can_use_return_insn): Return false for interrupt handler.
19687 (riscv_function_ok_for_sibcall): Likewise.
19688 (riscv_set_current_function): Add interrupt handler support.
19689 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
19690 * config/riscv/riscv.md (UNSPECV_MRET): New.
19692 (riscv_frflags, riscv_fsflags): Use tab after opcode.
19694 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
19696 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
19698 PR tree-optimization/85712
19699 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
19700 this candidate has already been replaced in-situ by a copy.
19702 2018-05-25 Jason Merrill <jason@redhat.com>
19704 PR c++/80485 - inline function non-zero address.
19705 * symtab.c (nonzero_address): Check DECL_COMDAT.
19707 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
19710 * config/alpha/alpha.md (ashlsi3): New insn pattern.
19711 (*ashlsi_se): Rename from *ashldi_se. Define as sign
19712 extension of SImode operation. Use const123_operand predicate.
19713 (*saddsi_1): Remove.
19714 (*saddl_se_1): Ditto.
19715 (*ssubsi_1): Ditto.
19716 (*ssubl_se_1): Ditto.
19717 * config/alpha/predicates.md (const123_operand): New predicate.
19718 * config/alpha/constraints.md (P): Use IN_RANGE.
19720 2018-05-25 Richard Biener <rguenther@suse.de>
19722 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
19724 (ref_maybe_used_by_stmt_p): Likewise.
19725 (stmt_may_clobber_ref_p): Likewise.
19726 (stmt_may_clobber_ref_p_1): Likewise.
19727 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
19729 (ref_maybe_used_by_stmt_p): Likewise.
19730 (stmt_may_clobber_ref_p): Likewise.
19731 (stmt_may_clobber_ref_p_1): Likewise.
19732 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
19733 the alias oracle to disambiguate DRs with stmts DR analysis
19735 (vect_analyze_data_refs): Do not give up on not analyzable
19736 DRs for BB vectorization. Remove code truncating the dataref
19739 2018-05-25 Jakub Jelinek <jakub@redhat.com>
19742 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
19743 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
19744 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
19746 2018-05-25 Richard Biener <rguenther@suse.de>
19748 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
19749 function, combining stmt data ref gathering and fatal analysis
19751 (vect_analyze_data_refs): Remove now redudnant code and simplify.
19752 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
19753 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
19754 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
19755 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
19757 2018-05-25 Bin Cheng <bin.cheng@arm.com>
19759 PR tree-optimization/85720
19760 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
19761 SCC if all partitions are builtins.
19762 (version_loop_by_alias_check): New parameter. Generate cancelable
19763 runtime alias check if all partitions are builtins.
19764 (distribute_loop): Update call to above function.
19766 2018-05-25 Bin Cheng <bin.cheng@arm.com>
19768 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
19769 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
19770 (parm_default_def_partition_arg): Ditto.
19771 (set_parm_default_def_partition): Ditto.
19772 (get_parm_default_def_partitions): Ditto and make it static.
19773 (get_undefined_value_partitions): Ditto and make it static.
19774 (remove_ssa_form): Refactor call to init_var_map here.
19775 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
19776 computation for loop region.
19777 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
19778 (register_default_def): Delete.
19779 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
19780 (parm_default_def_partition_arg): Ditto.
19781 (set_parm_default_def_partition): Ditto.
19782 (get_parm_default_def_partitions): Ditto and make it static.
19783 (get_undefined_value_partitions): Ditto and make it static.
19784 (coalesce_with_default, coalesce_with_default): Update comment.
19785 (create_coalesce_list_for_region): New func factored out from
19786 create_outofssa_var_map.
19787 (populate_coalesce_list_for_outofssa): New func factored out from
19788 create_outofssa_var_map and coalesce_ssa_name.
19789 (create_outofssa_var_map): Delete.
19790 (coalesce_ssa_name): Refactor to support live range computation.
19791 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
19792 (get_parm_default_def_partitions): Delete.
19793 (get_undefined_value_partitions): Ditto.
19794 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
19795 computation for loop region.
19796 (new_tree_live_info, loe_visit_block): Ditto.
19797 (live_worklist, set_var_live_on_entry): Ditto.
19798 (calculate_live_on_exit, verify_live_on_entry): Ditto.
19799 * tree-ssa-live.h (struct _var_map): New fields.
19800 (init_var_map): Change decl.
19801 (region_contains_p): New.
19803 2018-05-25 Bin Cheng <bin.cheng@arm.com>
19805 * tree-ssa-live.h (live_merge_and_clear): Delete.
19807 2018-05-25 Richard Biener <rguenther@suse.de>
19810 * tree-dump.c (dequeue_and_dump): Remove access to removed
19811 operand 2 of a SWITCH_EXPR.
19813 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
19815 * doc/sourcebuild.texi (vect_double_cond_arith): Include
19816 multiplication and division.
19817 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
19818 (cond_udiv@var{m}, cond_umod@var{m}): Document.
19819 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
19820 (cond_udiv_optab, cond_umod_optab): New optabs.
19821 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
19822 (IFN_COND_RDIV): New internal functions.
19823 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
19824 TRUNC_MOD_EXPR and RDIV_EXPR.
19825 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
19826 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
19828 (SVE_INT_BINARY): Include mult.
19829 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
19830 (optab, sve_int_op): Handle mult.
19831 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
19833 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
19834 for SVE_INT_BINARY_SD.
19836 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
19838 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
19839 (optab, sve_int_op): Handle div and udiv.
19840 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
19841 for SVE_INT_BINARY_SD.
19842 (*<optab><mode>3): New insn for the same.
19844 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
19846 * tree-vect-patterns.c: Include predict.h.
19847 (vect_recog_divmod_pattern): Restrict check for division support
19848 to when optimizing for size.
19850 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
19852 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
19853 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
19854 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
19855 (gimple_match_op::set_op): Likewise.
19856 (gimple_resimplify4): Declare.
19857 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
19858 (expr::gen_transform): Likewise.
19859 (decision_tree::gen): Generate a simplification routine for 4 operands.
19860 * gimple-match-head.c (gimple_simplify): Add an overload for
19861 4 operands. In the top-level function, handle up to 4 call
19862 arguments and call gimple_resimplify4.
19863 (gimple_resimplify4): New function.
19864 (build_call_internal): Pass a fourth operand.
19865 (maybe_push_to_seq): Likewise.
19866 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
19867 Fold VEC_COND_EXPRs of an operation and a default value into
19868 an IFN_COND_* function if possible.
19869 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
19871 (SVE_COND_FP_BINARY): Include them.
19872 (optab, sve_fp_op): Handle them.
19873 (SVE_INT_BINARY_REV): New code iterator.
19874 (SVE_COND_FP_BINARY_REV): New int iterator.
19875 (commutative): New int attribute.
19876 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
19878 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
19880 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
19881 (*cond_<optab><mode>): New patterns for reversed operands.
19883 2018-05-25 Richard Biener <rguenther@suse.de>
19885 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
19886 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
19887 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
19888 (STMT_VINFO_GROUPED_ACCESS): Adjust.
19889 * tree-vect-data-refs.c (everywhere): Adjust users.
19890 * tree-vect-loop.c (everywhere): Likewise.
19891 * tree-vect-slp.c (everywhere): Likewise.
19892 * tree-vect-stmts.c (everywhere): Likewise.
19893 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
19895 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19897 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
19899 (gcc_cv_as_section_exclude): ... this.
19900 Try Solaris as #exclude syntax.
19901 * configure: Regenerate.
19902 * config.in: Regenerate.
19903 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
19905 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
19906 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
19908 * varasm.c (default_elf_asm_named_section): Don't check if
19909 HAVE_GAS_SECTION_EXCLUDE is defined.
19911 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
19913 * doc/md.texi: Update the documentation of the cond_* optabs
19914 to mention the new final operand. Fix GET_MODE_NUNITS call.
19915 Describe the scalar case too.
19916 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
19917 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
19919 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
19920 (get_conditional_internal_fn): Update comment.
19921 * tree-vect-loop.c (vectorizable_reduction): Pass the original
19922 accumulator value as a final argument to conditional functions.
19923 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
19924 a define_expand and add an "else" operand. Assert for now that
19925 the else operand is equal to operand 2. Use SVE_INT_BINARY and
19926 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
19927 (*cond_<optab><mode>): New patterns.
19928 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
19929 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
19930 (UNSPEC_COND_EOR): Delete.
19931 (optab): Remove associated mappings.
19932 (SVE_INT_BINARY): New code iterator.
19933 (sve_int_op): Remove int attribute and add "minus" to the code
19935 (SVE_COND_INT_OP): Delete.
19936 (SVE_COND_FP_OP): Rename to...
19937 (SVE_COND_FP_BINARY): ...this.
19939 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
19941 * optabs.c (can_reuse_operands_p): New function.
19942 (maybe_legitimize_operands): Try to reuse the results for
19945 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
19947 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
19948 Add {q} suffix to insn mnemonic.
19950 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
19952 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
19953 (msp430_warn_func_return): New.
19955 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
19957 * fold-const.c (tree_nonzero_bits): New function.
19958 * fold-const.h (tree_nonzero_bits): Likewise.
19959 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
19960 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
19962 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
19966 * varasm.c (assemble_alias): Check ifunc_resolver only on
19969 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
19972 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
19973 when memory input operand is handled.
19975 2018-05-24 Luis Machado <luis.machado@linaro.org>
19977 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
19979 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
19981 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
19983 * match.pd: Delay FMA folds until after vectorization.
19985 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
19988 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
19989 address check not strict.
19991 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
19993 * gimple-match.h (gimple_match_op): New class.
19994 (mprts_hook): Replace parameters with a gimple_match_op *.
19995 (maybe_build_generic_op): Likewise.
19996 (gimple_simplified_result_is_gimple_val): Replace parameters with
19997 a const gimple_match_op *.
19998 (gimple_simplify): Replace code_helper * and tree * parameters with
19999 a gimple_match_op * parameter.
20000 (gimple_resimplify1): Replace code_helper *, tree and tree *
20001 parameters with a gimple_match_op * parameter.
20002 (gimple_resimplify2): Likewise.
20003 (gimple_resimplify3): Likewise.
20004 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
20005 parameters with a gimple_match_op * parameter.
20006 * gimple-match-head.c (gimple_simplify): Change prototypes of
20007 auto-generated functions to take a gimple_match_op * instead of
20008 separate code_helper * and tree * parameters. Make the same
20009 change in the top-level overload and update calls to the
20010 gimple_resimplify routines. Update calls to the auto-generated
20011 functions and to maybe_push_res_to_seq in the publicly-facing
20012 operation-specific gimple_simplify overloads.
20013 (gimple_match_op::MAX_NUM_OPS): Define.
20014 (gimple_resimplify1): Replace rcode and ops with a single res_op
20015 parameter. Update call to gimple_simplify.
20016 (gimple_resimplify2): Likewise.
20017 (gimple_resimplify3): Likewise.
20018 (mprts_hook): Replace parameters with a gimple_match_op *.
20019 (maybe_build_generic_op): Likewise.
20020 (build_call_internal): Replace type, nargs and ops with
20021 a gimple_match_op *.
20022 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
20023 with a single gimple_match_op *. Update calls to mprts_hook,
20024 build_call_internal and gimple_simplified_result_is_gimple_val.
20025 Factor out code that is common to the tree_code and combined_fn cases.
20026 * genmatch.c (expr::gen_transform): Replace tem_code and
20027 tem_ops with a gimple_match_op called tem_op. Update calls
20028 to the gimple_resimplify functions and maybe_push_res_to_seq.
20029 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
20030 res_ops. Update call to the gimple_resimplify functions.
20031 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
20032 (decision_tree::gen): Make the functions take a gimple_match_op *
20033 called res_op instead of separate res_code and res_ops parameters.
20034 Update call accordingly.
20035 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
20036 and ops with a single res_op parameter. Update calls to
20037 maybe_build_generic_op and maybe_push_res_to_seq.
20038 (fold_stmt_1): Update calls to gimple_simplify and
20039 replace_stmt_with_simplification.
20040 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
20041 and gimple_simplified_result_is_gimple_val.
20042 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
20044 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
20045 with a gimple_match_op *.
20046 (vn_nary_build_or_lookup): Likewise. Update call to
20047 vn_nary_build_or_lookup_1.
20048 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
20049 gimple_match_op *. Update calls to the gimple_resimplify routines
20050 and to gimple_simplified_result_is_gimple_val.
20051 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
20052 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
20053 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
20054 (visit_nary_op): Likewise.
20055 (visit_reference_op_load): Likewise.
20057 2018-05-23 Luis Machado <luis.machado@linaro.org>
20059 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
20060 modifier for printing the step amount.
20062 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
20065 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
20068 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
20070 * doc/sourcebuild.texi (Endianness): New subsubsection.
20072 2018-05-23 Luis Machado <luis.machado@linaro.org>
20074 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20075 <prefetch_dynamic_strides>: New const bool field.
20076 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20077 prefetch_dynamic_strides.
20078 (exynosm1_prefetch_tune): Likewise.
20079 (thunderxt88_prefetch_tune): Likewise.
20080 (thunderx_prefetch_tune): Likewise.
20081 (thunderx2t99_prefetch_tune): Likewise.
20082 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
20084 (aarch64_override_options_internal): Update to set
20085 PARAM_PREFETCH_DYNAMIC_STRIDES.
20086 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20087 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20088 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20089 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20090 prefetch-dynamic-strides setting.
20092 2018-05-23 Luis Machado <luis.machado@linaro.org>
20094 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20095 <minimum_stride>: New const int field.
20096 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20097 minimum_stride field defaulting to -1.
20098 (exynosm1_prefetch_tune): Likewise.
20099 (thunderxt88_prefetch_tune): Likewise.
20100 (thunderx_prefetch_tune): Likewise.
20101 (thunderx2t99_prefetch_tune): Likewise.
20102 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
20103 <default_opt_level>: Set to 3.
20104 (aarch64_override_options_internal): Update to set
20105 PARAM_PREFETCH_MINIMUM_STRIDE.
20106 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20107 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20108 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20109 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20110 stride is constant and is below the minimum stride threshold.
20112 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20114 * config/arm/arm-cpus.in (mode26): Delete.
20115 (armv4): Delete mode26 reference.
20116 * config/arm/arm.c (arm_configure_build_target): Delete use of
20119 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
20121 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
20123 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
20124 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
20126 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
20127 (floatunsdidf2): Ditto.
20129 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
20131 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
20132 (fixuns_trunc<mode>si2_avx512f): Ditto.
20133 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
20134 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
20135 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
20137 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
20139 PR rtl-optimization/79985
20140 * df-scan.c (df_insn_refs_collect): Remove special case for
20141 global registers and asm statements.
20143 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
20145 * extend.texi (Global Register Variables): Rewrite the bullet list.
20146 Note that the register is available for allocation. Note that access
20147 via inline asm must use constraints. Add note about async-signal
20148 handlers. Remove paragraph about automagic register selection.
20150 2018-05-23 Richard Biener <rguenther@suse.de>
20152 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
20153 of fixed offset from memset VN.
20155 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
20157 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
20158 first_interp field.
20159 (alloc_cand_and_find_basis): Initialize first_interp field.
20160 (slsr_process_mul): Modify first_interp field.
20161 (slsr_process_add): Likewise.
20162 (slsr_process_cast): Modify first_interp field for each new
20164 (slsr_process_copy): Likewise.
20165 (dump_candidate): Dump first_interp field.
20166 (replace_mult_candidate): Process all interpretations, not just
20168 (replace_rhs_if_not_dup): Likewise.
20169 (replace_one_candidate): Likewise.
20171 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
20173 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
20175 (aarch64_needs_frame_chain): New function.
20176 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
20178 2018-05-23 Sudakshina Das <sudi.das@arm.com>
20181 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
20182 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
20183 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
20184 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
20185 as true for strict-align.
20186 (aarch64_can_inline_p): Perform checks even when callee has no
20187 attributes to check for strict alignment.
20188 * doc/extend.texi (AArch64 Function Attributes): Document
20190 * doc/invoke.texi: (AArch64 Options): Likewise.
20192 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
20194 PR tree-optimization/85853
20195 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
20196 the handling of the root of the node to...
20197 (vect_slp_analyze_node_operations_1): ...this new function,
20198 and run the whole thing with the child nodes' def types
20199 set according to their SLP node's def type.
20201 2018-05-23 Richard Biener <rguenther@suse.de>
20203 PR middle-end/85874
20204 * tree-data-ref.c (create_runtime_alias_checks): Defer
20205 and ignore overflow warnings.
20207 2018-05-23 Yury Gribov <tetra2005@gmail.com>
20209 PR tree-optimization/85822
20210 * tree-vrp.c (is_masked_range_test): Fix handling of negative
20213 2018-05-23 Richard Biener <rguenther@suse.de>
20215 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
20216 memset constants via native_interpret_expr.
20218 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
20221 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
20223 (cgraph_node::create_alias): Likewise.
20224 (cgraph_node::get_availability): Check ifunc_resolver instead
20225 of looking up ifunc attribute.
20226 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
20227 * varasm.c (do_assemble_alias): Likewise.
20228 (assemble_alias): Likewise.
20229 (default_binds_local_p_3): Likewise.
20230 * cgraph.h (cgraph_node): Add ifunc_resolver.
20231 (cgraph_node::only_called_directly_or_aliased_p): Return false
20232 for IFUNC resolver.
20233 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
20235 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
20236 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
20237 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
20238 instead of looking up ifunc attribute.
20240 2018-05-22 Luis Machado <luis.machado@linaro.org>
20242 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
20244 2018-05-22 Martin Sebor <msebor@redhat.com>
20246 PR middle-end/85359
20247 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
20248 only when expasion succeeds.
20249 (expand_builtin_strcmp): Same.
20250 (expand_builtin_strncmp): Same.
20252 2018-05-22 Martin Sebor <msebor@redhat.com>
20254 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
20256 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
20257 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20259 * config/aarch64/aarch64-ldpstp.md: Replace uses of
20260 aarch64_mem_pair_operand with memory_operand and delete operand swapping
20262 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
20263 Add check for legitimate_address.
20264 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
20265 (aarch64_swap_ldrstr_operands): New.
20266 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
20269 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
20270 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20272 * config/aarch64/aarch64.md: New patterns to generate stp
20274 (store_pair_sw, store_pair_dw): New patterns to generate stp for
20275 single words and double words.
20276 (load_pair_sw, load_pair_dw): Likewise.
20277 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
20279 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
20281 * config/aarch64/aarch64-ldpstp.md: Modify peephole
20282 for different mode ldpstp and add peephole for merged zero stores.
20283 Likewise for loads.
20284 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
20286 (aarch64_gen_store_pair): Rename calls to match new patterns.
20287 (aarch64_gen_load_pair): Rename calls to match new patterns.
20288 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
20289 (load_pair<DREG:mode><DREG2:mode>): ... This.
20290 (store_pair<mode>): Rename to...
20291 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
20292 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
20293 New mode iterators.
20294 (V_INT_EQUIV): Handle SImode.
20295 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
20298 2018-05-22 Martin Sebor <msebor@redhat.com>
20301 * calls.c (maybe_warn_nonstring_arg): Use string length to set
20302 or ajust the presumed bound on an operation to avoid unnecessary
20305 2018-05-22 Martin Sebor <msebor@redhat.com>
20307 PR tree-optimization/85826
20308 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
20309 assuming that a DECL necesarily has a constant size.
20311 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
20313 PR middle-end/85862
20314 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
20316 2018-05-22 Richard Biener <rguenther@suse.de>
20318 PR tree-optimization/85834
20319 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
20320 non-constant and non-zero memset arguments.
20322 2018-05-22 Martin Liska <mliska@suse.cz>
20325 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
20327 2018-05-22 Richard Biener <rguenther@suse.de>
20329 PR tree-optimization/85863
20330 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
20331 comparisons when vectype is specified.
20332 (vectorizable_condition): Do not specify vectype for
20333 vect_is_simple_cond when SLP vectorizing.
20335 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
20338 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
20339 define __ibm128 as long double.
20340 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
20341 as a distinct type when IEEE 128-bit support is enabled.
20342 (init_float128_ieee): Fix up conversions between IFmode and IEEE
20343 128-bit types to use the correct functions.
20344 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
20345 convert between 128-bit floating point types that have different
20346 modes but the same representation, instead of using gen_lowpart to
20348 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
20350 (IFKF_reg): New attributes to give the register constraints for
20352 (extend<mode>tf2_internal): New insns to mark an explicit
20353 conversion between 128-bit floating point types that have a
20354 different mode but share the same representation.
20356 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
20358 PR tree-optimization/85814
20359 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
20360 a null return from get_strinfo when unsharing the next
20361 strinfo in the chain.
20363 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
20366 * varasm.c (weak_finish): Clean up weak_decls.
20368 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20370 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
20371 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
20372 UNSPEC_UADALP values.
20373 * config/aarch64/iterators.md (ABAL): New int iterator.
20376 (sur): Add mappings for the above.
20377 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
20379 (aarch64_<sur>abal<mode>_4): Likewise.
20380 (aarch64_<sur>adalp<mode>_3): Likewise.
20381 (<sur>sadv16qi): New define_expand.
20383 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
20385 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
20386 (*movdf_internal): Ditto.
20387 (*rcpsf2_sse): Ditto.
20388 (*rsqrtsf2_sse): Ditto.
20389 (*sqrt<mode>2_sse): Ditto.
20391 2018-05-21 Tamar Christina <tamar.christina@arm.com>
20393 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
20395 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
20396 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
20397 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
20398 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
20400 * config/aarch64/arm_neon.h: Likewise.
20401 * config/aarch64/iterators.md (VQ_I): New.
20403 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
20405 * config.gcc: Add arc/t-multilib-linux to tmake_file for
20407 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
20410 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
20412 * config/nds32/constraints.md (S): New constraint.
20413 * config/nds32/nds32.md (call_internal): Use constraint S.
20414 (call_value_internal): Likewise.
20415 (sibcall_internal): Likewise.
20416 (sibcall_value_internal): Likewise.
20418 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
20419 Chung-Ju Wu <jasonwucj@gmail.com>
20421 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
20422 into consideration.
20424 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
20425 Chung-Ju Wu <jasonwucj@gmail.com>
20427 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
20428 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
20429 (nds32_rtx_costs_impl): Simplify.
20430 (nds32_address_cost_impl): Simplify.
20431 (nds32_init_rtx_costs): New function.
20432 (nds32_rtx_costs_speed_prefer): Likewise.
20433 (nds32_rtx_costs_size_prefer): Likewise.
20434 (nds32_address_cost_speed_prefer): Likewise.
20435 (nds32_address_cost_speed_fwprop): Likewise.
20436 (nds32_address_cost_size_prefer): Likewise.
20437 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
20438 * config/nds32/nds32.c (nds32_option_override): Use
20439 nds32_init_rtx_costs function.
20441 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
20443 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
20444 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
20445 (TARGET_PIPELINE_N8): Likewise.
20446 (TARGET_PIPELINE_N10): Likewise.
20447 (TARGET_PIPELINE_N13): Likewise.
20448 (TARGET_PIPELINE_GRAYWOLF): Likewise.
20450 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
20452 * config/nds32/nds32-fpu.md: Update copyright year.
20454 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
20456 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
20458 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
20460 * config/nds32/nds32.c
20461 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
20462 * config/nds32/nds32.opt (minline-asm-r15): New option.
20464 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
20466 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
20468 * config/nds32/nds32.md (abssi2): New pattern.
20470 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
20472 * config/i386/i386.md (rex64namesuffix): New mode attribute.
20473 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
20474 Merge insn pattern from sse_cvtsi2ss<round_name> and
20475 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
20476 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
20477 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
20478 using SWI48 mode iterator.
20479 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
20480 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
20481 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
20482 pattern from sse_cvttss2si<round_saeonly_name>
20483 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
20484 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
20485 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
20486 using SWI48 mode iterator.
20487 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
20488 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
20489 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
20490 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
20491 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
20492 using SWI48 mode iterator.
20493 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
20494 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
20495 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
20496 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
20497 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
20498 SWI48 mode iterator.
20499 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
20500 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
20501 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
20502 pattern from sse_cvttsd2si<round_saeonly_name>
20503 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
20505 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
20507 * config/nds32/nds32-md-auxiliary.c
20508 (nds32_valid_smw_lwm_base_p): Refine.
20509 (nds32_output_smw_single_word): Refine.
20510 (nds32_output_smw_double_word): New.
20511 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
20513 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
20515 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
20516 (nds32_output_stack_pop): Refine.
20517 (nds32_expand_unaligned_load): Refine.
20518 (nds32_expand_unaligned_store): Refine.
20520 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
20521 Chung-Ju Wu <jasonwucj@gmail.com>
20523 * config/nds32/constants.md: Add TP_REGNUM constant.
20524 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
20525 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
20527 * config/nds32/nds32-doubleword.md: Consider flag_pic.
20528 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
20529 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
20530 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
20531 and PIC code generation.
20532 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
20534 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
20536 * config/nds32/nds32.md: Support TLS and PIC.
20537 * config/nds32/nds32.c: Support TLS and PIC.
20538 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
20539 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
20542 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
20544 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
20545 mode with E_ prefix.
20547 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
20548 Chung-Ju Wu <jasonwucj@gmail.com>
20550 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
20551 * config/nds32/nds32-md-auxiliary.c
20552 (symbolic_reference_mentioned_p): New.
20553 (nds32_legitimize_ict_address): New.
20554 (nds32_expand_ict_move): New.
20555 (nds32_indirect_call_referenced_p): New.
20556 (nds32_symbol_binds_local_p): Delete.
20557 (nds32_long_call_p): Modify.
20558 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
20559 * config/nds32/nds32-protos.h
20560 (symbolic_reference_mentioned_p): Declare.
20561 (nds32_legitimize_ict_address): Declare.
20562 (nds32_expand_ict_move): Declare.
20563 (nds32_indirect_call_referenced_p): Declare.
20564 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
20565 (nds32_relax_group): Use nds32_ict_const_p as condition.
20566 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
20567 (nds32_asm_file_start): Output ict_model directive in asm code.
20568 (nds32_legitimate_address_p): Consider indirect call.
20569 (nds32_print_operand): Consider indirect call.
20570 (nds32_print_operand_address): Consider indirect call.
20571 (nds32_insert_attributes): Handle "indirect_call" attribute.
20572 (TARGET_LEGITIMATE_ADDRESS_P): Define.
20573 (TARGET_LEGITIMATE_CONSTANT_P): Define.
20574 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
20575 (TARGET_DELEGITIMIZE_ADDRESS): Define.
20576 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
20577 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
20578 (TARGET_ICT_MODEL_SMALL): Define.
20579 (TARGET_ICT_MODEL_LARGE): Define.
20580 * config/nds32/nds32.md (movsi): Consider ict model.
20581 (call, call_value): Consider ict model.
20582 (sibcall, sibcall_value): Consider ict model.
20583 * config/nds32/nds32.opt (mict-model): New option.
20584 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
20587 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
20588 Monk Chiang <sh.chiang04@gmail.com>
20589 Jim Wilson <jimw@sifive.com>
20591 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
20592 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
20593 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
20594 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
20595 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
20596 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
20597 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
20598 compute save_libcall_adjustment properly.
20599 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
20600 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
20601 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
20602 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
20603 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
20604 (ABI_SPEC): Handle mabi=ilp32e.
20605 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
20606 (RVE): Add RVE mask.
20607 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
20608 <-march>: Add rv32e as an example.
20610 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
20613 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
20614 (intra_create_variable_infos): Handle C++ constructors.
20616 2018-05-18 Martin Liska <mliska@suse.cz>
20618 * passes.def: Remove a redundant pass.
20620 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
20623 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
20625 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20627 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
20629 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
20631 (armv2, armv2a, armv3, armv3m): Delete architectures.
20632 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
20633 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
20634 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
20636 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
20637 (*mulsidi3adddi): Likewise.
20638 (mulsidi3): Likewise.
20639 (*mulsidi3_nov6): Likewise.
20640 (umulsidi3): Likewise.
20641 (umulsidi3_nov6): Likewise.
20642 (umaddsidi4): Likewise.
20643 (*umulsidi3adddi): Likewise.
20644 (smulsi3_highpart): Likewise.
20645 (*smulsi3_highpart_nov6): Likewise.
20646 (umulsi3_highpart): Likewise.
20647 (*umulsi3_highpart_nov6): Likewise.
20648 * config/arm/arm.h (arm_arch3m): Delete.
20649 * config/arm/arm.c (arm_arch3m): Delete.
20650 (arm_option_override_internal): Update armv3-related comment.
20651 (arm_configure_build_target): Delete use of isa_bit_mode32.
20652 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
20653 (arm_rtx_costs_internal): Delete check of arm_arch3m.
20654 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
20655 (mulsa3): Likewise.
20656 (mulusa3): Likewise.
20657 * config/arm/arm-protos.h (arm_arch3m): Delete.
20658 * config/arm/arm-tables.opt: Regenerate.
20659 * config/arm/arm-tune.md: Likewise.
20660 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
20661 deleted architectures.
20663 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20665 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
20666 (armv5t, armv5te): New features.
20667 (ARMv5, ARMv5e): Delete fgroups.
20668 (ARMv5t, ARMv5te): Adjust for above changes.
20669 (ARMv6m): Likewise.
20670 (armv5, armv5e): Delete arches.
20671 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
20673 (*call_reg_arm): Likewise.
20674 (*call_value_reg_armv5): Likewise.
20675 (*call_value_reg_arm): Likewise.
20676 (*call_symbol): Likewise.
20677 (*call_value_symbol): Likewise.
20678 (*sibcall_insn): Likewise.
20679 (*sibcall_value_insn): Likewise.
20680 (clzsi2): Likewise.
20681 (prefetch): Likewise.
20682 (define_split and define_peephole2 dependent on arm_arch5):
20684 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
20686 (TARGET_ARM_QBIT): Likewise.
20687 (TARGET_DSP_MULTIPLY): Likewise.
20688 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
20689 (arm_arch5, arm_arch5e): Delete.
20690 (arm_arch5t, arm_arch5te): Declare.
20691 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
20692 (arm_arch5t): Declare.
20693 (arm_option_reconfigure_globals): Update for the above.
20694 (arm_options_perform_arch_sanity_checks): Update comment, replace
20695 use of arm_arch5 with arm_arch5t.
20696 (use_return_insn): Likewise.
20697 (arm_emit_call_insn): Likewise.
20698 (output_return_instruction): Likewise.
20699 (arm_final_prescan_insn): Likewise.
20700 (arm_coproc_builtin_available): Likewise.
20701 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
20702 arm_arch5e with arm_arch5t and arm_arch5te.
20703 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
20704 (arm_arch5t, arm_arch5te): Declare.
20705 * config/arm/arm-tables.opt: Regenerate.
20706 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
20707 * config/arm/t-multilib: Likewise.
20708 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
20709 instead of arm_arch5.
20710 (*call_reg_thumb1): Likewise.
20711 (*call_value_reg_thumb1_v5): Likewise.
20712 (*call_value_reg_thumb1): Likewise.
20713 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
20715 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
20717 2018-05-18 Martin Liska <mliska@suse.cz>
20719 PR gcov-profile/84846
20720 * doc/gcov.texi: Document -t option of gcov tool.
20722 2018-05-18 Martin Liska <mliska@suse.cz>
20724 PR gcov-profile/84846
20725 * gcov.c (print_usage): Add new -t option.
20726 (process_args): Handle the option.
20727 (generate_results): Use stdout as output when requested by
20730 2018-05-18 Martin Liska <mliska@suse.cz>
20732 PR gcov-profile/84846
20733 * coverage.c (coverage_init): Write PWD to .gcno file.
20734 * doc/gcov.texi: Document how working directory is printed.
20735 * gcov-dump.c (dump_gcov_file): Print PWD.
20736 * gcov.c (output_intermediate_file): Likewise.
20737 (read_graph_file): Read PWD string.
20738 (output_lines): Print PWD.
20740 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20742 PR middle-end/85817
20743 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
20744 for retval and return false if all args to phi are zero.
20746 2018-05-18 Richard Biener <rguenther@suse.de>
20748 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
20750 (evrp_dom_walker::before_dom_children): Call it.
20752 2018-05-18 Richard Biener <rguenther@suse.de>
20754 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
20755 results when processing array refs with variable index.
20757 2018-05-18 Toon Moene <toon@moene.org>
20759 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
20760 directly after that of -floop-interchange. Indicate that both
20761 options are enabled by default when specifying -O3.
20763 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20765 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
20766 iterator. Delete separate integer-mode vec_set<mode> expander.
20767 (aarch64_simd_vec_setv2di): Delete.
20768 (vec_setv2di): Delete.
20769 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
20770 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
20771 the "w, r" alternative.
20773 2018-05-18 Martin Liska <mliska@suse.cz>
20775 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
20776 * tree-pass.h (make_pass_lower_switch_O0): New function.
20777 * tree-switch-conversion.c (node_has_low_bound): Remove.
20778 (node_has_high_bound): Likewise.
20779 (node_is_bounded): Likewise.
20780 (class pass_lower_switch): Make it a template type and create
20782 (pass_lower_switch::execute): Add template argument.
20783 (make_pass_lower_switch): New function.
20784 (make_pass_lower_switch_O0): New function.
20785 (do_jump_if_equal): Remove.
20786 (emit_case_nodes): Simplify to just handle all 3 cases and leave
20787 all the hard work to tree optimization passes.
20789 2018-05-18 Martin Liska <mliska@suse.cz>
20791 * dbgcnt.c (limit_low): Renamed from limit.
20792 (limit_high): New variable.
20793 (dbg_cnt_is_enabled): Check for upper limit.
20794 (dbg_cnt): Adjust dumping.
20795 (dbg_cnt_set_limit_by_index): Add new argument for high
20797 (dbg_cnt_set_limit_by_name): Likewise.
20798 (dbg_cnt_process_single_pair): Parse new format.
20799 (dbg_cnt_process_opt): Use strtok.
20800 (dbg_cnt_list_all_counters): Remove 'value' and add
20802 * doc/invoke.texi: Document changes.
20804 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
20806 * doc/sourcebuild.texi (scalar_all_fma): Document.
20807 * tree.def (FMA_EXPR): Delete.
20808 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
20809 * internal-fn.c (ternary_direct): New macro.
20810 (expand_ternary_optab_fn): Likewise.
20811 (direct_ternary_optab_supported_p): Likewise.
20812 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
20813 * builtins.c (fold_builtin_fma): Delete.
20814 (fold_builtin_3): Don't call it.
20815 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
20816 * expr.c (expand_expr_real_2): Likewise.
20817 * fold-const.c (operand_equal_p): Likewise.
20818 (fold_ternary_loc): Likewise.
20819 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
20820 * gimple.c (DEFTREECODE): Likewise.
20821 * gimplify.c (gimplify_expr): Likewise.
20822 * optabs-tree.c (optab_for_tree_code): Likewise.
20823 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
20824 * tree-eh.c (operation_could_trap_p): Likewise.
20825 (stmt_could_throw_1_p): Likewise.
20826 * tree-inline.c (estimate_operator_cost): Likewise.
20827 * tree-pretty-print.c (dump_generic_node): Likewise.
20828 (op_code_prio): Likewise.
20829 * tree-ssa-loop-im.c (stmt_cost): Likewise.
20830 * tree-ssa-operands.c (get_expr_operands): Likewise.
20831 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
20832 * fold-const-call.h (fold_fma): Delete.
20833 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
20834 CFN_FNMA and CFN_FNMS.
20835 (fold_fma): Delete.
20836 * genmatch.c (combined_fn): New enum.
20837 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
20838 (commutative_op): New function.
20839 (commutate): Use it. Handle more than 2 operands.
20840 (dt_operand::gen_gimple_expr): Use commutative_op.
20841 (parser::parse_expr): Allow :c to be used with non-binary
20842 operators if the commutative operand is known.
20843 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
20844 CFN_FMS, CFN_FNMA and CFN_FNMS.
20845 (backprop::process_assign_use): Remove FMA_EXPR handling.
20846 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
20847 (gen_hsa_fma): New function.
20848 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
20849 IFN_FNMA and IFN_FNMS.
20850 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
20851 * gimple-fold.h (follow_all_ssa_edges): Declare.
20852 * gimple-fold.c (follow_all_ssa_edges): New function.
20853 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
20854 gimple_build interface and use follow_all_ssa_edges to fold the result.
20855 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
20856 instead of checking for optabs directly.
20857 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
20858 rather than FMA_EXPRs.
20859 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
20860 call to IFN_FMA instead of an FMA_EXPR.
20862 2018-05-17 Jim Wilson <jimw@sifive.com>
20864 * expr.c (do_tablejump): When converting index to Pmode, if we have a
20865 sign extended promoted subreg, and the range does not have the sign bit
20866 set, then do a sign extend.
20868 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
20869 test, check for sign extended subreg and/or constant operands, and
20870 do a sign extend in that case.
20872 2018-05-17 Steve Ellcey <sellcey@cavium.com>
20874 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
20875 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
20877 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
20878 Change logics_shift_reg to logics_shift_imm.
20879 (thunderx2t99_fp_loadpair_basic): Delete.
20880 (thunderx2t99_fp_storepair_basic): Delete.
20881 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
20882 (thunderx2t99_asimd_polynomial): Delete.
20883 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
20884 and neon_fp_mul_d_scalar_q.
20885 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
20886 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
20887 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
20888 (thunderx2t99_asimd_lut): Add missing tbl types.
20889 (thunderx2t99_asimd_ext): Delete.
20890 (thunderx2t99_asimd_load1_1_mult): Delete.
20891 (thunderx2t99_asimd_load1_2_mult): Delete.
20892 (thunderx2t99_asimd_load1_ldp): New.
20893 (thunderx2t99_asimd_load1): New.
20894 (thunderx2t99_asimd_load2): Add missing *load2* types.
20895 (thunderx2t99_asimd_load3): New.
20896 (thunderx2t99_asimd_load4): New.
20897 (thunderx2t99_asimd_store1_1_mult): Delete.
20898 (thunderx2t99_asimd_store1_2_mult): Delete.
20899 (thunderx2t99_asimd_store2_mult): Delete.
20900 (thunderx2t99_asimd_store2_onelane): Delete.
20901 (thunderx2t99_asimd_store_stp): New.
20902 (thunderx2t99_asimd_store1): New.
20903 (thunderx2t99_asimd_store2): New.
20904 (thunderx2t99_asimd_store3): New.
20905 (thunderx2t99_asimd_store4): New.
20907 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
20909 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
20910 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
20912 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
20913 Segher Boessenkool <segher@kernel.crashing.org>
20916 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
20919 2018-05-17 Richard Biener <rguenther@suse.de>
20921 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
20922 for pruning loop and prune defs feeding only already visited PHIs.
20924 2018-05-17 Richard Biener <rguenther@suse.de>
20926 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
20928 2018-05-17 Bin Cheng <bin.cheng@arm.com>
20929 Richard Biener <rguenther@suse.de>
20931 PR tree-optimization/85793
20932 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
20933 for VMAT_ELEMENTWISE.
20935 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
20937 * internal-fn.h (lookup_internal_fn): Declare
20938 * internal-fn.c (lookup_internal_fn): New function.
20939 * gimple.c (gimple_build_call_from_tree): Handle calls to
20940 internal functions.
20941 * gimple-pretty-print.c (dump_gimple_call): Print "." before
20942 internal function names.
20943 * tree-pretty-print.c (dump_generic_node): Likewise.
20944 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
20946 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
20948 * gimple-fold.h (gimple_build): Make the function forms take
20949 combined_fn rather than built_in_function.
20950 (gimple_simplify): Likewise.
20951 * gimple-match-head.c (gimple_simplify): Likewise.
20952 * gimple-fold.c (gimple_build): Likewise.
20953 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
20954 rather than gimple_build_call_internal.
20955 (get_initial_defs_for_reduction): Likewise.
20956 (vect_create_epilog_for_reduction): Likewise.
20957 (vectorizable_live_operation): Likewise.
20959 2018-05-17 Martin Liska <mliska@suse.cz>
20961 * gimple-ssa-sprintf.c (format_directive): Do not use
20962 space in between 'G_' and '('.
20964 2018-05-17 Jakub Jelinek <jakub@redhat.com>
20967 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
20968 even if the mask is not all ones.
20971 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
20973 (ix86_gimple_fold_builtin): Likewise.
20976 * config/i386/i386.c: Include tree-vector-builder.h.
20977 (ix86_vector_shift_count): New function.
20978 (ix86_fold_builtin): Fold shift builtins by scalar count.
20979 (ix86_gimple_fold_builtin): Likewise.
20981 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
20982 _mm512_setzero): New intrinsics.
20984 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
20985 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20987 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
20988 code generation for cases where splatting a value is not useful.
20989 * simplify-rtx.c (simplify_ternary_operation): Simplify
20990 vec_merge across a vec_duplicate and a paradoxical subreg forming
20991 a vector mode to a vec_concat.
20993 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
20995 * config.gcc: Support "goldmont-plus".
20996 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
20998 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20999 PROCESSOR_GOLDMONT_PLUS.
21000 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
21001 (processor_target_table): Add "goldmont-plus".
21002 (PTA_GOLDMONT_PLUS): Define.
21003 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
21004 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
21005 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
21006 (fold_builtin_cpu): Add "goldmont-plus".
21007 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
21008 (ix86_option_override_internal): Add "goldmont-plus".
21009 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
21010 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
21011 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
21012 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
21014 2018-05-17 Richard Biener <rguenther@suse.de>
21016 PR tree-optimization/85757
21017 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
21018 remove defs that only feed that PHI from further processing.
21020 2018-05-16 Jim Wilson <jimw@sifive.com>
21022 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
21024 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
21026 2018-05-16 Mark Wielaard <mark@klomp.org>
21028 * dwarf2out.c (count_index_strings): New function.
21029 (output_indirect_strings): Call count_index_strings and generate
21030 header for dwarf_version >= 5.
21032 2018-05-16 Mark Wielaard <mark@klomp.org>
21034 * dwarf2out.c (dwarf_FORM): New function.
21035 (set_indirect_string): Use dwarf_FORM.
21036 (reset_indirect_string): Likewise.
21037 (size_of_die): Likewise.
21038 (value_format): Likewise.
21039 (output_die): Likewise.
21040 (add_skeleton_AT_string): Likewise.
21041 (output_macinfo_op): Likewise.
21042 (index_string): Likewise.
21043 (output_index_string_offset): Likewise.
21044 (output_index_string): Likewise.
21045 (count_index_strings): Likewise.
21047 2018-05-16 Carl Love <cel@us.ibm.com>
21049 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
21050 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
21052 2018-05-16 Martin Jambor <mjambor@suse.cz>
21054 * ipa-prop.c (ipa_free_all_edge_args): Remove.
21055 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
21057 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
21059 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
21060 (fnma<mode>4): Likewise.
21061 (fms<mode>4): Likewise.
21062 (fnms<mode>4): Likewise.
21063 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
21064 (aarch64_fnma<mode>4): Likewise.
21065 (aarch64_fms<mode>4): Likewise.
21066 (aarch64_fnms<mode>4): Likewise.
21067 (aarch64_fnmadd<mode>4): Likewise.
21069 2018-05-16 Jason Merrill <jason@redhat.com>
21071 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
21073 2018-05-16 Richard Biener <rguenther@suse.de>
21075 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
21076 (dump_stmt_cost): Declare.
21077 (add_stmt_cost): Dump cost we add.
21078 (add_stmt_costs): New function.
21079 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
21080 No longer exported.
21081 (vect_analyze_stmt): Adjust prototype.
21082 (vectorizable_condition): Likewise.
21083 (vectorizable_live_operation): Likewise.
21084 (vectorizable_reduction): Likewise.
21085 (vectorizable_induction): Likewise.
21086 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
21087 cost vector to pass to vectorizable_ and record afterwards.
21088 (vect_model_reduction_cost): Take cost vector argument and adjust.
21089 (vect_model_induction_cost): Likewise.
21090 (vectorizable_reduction): Likewise.
21091 (vectorizable_induction): Likewise.
21092 (vectorizable_live_operation): Likewise.
21093 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
21094 SLP_TREE_NUMBER_OF_VEC_STMTS.
21095 (vect_analyze_slp_cost_1): Remove.
21096 (vect_analyze_slp_cost): Likewise.
21097 (vect_slp_analyze_node_operations): Take visited args and
21098 a target cost vector. Avoid processing already visited stmt sets.
21099 (vect_slp_analyze_operations): Use a local cost vector to gather
21100 costs and register those of non-discarded instances.
21101 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
21102 (vect_schedule_slp_instance): Remove copying of
21103 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
21105 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
21106 adding cost. Record cost entry location.
21107 (vect_prologue_cost_for_slp_op): Function to compute cost of
21108 a constant or invariant generated for SLP vect in the prologue,
21109 split out from vect_analyze_slp_cost_1.
21110 (vect_model_simple_cost): Make static. Adjust for SLP costing.
21111 (vect_model_promotion_demotion_cost): Likewise.
21112 (vect_model_store_cost): Likewise, make static.
21113 (vect_model_load_cost): Likewise.
21114 (vectorizable_bswap): Add cost vector arg and adjust.
21115 (vectorizable_call): Likewise.
21116 (vectorizable_simd_clone_call): Likewise.
21117 (vectorizable_conversion): Likewise.
21118 (vectorizable_assignment): Likewise.
21119 (vectorizable_shift): Likewise.
21120 (vectorizable_operation): Likewise.
21121 (vectorizable_store): Likewise.
21122 (vectorizable_load): Likewise.
21123 (vectorizable_condition): Likewise.
21124 (vectorizable_comparison): Likewise.
21125 (can_vectorize_live_stmts): Likewise.
21126 (vect_analyze_stmt): Likewise.
21127 (vect_transform_stmt): Adjust calls to vectorizable_*.
21128 * tree-vectorizer.c: Include gimple-pretty-print.h.
21129 (dump_stmt_cost): New function.
21131 2018-05-16 Richard Biener <rguenther@suse.de>
21133 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
21134 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
21135 * tree-ssa-dse.c: Include tree-ssa-loop.h.
21136 (check_name): New callback.
21137 (dse_classify_store): Track cycles via a visited bitmap of PHI
21138 defs and simplify handling of in-loop and across loop dead stores
21139 and properly fail for loop-variant refs. Handle byte-tracking with
21140 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
21143 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
21145 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
21146 (vect_get_mask_type_for_stmt): Likewise.
21147 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
21149 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
21150 to determine the statement's vector type and the vector type that
21151 should be used for calculating nunits. Deal with cases in which
21152 the type has to be deferred.
21153 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
21154 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
21155 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
21156 (vect_determine_vf_for_stmt): New functions, split out from...
21157 (vect_determine_vectorization_factor): ...here.
21158 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
21159 (vect_get_mask_type_for_stmt): New functions, split out from
21160 vect_determine_vectorization_factor.
21162 2018-05-16 Richard Biener <rguenther@suse.de>
21164 * tree-cfg.c (verify_gimple_assign_ternary): Properly
21165 verify the [VEC_]COND_EXPR embedded comparison.
21167 2018-05-15 Martin Sebor <msebor@redhat.com>
21169 PR tree-optimization/85753
21170 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
21171 RECORD_TYPE in addition to ARRAY_TYPE.
21173 2018-05-15 Martin Sebor <msebor@redhat.com>
21175 PR middle-end/85643
21176 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
21178 2018-05-15 Richard Biener <rguenther@suse.de>
21180 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
21181 add by_clobber_p one. Change algorithm to collect all defs
21182 representing uses we need to walk and try reducing them to
21183 a single one before failing.
21184 (dse_dom_walker::dse_optimize_stmt): Adjust.
21186 2018-05-13 Mark Wielaard <mark@klomp.org>
21188 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
21189 (size_of_loc_descr): Likewise.
21190 (output_loc_operands): Likewise.
21191 (output_loc_operands_raw): Likewise.
21192 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
21193 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
21194 (hash_loc_operands): Likewise.
21195 (compare_loc_operands): Likewise.
21197 2018-05-14 Mark Wielaard <mark@klomp.org>
21199 * dwarf2out.c (count_index_addrs): New function.
21200 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
21202 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21204 PR tree-optimization/83648
21205 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
21206 return value as malloc candidate.
21208 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21211 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
21212 param as true in call to suggest_attribute.
21214 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
21216 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
21217 -mreadonly-in-sdata.
21219 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21221 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
21223 (aarch64_crypto_aesd_fused): Likewise.
21225 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
21227 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
21228 (movsi_aarch64): Likewise.
21229 (load_pairsi): Likewise.
21230 (load_pairdi): Likewise.
21231 (store_pairsi): Likewise.
21232 (store_pairdi): Likewise.
21233 (load_pairsf): Likewise.
21234 (load_pairdf): Likewise.
21235 (store_pairsf): Likewise.
21236 (store_pairdf): Likewise.
21237 (zero_extend): Likewise.
21238 (trunc): Swap alternatives.
21239 (fcvt_target): Add '?' to prefer w over r.
21241 2018-05-14 Jakub Jelinek <jakub@redhat.com>
21244 * config/i386/i386.md: Disallow non-commutative arithmetics in
21245 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
21246 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
21247 in the peephole2 before it.
21249 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
21251 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
21252 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
21253 (ix86_handle_option): Handle -mcldemote.
21254 * config.gcc: New header.
21255 * config/i386/cldemoteintrin.h: New file.
21256 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
21257 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
21259 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21260 OPTION_MASK_ISA_CLDEMOTE.
21261 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
21262 (ix86_valid_target_attribute_inner_p): Ditto.
21263 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
21264 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
21265 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
21266 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
21267 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
21269 * config/i386/i386.opt: Add -mcldemote.
21270 * config/i386/x86intrin.h: New header.
21271 * doc/invoke.texi: Add -mcldemote.
21273 2018-05-14 Richard Biener <rguenther@suse.de>
21275 * doc/match-and-simplify.texi: Adjust :s documentation.
21277 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
21279 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
21280 intended memcpy size.
21281 (REORDER_45): Likewise.
21283 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
21285 * sort.cc: New file.
21286 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
21287 * vec.c (qsort_chk): Use gcc_qsort.
21288 * Makefile.in (OBJS-libcommon): Add sort.o.
21289 (build/sort.o): New target. Use it...
21290 (BUILD_RTL): ... here, and...
21291 (build/gencfn-macros): ... here, and...
21292 (build/genmatch): ... here.
21294 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
21295 Chung-Ju Wu <jasonwucj@gmail.com>
21297 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
21298 * config/nds32/nds32-graywolf.md: New file.
21299 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
21300 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
21302 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
21303 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
21304 * config/nds32/nds32.md (pipeline_model): Add graywolf.
21305 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
21306 * config/nds32/pipelines.md: Include n15 settings.
21308 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
21309 Chung-Ju Wu <jasonwucj@gmail.com>
21311 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
21312 * config/nds32/nds32-n13.md: New file.
21313 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
21314 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
21316 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
21317 * config/nds32/nds32.md (pipeline_model): Add n13.
21318 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
21319 * config/nds32/pipelines.md: Include n13 settings.
21321 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
21322 Chung-Ju Wu <jasonwucj@gmail.com>
21324 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
21325 * config/nds32/nds32-n10.md: New file.
21326 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
21327 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
21329 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
21330 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
21331 * config/nds32/nds32.md (pipeline_model): Add n10.
21332 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
21333 * config/nds32/pipelines.md: Include n10 settings.
21335 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
21336 Kito Cheng <kito.cheng@gmail.com>
21337 Chung-Ju Wu <jasonwucj@gmail.com>
21339 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
21340 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
21341 Add enum values for DSP extension instructions.
21342 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
21344 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
21345 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
21346 New code iterators.
21347 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
21348 * config/nds32/nds32-dspext.md: New file for DSP implementation.
21349 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
21350 * config/nds32/nds32-intrinsic.md: Likewise.
21351 * config/nds32/nds32_intrinsic.h: Likewise.
21352 * config/nds32/nds32-md-auxiliary.c: Likewise.
21353 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
21354 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
21355 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
21356 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
21357 * config/nds32/nds32-protos.h: New declarations for DSP extension.
21358 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
21359 TYPE_DMAC in switch statement.
21360 * config/nds32/nds32.c: New checking and implementation for DSP
21361 extension instructions.
21362 * config/nds32/nds32.h: Likewise.
21363 * config/nds32/nds32.md: Likewise.
21364 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
21365 * config/nds32/predicates.md: Implement new predicates for DSP
21368 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
21370 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
21371 Reformat alternatives and attributes so it is easier to identify
21372 which constraints/attributes go with which instruction.
21373 (mov<mode>_hardfloat32, FMOVE64): Likewise.
21374 (mov<mode>_softfloat32, FMOVE64): Likewise.
21375 (mov<mode>_hardfloat64, FMOVE64): Likewise.
21376 (mov<mode>_softfloat64, FMOVE64): Likewise.
21378 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
21380 * doc/extend.texi (PowerPC Built-in Functions): Rename this
21382 (Basic PowerPC Built-in Functions): The new name of the
21383 subsection previously known as "PowerPC Built-in Functions".
21384 (Basic PowerPC Built-in Functions Available on all Configurations):
21386 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
21387 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
21388 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
21389 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
21391 2018-05-11 Martin Jambor <mjambor@suse.cz>
21394 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
21397 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
21400 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
21402 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
21404 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
21405 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
21406 (ix86_handle_option): Handle -mwaitpkg.
21407 * config.gcc: New header.
21408 * config/i386/cpuid.h (bit_WAITPKG): New bit.
21409 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
21410 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
21412 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21413 OPTION_MASK_ISA_WAITPKG.
21414 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
21415 (ix86_option_override_internal): Add PTA_WAITPKG.
21416 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
21417 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
21418 IX86_BUILTIN_TPAUSE.
21419 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
21420 __builtin_ia32_umwait and __builtin_ia32_tpause.
21421 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
21422 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
21423 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
21424 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
21425 UNSPECV_TPAUSE): New.
21426 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
21427 * config/i386/i386.opt: Add -mwaitpkg.
21428 * config/i386/waitpkgintrin.h: New file.
21429 * config/i386/x86intrin.h: New header.
21430 * doc/invoke.texi: Add -mwaitpkg.
21432 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
21435 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
21437 (cortex-m0): Use armv6s-m isa.
21438 (cortex-m0plus): Likewise.
21439 (cortex-m1): Likewise.
21440 (cortex-m0.small-multiply): Likewise.
21441 (cortex-m0plus.small-multiply): Likewise.
21442 (cortex-m1.small-multiply): Likewise.
21444 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
21445 Jakub Jelinek <jakub@redhat.com>
21447 PR tree-optimization/85692
21448 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
21449 source permute as well.
21451 2018-05-11 Martin Liska <mliska@suse.cz>
21454 * doc/extend.texi: Document LLVM style format for no_sanitize
21457 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
21459 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
21460 mode_supports_vsx_dform_quad to mode_supports_dq_form.
21461 (mode_supports_vsx_dform_quad): Likewise.
21462 (mode_supports_vmx_dform): Move these functions to be next to the
21463 other mode_supports functions.
21464 (mode_supports_dq_form): Likewise.
21465 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
21466 mode_supports_dq_form.
21467 (reg_offset_addressing_ok_p): Likewise.
21468 (offsettable_ok_by_alignment): Likewise.
21469 (rs6000_legitimate_offset_address_p): Likewise.
21470 (legitimate_lo_sum_address_p): Likewise.
21471 (rs6000_legitimize_address): Likewise.
21472 (rs6000_legitimize_reload_address): Likewise.
21473 (rs6000_secondary_reload_inner): Likewise.
21474 (rs6000_preferred_reload_class): Likewise.
21475 (rs6000_output_move_128bit): Likewise.
21477 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
21479 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
21480 Generate SImode target register for null target.
21481 <case IX86_BUILTIN_XGETBV>: Ditto.
21482 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
21483 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
21485 2018-05-10 Carl Love <cel@us.ibm.com>
21487 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
21488 dcbtt and dcbtstt if operands[2] is 0.
21490 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
21493 * config/i386/sse.md (usadv64qi): New expander.
21495 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
21497 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
21498 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
21499 -maltivec=be support.
21500 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
21501 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
21502 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
21503 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
21504 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
21505 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
21506 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
21507 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
21508 altivec_vsumsws): Adjust.
21509 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
21510 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
21511 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
21513 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
21514 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
21515 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
21516 (altivec_lve<VI_char>x): Delete expand.
21517 (*altivec_lve<VI_char>x_internal): Rename to...
21518 (altivec_lve<VI_char>x): ... this.
21519 (altivec_lvxl_<mode>): Delete expand.
21520 (*altivec_lvxl_<mode>_internal): Rename to ...
21521 (altivec_lvxl_<mode>): ... this.
21522 (altivec_stvxl_<mode>): Delete expand.
21523 (*altivec_stvxl_<mode>_internal): Rename to ...
21524 (altivec_stvxl_<mode>): ... this.
21525 (altivec_stve<VI_char>x): Delete expand.
21526 (*altivec_stve<VI_char>x_internal): Rename to ...
21527 (altivec_stve<VI_char>x): ... this.
21528 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
21529 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
21530 reduc_plus_scal_<mode>): Adjust.
21531 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
21533 (rs6000_cpu_cpp_builtins): Adjust.
21534 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
21535 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
21536 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
21537 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
21538 -maltivec=be support.
21539 (rs6000_split_vec_extract_var): Adjust.
21540 (rs6000_split_v4si_init): Adjust.
21541 (swap_selector_for_mode): Delete.
21542 (altivec_expand_lvx_be, altivec_expand_stvx_be,
21543 altivec_expand_stvex_be): Delete.
21544 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
21545 -maltivec=be support.
21546 (rs6000_gimple_fold_builtin): Ditto.
21547 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
21549 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
21550 (TARGET_DIRECT_MOVE_64BIT): Adjust.
21551 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
21552 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
21553 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
21554 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
21555 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
21556 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
21557 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
21558 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
21559 anonymous split): Adjust.
21560 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
21561 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
21563 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
21565 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
21566 when --with-gxx-include-dir is also specified.
21567 * configure: Regenerate.
21569 2018-05-09 Jim Wilson <jimw@sifive.com>
21572 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
21573 * config/riscv/t-withmultilib: New.
21574 * config/riscv/withmultilib.h: New.
21575 * doc/install.texi: Document RISC-V --with-multilib-list support.
21577 2018-05-09 Richard Biener <rguenther@suse.de>
21579 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
21581 (vect_bb_vectorization_profitable_p): Adjust. Compute
21582 actual scalar cost using the cost vector and the add_stmt_cost
21585 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
21587 PR rtl-optimization/85645
21588 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
21589 in the REG_CFA_REGISTER note for LR, don't leave it empty.
21591 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
21593 PR rtl-optimization/85645
21594 * shrink-wrap.c (spread_components): Return a boolean saying if
21595 anything was changed.
21596 (try_shrink_wrapping_separate): Iterate spread_components until
21597 nothing changes anymore.
21599 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
21601 PR rtl-optimization/85645
21602 * regrename.c (build_def_use): Also kill the chains that include the
21603 destination of a REG_CFA_REGISTER note.
21605 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
21607 PR rtl-optimization/85645
21608 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
21609 insn that has a REG_CFA_REGISTER note.
21611 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
21613 * cfgexpand.c (expand_clobber): New function.
21614 (expand_gimple_stmt_1): Use it.
21615 * tree-vect-stmts.c (vect_clobber_variable): New function,
21617 (vectorizable_simd_clone_call): ...here.
21618 (vectorizable_store): Emit a clobber either side of an
21619 IFN_STORE_LANES sequence.
21620 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
21622 2018-05-09 Tom de Vries <tom@codesourcery.com>
21625 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
21626 (define_insn "trap_if_false"): Add exit after trap.
21628 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
21630 PR rtl-optimization/85638
21631 * bb-reorder.c: Include common/common-target.h.
21632 (create_forwarder_block): New function extracted from...
21633 (fix_up_crossing_landing_pad): ...here. Rename into...
21634 (dw2_fix_up_crossing_landing_pad): ...this.
21635 (sjlj_fix_up_crossing_landing_pad): New function.
21636 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
21637 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
21638 from both partitions and exit the loop after one iteration.
21640 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
21643 * doc/extend.texi (PowerPC Built-in Functions): Rename this
21645 (Basic PowerPC Built-in Functions): The new name of the
21646 subsection previously known as "PowerPC Built-in Functions".
21647 (Basic PowerPC Built-in Functions Available on all Configurations):
21649 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
21651 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
21653 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
21655 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
21658 2018-05-08 Jim Wilson <jimw@sifive.com>
21660 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
21661 (LD_EMUL_SUFFIX): New.
21662 (LINK_SPEC): Use it.
21664 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
21666 * doc/extend.texi (PowerPC Built-in Functions): Rename this
21668 (Basic PowerPC Built-in Functions): The new name of the
21669 subsection previously known as "PowerPC Built-in Functions".
21670 (Basic PowerPC Built-in Functions Available on all Configurations):
21672 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
21674 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
21676 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
21678 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
21681 2018-05-08 Jakub Jelinek <jakub@redhat.com>
21684 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
21685 after cmpelim optimization.
21687 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
21689 * config.gcc: Support "goldmont".
21690 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
21691 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21692 PROCESSOR_GOLDMONT.
21693 * config/i386/i386.c (m_GOLDMONT): Define.
21694 (processor_target_table): Add "goldmont".
21695 (PTA_GOLDMONT): Define.
21696 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
21697 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
21698 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
21699 (fold_builtin_cpu): Add "goldmont".
21700 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
21701 (ix86_option_override_internal): Add "goldmont".
21702 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
21703 (processor_type): Add PROCESSOR_GOLDMONT.
21704 * config/i386/i386.md: Add CPU "glm".
21705 * config/i386/glm.md: New file.
21706 * config/i386/x86-tune.def: Add m_GOLDMONT.
21707 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
21709 2018-05-08 Jakub Jelinek <jakub@redhat.com>
21712 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
21714 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
21715 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
21716 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
21719 * config/i386/i386.c (ix86_fold_builtin): Handle
21720 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
21723 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
21724 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
21726 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
21729 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
21730 (check_arch): Likewise.
21731 (check_fpu): Return the result rather than printing it.
21732 (end arch): Fix operator precedence.
21733 (end cpu): Likewise.
21734 (END): Print the result from check_fpu.
21736 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
21737 Alan Hayward <alan.hayward@arm.com>
21738 David Sherwood <david.sherwood@arm.com>
21740 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
21741 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
21742 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
21743 (*fcmuo<mode>_and): New patterns.
21745 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
21747 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
21748 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
21749 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
21750 (cmp_op, sve_imm_con): New code attributes.
21751 (SVE_COND_INT_CMP, imm_con): Delete.
21752 (cmp_op): Remove above unspecs from int attribute.
21753 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
21755 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
21756 comparison-specific unspecs.
21757 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
21758 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
21759 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
21760 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
21761 (*vec_fcm<cmp_op><mode>): Rename to...
21762 (*fcm<cmp_op><mode>): ...this and adjust likewise.
21763 (*vec_fcmuo<mode>): Rename to...
21764 (*fcmuo<mode>): ...this and adjust likewise.
21765 (*pred_fcm<cmp_op><mode>): New pattern.
21766 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
21767 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
21769 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
21771 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
21772 (aarch64_emit_sve_predicated_cond): New function.
21773 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
21774 (aarch64_emit_unspec_cond_or): Replace with...
21775 (aarch64_emit_sve_or_conds): ...this new function. Use
21776 aarch64_emit_sve_ptrue_op for the individual comparisons and
21777 aarch64_emit_binop to OR them together.
21778 (aarch64_emit_inverted_unspec_cond): Replace with...
21779 (aarch64_emit_sve_inverted_cond): ...this new function. Use
21780 aarch64_emit_sve_ptrue_op for the comparison and
21781 aarch64_emit_unop to invert the result.
21782 (aarch64_expand_sve_vec_cmp_float): Update after the above
21783 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
21785 2018-05-07 Nathan Sidwell <nathan@acm.org>
21787 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
21788 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
21789 (Backwards Compatibility): Likewise.
21791 2018-05-07 Luis Machado <luis.machado@linaro.org>
21795 2018-05-07 Luis Machado <luis.machado@linaro.org>
21797 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21798 <prefetch_dynamic_strides>: New const bool field.
21799 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21800 prefetch_dynamic_strides.
21801 (exynosm1_prefetch_tune): Likewise.
21802 (thunderxt88_prefetch_tune): Likewise.
21803 (thunderx_prefetch_tune): Likewise.
21804 (thunderx2t99_prefetch_tune): Likewise.
21805 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
21807 (aarch64_override_options_internal): Update to set
21808 PARAM_PREFETCH_DYNAMIC_STRIDES.
21809 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
21810 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
21811 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
21812 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
21813 prefetch-dynamic-strides setting.
21815 2018-05-07 Luis Machado <luis.machado@linaro.org>
21817 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21818 <minimum_stride>: New const int field.
21819 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21820 minimum_stride field.
21821 (exynosm1_prefetch_tune): Likewise.
21822 (thunderxt88_prefetch_tune): Likewise.
21823 (thunderx_prefetch_tune): Likewise.
21824 (thunderx2t99_prefetch_tune): Likewise.
21825 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
21826 (aarch64_override_options_internal): Update to set
21827 PARAM_PREFETCH_MINIMUM_STRIDE.
21828 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
21829 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
21830 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
21831 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
21832 stride is constant and is below the minimum stride threshold.
21834 2018-05-07 Luis Machado <luis.machado@linaro.org>
21836 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
21839 2018-05-07 Luis Machado <luis.machado@linaro.org>
21841 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21842 <prefetch_dynamic_strides>: New const bool field.
21843 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21844 prefetch_dynamic_strides.
21845 (exynosm1_prefetch_tune): Likewise.
21846 (thunderxt88_prefetch_tune): Likewise.
21847 (thunderx_prefetch_tune): Likewise.
21848 (thunderx2t99_prefetch_tune): Likewise.
21849 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
21851 (aarch64_override_options_internal): Update to set
21852 PARAM_PREFETCH_DYNAMIC_STRIDES.
21853 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
21854 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
21855 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
21856 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
21857 prefetch-dynamic-strides setting.
21859 2018-05-07 Luis Machado <luis.machado@linaro.org>
21861 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
21862 <minimum_stride>: New const int field.
21863 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
21864 minimum_stride field.
21865 (exynosm1_prefetch_tune): Likewise.
21866 (thunderxt88_prefetch_tune): Likewise.
21867 (thunderx_prefetch_tune): Likewise.
21868 (thunderx2t99_prefetch_tune): Likewise.
21869 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
21870 (aarch64_override_options_internal): Update to set
21871 PARAM_PREFETCH_MINIMUM_STRIDE.
21872 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
21873 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
21874 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
21875 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
21876 stride is constant and is below the minimum stride threshold.
21878 2018-05-06 Jakub Jelinek <jakub@redhat.com>
21881 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
21882 the type is addressable. Don't force op into register if it has
21885 2018-05-05 Roland McGrath <mcgrathr@google.com>
21888 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
21889 any section for which we don't know a specific type it should have,
21890 regardless of name. Previously this was done only for the exact
21891 names ".init_array", ".fini_array", and ".preinit_array".
21892 (default_elf_asm_named_section): Add comment about
21893 relationship with default_section_type_flags and SECTION_NOTYPE.
21894 (get_section): Don't consider it a type conflict if one side has
21895 SECTION_NOTYPE and the other doesn't, as long as neither has the
21896 SECTION_BSS et al used in the default_section_type_flags logic.
21898 2018-05-05 Tom de Vries <tom@codesourcery.com>
21901 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
21902 (workaround_barsyncs): New function.
21903 (nvptx_reorg): Use workaround_barsyncs.
21904 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
21905 (define_expand "nvptx_membar_cta"): New define_expand.
21906 (define_insn "*nvptx_membar_cta"): New insn.
21908 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
21910 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
21911 To improve optimization opportunities.
21912 * builtin-types.def: The new needed builtin types for the above.
21914 2018-05-04 Richard Biener <rguenther@suse.de>
21916 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
21917 * gimple-ssa-store-merging.c
21918 (imm_store_chain_info::output_merged_store): Remove redundant create,
21919 release split_store vector contents on failure.
21920 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
21921 scalar stmt vector on cache hit.
21923 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
21925 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
21927 * config.gcc (powerpc-xilinx-eabi*): Remove.
21928 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
21930 (fusion_addis_mem_combo_load): Ditto.
21931 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
21933 (rs6000_cpu_cpp_builtins): Ditto.
21934 * config/rs6000/rs6000-linux.c
21935 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
21936 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
21937 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
21939 (rs6000_setup_reg_addr_masks): Ditto.
21940 (rs6000_init_hard_regno_mode_ok): Ditto.
21941 (rs6000_option_override_internal): Ditto.
21942 (legitimate_lo_sum_address_p): Ditto.
21943 (rs6000_legitimize_address): Ditto.
21944 (rs6000_legitimize_reload_address): Ditto.
21945 (rs6000_legitimate_address_p): Ditto.
21946 (abi_v4_pass_in_fpr): Ditto.
21947 (setup_incoming_varargs): Ditto.
21948 (rs6000_gimplify_va_arg): Ditto.
21949 (rs6000_split_multireg_move): Ditto.
21950 (rs6000_savres_strategy): Ditto.
21951 (rs6000_emit_prologue_components): Ditto.
21952 (rs6000_emit_epilogue_components): Ditto.
21953 (rs6000_emit_prologue): Ditto.
21954 (rs6000_emit_epilogue): Ditto.
21955 (rs6000_elf_file_end): Ditto.
21956 (rs6000_function_value): Ditto.
21957 (rs6000_libcall_value): Ditto.
21958 * config/rs6000/rs6000.h: Ditto.
21959 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
21960 (TARGET_MINMAX): ... this. New.
21961 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
21962 * config/rs6000/rs6000.md: Remove Xilinx FP support.
21963 (*movsi_internal1_single): Delete.
21964 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
21965 mfpu=, mxilinx-fpu): Delete.
21966 * config/rs6000/singlefp.h: Delete.
21967 * config/rs6000/sysv4.h: Remove Xilinx FP support.
21968 * config/rs6000/t-rs6000: Ditto.
21969 * config/rs6000/t-xilinx: Delete.
21970 * config/rs6000/titan.md: Adjust for fp_type removal.
21971 * config/rs6000/vsx.md: Remove Xilinx FP support.
21972 (VStype_simple): Delete.
21973 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
21974 * config/rs6000/xfpu.h: Delete.
21975 * config/rs6000/xfpu.md: Delete.
21976 * config/rs6000/xilinx.h: Delete.
21977 * config/rs6000/xilinx.opt: Delete.
21978 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
21979 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
21981 2018-05-04 Tom de Vries <tom@codesourcery.com>
21984 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
21987 2018-05-04 Richard Biener <rguenther@suse.de>
21989 PR middle-end/85627
21990 * tree-complex.c (update_complex_assignment): We are always in SSA form.
21991 (expand_complex_div_wide): Likewise.
21992 (expand_complex_operations_1): Likewise.
21993 (expand_complex_libcall): Preserve EH info of the original stmt.
21994 (tree_lower_complex): Handle removed blocks.
21995 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
21996 on complex multiplication and division libcall builtins.
21998 2018-05-04 Richard Biener <rguenther@suse.de>
22000 PR middle-end/85574
22001 * fold-const.c (negate_expr_p): Restrict negation of operand
22002 zero of a division to when we know that can happen without
22004 (fold_negate_expr_1): Likewise.
22006 2018-05-04 Jakub Jelinek <jakub@redhat.com>
22009 * real.h (real_nextafter): Declare.
22010 * real.c (real_nextafter): New function.
22011 * fold-const-call.c (fold_const_nextafter): New function.
22012 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
22013 CASE_CFN_NEXTTOWARD.
22014 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
22015 even when arg1_mode is different from arg0_mode.
22017 2018-05-03 Nathan Sidwell <nathan@acm.org>
22019 * doc/extend.texi (Deprecated Features): Remove
22020 -ffriend-injection.
22021 (Backwards Compatibility): Likewise.
22022 * doc/invoke.texi (C++ Language Options): Likewise.
22023 (C++ Dialect Options): Likewise.
22025 2018-05-03 Jakub Jelinek <jakub@redhat.com>
22028 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
22029 _mm512_mask_mullox_epi64): New intrinsics.
22031 2018-05-03 Tom de Vries <tom@codesourcery.com>
22034 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
22035 dump files): Add offload-tree.
22037 2018-05-03 Richard Biener <rguenther@suse.de>
22039 PR tree-optimization/85615
22040 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
22041 to loops not nested in BBs loop father to avoid creating multi-entry
22044 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22046 PR tree-optimization/70291
22047 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
22048 arguments. Change return type to tree. Emit libcall as a new
22049 statement rather than replacing existing one when inplace_p is true.
22050 (expand_complex_multiplication_components): New function.
22051 (expand_complex_multiplication): Expand floating-point complex
22052 multiplication using the above.
22053 (expand_complex_division): Rename inner_type parameter to type.
22054 Update expand_complex_libcall call-site.
22055 (expand_complex_operations_1): Update expand_complex_multiplication
22056 and expand_complex_division call-sites.
22058 2018-05-02 Jakub Jelinek <jakub@redhat.com>
22061 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
22062 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
22063 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
22064 the highest significant bit of the shift count mask is clear. In
22065 check whether and[sq]i3 is needed verify that all significant bits
22066 of the shift count other than the highest are set.
22068 2018-05-02 Tom de Vries <tom@codesourcery.com>
22071 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
22072 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
22073 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
22074 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
22075 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
22076 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
22077 __builtin_goacc_parlevel_size.
22079 2018-05-02 Richard Biener <rguenther@suse.de>
22081 PR tree-optimization/85597
22082 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
22083 do not use split vect_get_vec_defs call but call vect_get_slp_defs
22086 2018-05-02 Tom de Vries <tom@codesourcery.com>
22089 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
22090 dump files): Add ltrans-tree.
22092 2018-05-02 Tom de Vries <tom@codesourcery.com>
22095 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
22096 dump files): Add wpa-ipa.
22098 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
22100 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
22101 powerpc*-*-linux*paired* target.
22102 * config/rs6000/750cl.h: Delete.
22103 * config/rs6000/paired.h: Delete.
22104 * config/rs6000/paired.md: Delete.
22105 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
22107 * config/rs6000/rs6000-builtin.def: Remove paired float support.
22108 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
22109 comment. Remove paired float support.
22110 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
22111 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
22113 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
22114 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
22116 * config/rs6000/rs6000.c: Remove paired float support.
22117 (paired_expand_vector_init, paired_expand_vector_move,
22118 paired_emit_vector_compare, paired_emit_vector_cond_expr,
22119 (paired_expand_lv_builtin, paired_expand_stv_builtin,
22120 paired_expand_builtin, paired_expand_predicate_builtin,
22121 paired_init_builtins): Delete.
22122 * config/rs6000/rs6000.h: Remove paired float support.
22123 * config/rs6000/rs6000.md: Remove paired float support.
22124 (move_from_CR_ov_bit): Delete.
22125 * config/rs6000/rs6000.opt (mpaired): Delete.
22126 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
22127 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
22129 2018-05-02 Richard Biener <rguenther@suse.de>
22131 PR middle-end/85567
22132 * gimplify.c (gimplify_save_expr): When in SSA form allow
22133 SAVE_EXPRs to compute to SSA vars.
22135 2018-05-02 Jakub Jelinek <jakub@redhat.com>
22138 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
22139 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
22140 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
22141 clobber operands[2], instead use a new pseudo. Formatting fixes.
22143 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
22145 PR tree-optimization/85586
22146 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
22147 exit early for statements in the same group if the accesses are
22150 2018-05-02 Tom de Vries <tom@codesourcery.com>
22153 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
22156 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
22158 PR tree-optimization/85143
22159 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
22161 2018-05-01 Tom de Vries <tom@codesourcery.com>
22164 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
22165 not found" error message.
22167 2018-05-01 Tom de Vries <tom@codesourcery.com>
22170 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
22171 * vec.c (test_ordered_remove_if): New function.
22172 (vec_c_tests): Call test_ordered_remove_if.
22173 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
22174 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
22175 * tree-vect-patterns.c (vect_pattern_recog_1): Use
22176 VEC_ORDERED_REMOVE_IF.
22178 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22180 PR tree-optimization/82665
22181 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
22182 pointer subtraction where arguments come from a memchr call.
22184 2018-05-01 Jakub Jelinek <jakub@redhat.com>
22186 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
22187 --push-state --as-needed and --pop-state instead of --as-needed and
22188 --no-as-needed if ld supports it.
22189 * configure: Regenerated.
22192 * doc/install.texi2html: Replace _002d with - and _002a with * in
22193 generated html files using sed.
22195 2018-04-30 David Malcolm <dmalcolm@redhat.com>
22198 * gcc-rich-location.c (blank_line_before_p): New function.
22199 (use_new_line): New function.
22200 (gcc_rich_location::add_fixit_insert_formatted): New function.
22201 * gcc-rich-location.h
22202 (gcc_rich_location::add_fixit_insert_formatted): New function.
22204 2018-04-30 David Malcolm <dmalcolm@redhat.com>
22206 * selftest.c (assert_streq): Rename "expected" and "actual" to
22207 "val1" and "val2". Extend NULL-handling to cover both inputs
22208 symmetrically, while still requiring both to be non-NULL for a pass.
22209 * selftest.h (assert_streq): Rename "expected" and "actual" to
22211 (ASSERT_EQ): Likewise.
22212 (ASSERT_EQ_AT): Likewise.
22213 (ASSERT_KNOWN_EQ): Likewise.
22214 (ASSERT_KNOWN_EQ_AT): Likewise.
22215 (ASSERT_NE): Likewise.
22216 (ASSERT_MAYBE_NE): Likewise.
22217 (ASSERT_MAYBE_NE_AT): Likewise.
22218 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
22219 the assertion to pass.
22220 (ASSERT_STREQ_AT): Likewise.
22222 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
22224 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
22225 interaction with -pie.
22227 2018-04-30 David Malcolm <dmalcolm@redhat.com>
22229 * selftest.h: Fix alphabetization of per-source-file selftest
22232 2018-04-30 Jason Merrill <jason@redhat.com>
22234 PR c++/61982 - dead stores to destroyed objects.
22235 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
22238 2018-04-30 Jason Merrill <jason@redhat.com>
22240 * tree.c (build_clobber): New.
22241 * tree.h: Declare it.
22242 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
22244 2018-04-30 David Malcolm <dmalcolm@redhat.com>
22246 * diagnostic-show-locus.c (layout::layout): Update for
22247 location_get_source_line returning a char_span.
22248 (struct char_span): Move to input.h.
22249 (struct correction): Update for fields in char_span becoming
22251 (struct source_line): Update for location_get_source_line
22252 returning a char_span.
22253 (layout::print_line): Likewise.
22254 * edit-context.c (edited_file::print_content): Likewise.
22255 (edited_file::print_diff_hunk): Likewise.
22256 (edited_file::print_run_of_changed_lines): Likewise.
22257 (edited_file::get_num_lines): Likewise.
22258 (edited_line::edited_line): Likewise.
22259 * final.c (asm_show_source): Likewise.
22260 * input.c (location_get_source_line): Convert return type
22261 from const char * to char_span, losing the final "line_len"
22263 (dump_location_info): Update for the above.
22264 (get_substring_ranges_for_loc): Likewise. Use a char_span
22265 when handling the literal within the line.
22266 (test_reading_source_line): Update for location_get_source_line
22267 returning a char_span.
22268 * input.h (class char_span): Move here from
22269 diagnostic-show-locus.c, converting from a struct to a class.
22270 Make data members private.
22271 (char_span::operator bool): New.
22272 (char_span::length): New.
22273 (char_span::get_buffer): New.
22274 (char_span::operator[]): New.
22275 (char_span::subspan): Make const.
22276 (char_span::xstrdup): New.
22277 (location_get_source_line): Convert return type from const char *
22278 to char_span, losing the final "line_size" param.
22280 2018-04-30 Jan Hubicka <jh@suse.cz>
22282 * lto-wrapper.c (ltrans_priorities): New static var.
22283 (cmp_priority): New.
22284 (run_gcc): Read priorities and if doing parallel build order
22285 the Makefile by them.
22287 2018-04-30 David Malcolm <dmalcolm@redhat.com>
22289 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
22291 2018-04-30 Richard Biener <rguenther@suse.de>
22293 * tree-cfg.c (verify_address): Remove base argument, add
22294 flag whether to check TREE_ADDRESSABLE and do that.
22295 (verify_expr): Remove.
22296 (verify_types_in_gimple_reference): Add pieces from verify_expr.
22297 (verify_gimple_assign_single): Likewise.
22298 (verify_gimple_switch): Likewise.
22299 (verify_expr_location_1): Dereference tp once. Add (disabled)
22300 piece from verify_expr.
22301 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
22303 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
22305 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
22307 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
22309 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
22310 (small_data_pattern): Likewise.
22311 (arc_rewrite_small_data): Likewise.
22312 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
22313 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
22314 (get_symbol_alignment): New function.
22315 (legitimate_small_data_address_p): Likewise.
22316 (legitimate_scaled_address): Update, call
22317 legitimate_small_data_address_p.
22318 (output_sdata): New static variable.
22319 (arc_print_operand): Update how we handle small data operands.
22320 (arc_print_operand_address): Likewise.
22321 (arc_legitimate_address_p): Update, use
22322 legitimate_small_data_address_p.
22323 (arc_rewrite_small_data_p): Remove.
22324 (arc_rewrite_small_data_1): Likewise.
22325 (arc_rewrite_small_data): Likewise.
22326 (small_data_pattern): Likewise.
22327 (compact_sda_memory_operand): Update to use
22328 legitimate_small_data_address_p and get_symbol_alignment.
22329 (prepare_move_operands): Don't rewite sdata pattern.
22330 (prepare_extend_operands): Remove.
22331 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
22333 (zero_extendqisi2): Likewise.
22334 (zero_extendhisi2): Likewise.
22335 (extendqihi2): Likewise.
22336 (extendqisi2): Likewise.
22337 (extendhisi2): Likewise.
22338 (addsi3): Likewise.
22339 (subsi3): Likewise.
22340 (andsi3): Likewise.
22341 * config/arc/constraints.md (Usd): Change it to memory constraint.
22343 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
22345 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
22346 as source of std instructions.
22347 * config/arc/arc.md (movsi_insn): Update pattern predicate to
22348 allow 6-bit constants as source for store instructions.
22349 (movdi_insn): Update instruction pattern to allow 6-bit constants
22350 as source for store instructions.
22352 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
22354 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
22356 2018-04-30 Nathan Sidwell <nathan@acm.org>
22357 Sandra Loosemore <sandra@codesourcery.com>
22359 * dumpfile.c (dump_open): Allow '-' for stdout.
22360 * doc/invoke.texi (Developer Options): Document dump filename
22361 determination early. Document stdin/stdout selection.
22363 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
22365 Microblaze Target: PIC data text relative
22367 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
22368 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
22370 * config/microblaze/microblaze.h (microblaze_constant_address_p):
22371 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
22372 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
22373 New addressing mode for data-text relative position indepenedent code.
22374 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
22375 'ADDRESS_SYMBOLIC_TXT_REL'.
22376 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
22377 (microblaze_legitimate_pic_operand): Exclude function calls from
22378 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
22379 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
22381 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
22382 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
22383 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
22384 for 'address + offset'.
22385 (microblaze_expand_prologue): Add new function prologue call for
22387 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
22388 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
22389 table in case of TARGET_PIC_DATA_TEXT_REL.
22390 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
22391 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
22392 Add new macros 'UNSPEC_TEXT',
22393 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
22394 + exclude function calls from 'UNSPEC_PLT' in case of data text
22396 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
22397 new target hook for generating address diff vector tables in case of
22399 * doc/tm.texi : Regenerate.
22400 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
22401 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
22402 of addr diff vector generation.
22403 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
22404 target hook definition.
22405 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
22406 Add default function for generate_pic_addr_diff_vec -> flag_pic.
22407 * doc/invoke.texi (Add new pic option): Add new microblaze pic
22408 option for data text relative.
22410 2018-04-30 Richard Biener <rguenther@suse.de>
22412 * tree-chrec.h (evolution_function_is_constant_p): Remove
22414 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
22416 2018-04-30 Richard Biener <rguenther@suse.de>
22419 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
22421 2018-04-30 Richard Biener <rguenther@suse.de>
22423 PR tree-optimization/28364
22424 PR tree-optimization/85275
22425 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
22426 copying first exit test.
22428 2018-04-28 Mark Wielaard <mark@klomp.org>
22430 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
22431 dwarf_version >= 5.
22432 (dwarf_AT): Handle DW_AT_addr_base.
22433 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
22435 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
22438 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
22439 (*ashl<dwi>3_doubleword_mask_1): Ditto.
22440 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
22441 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
22443 2018-04-28 Richard Biener <rguenther@suse.de>
22445 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
22446 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
22447 to reflect use. Only add interesting stmts.
22449 2018-04-27 Martin Jambor <mjambor@suse.cz>
22452 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
22453 the jump function allows for passing through aggregate values.
22455 2018-04-27 David Malcolm <dmalcolm@redhat.com>
22457 * input.h (in_system_header_at): Convert from macro to inline
22459 (from_macro_expansion_at): Likewise.
22460 (from_macro_definition_at): Likewise.
22462 2018-04-27 Jeff Law <law@redhat.com>
22464 * config.gcc: Mark tile* targets as deprecated/obsolete.
22466 2018-04-27 Richard Biener <rguenther@suse.de>
22468 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
22471 2018-04-27 Richard Biener <rguenther@suse.de>
22473 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
22475 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
22477 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
22478 with Yd constraint. Set "preferred_for_speed" attribute from
22479 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
22480 with Yd constraint.
22481 (*movdi_internal): Ditto.
22482 (movti_interunit splitters): Remove
22483 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
22484 (movdi_interunit splitters): Ditto.
22485 * config/i386/constraints.md (Ye): Remove.
22486 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
22488 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22491 * config/aarch64/constraints.md (Usg): Limit to 31.
22492 (Usj): Limit to 63.
22494 2018-04-27 Jakub Jelinek <jakub@redhat.com>
22496 PR tree-optimization/85529
22497 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
22498 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
22499 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
22500 zero extension or masking of the MSB bit.
22501 (optimize_range_tests): Add FIRST_BB argument, pass it through
22502 to optimize_range_tests_var_bound.
22503 (maybe_optimize_range_tests, reassociate_bb): Adjust
22504 optimize_range_tests callers.
22506 2018-04-26 Richard Biener <rguenther@suse.de>
22507 Jakub Jelinek <jakub@redhat.com>
22509 * cgraph.h (symbol_table): Just declare debug method here.
22510 * symtab.c (symbol_table::debug): Define.
22512 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
22514 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
22516 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
22518 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
22519 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
22520 (*movdi_internal): Substitute Yi and Yj constraint with x
22521 and Ym and Yn constraint with y constraint. Update "isa"
22522 attribute and set "preferred_for_speed" attribute from
22523 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
22524 (*movsi_internal): Ditto.
22525 (*movdf_internal): Ditto.
22526 (*movsf_internal): Ditto.
22527 (*zero_extendsidi2): Ditto.
22528 * config/i386/sse.md (vec_set<mode>_0): Ditto.
22529 (sse2_loadld): Ditto.
22530 (*vec_extract<ssevecmodelower>_0): Ditto.
22531 (*vec_extractv4si_0_zext_sse4): Ditto.
22532 (vec_concatv2di): Ditto.
22533 (*vec_dup<mode>): Ditto.
22534 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
22535 * config/i386/constraints.md (Yi): Remove.
22540 2018-04-26 Nathan Sidwell <nathan@acm.org>
22542 * dumpfile.c (dump_open): New.
22543 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
22544 (dump_finish): Detect stdio/stderr by value not name.
22546 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
22548 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
22550 2018-04-26 Tom de Vries <tom@codesourcery.com>
22553 * config/nvptx/nvptx.c (verify_neutering_jumps)
22554 (verify_neutering_labels): New function
22555 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
22557 2018-04-26 Tom de Vries <tom@codesourcery.com>
22560 * config/nvptx/nvptx.c (needs_neutering_p): New function.
22561 (nvptx_single): Use needs_neutering_p to skip over insns that do not
22564 2018-04-26 Richard Biener <rguenther@suse.de>
22565 Tom de Vries <tom@codesourcery.com>
22568 * lto-streamer-out.c (output_function): Fixup loops if required to match
22569 discovery done in the reader.
22571 2018-04-26 Richard Biener <rguenther@suse.de>
22573 PR tree-optimization/85116
22574 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
22575 have a loop exit from the single latch predecessor. Remove
22576 case of header with just condition.
22577 (ch_base::copy_headers): Exclude infinite loops from any
22579 (pass_ch::execute): Record exits.
22581 2018-04-26 Richard Biener <rguenther@suse.de>
22583 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
22584 prologue cost vector and pass it to vect_get_load_cost.
22585 (vect_get_peeling_costs_all_drs): Likewise.
22586 (vect_peeling_hash_get_lowest_cost): Likewise.
22587 (vect_enhance_data_refs_alignment): Likewise.
22589 2018-04-26 Richard Biener <rguenther@suse.de>
22591 PR middle-end/85450
22592 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
22593 checking of integer<->pointer conversions.
22594 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
22595 sign-/zero-extending pointer types.
22596 (expand_omp_for_static_chunk): Likewise.
22598 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
22599 Jean Lee <xiaoyur347@gmail.com>
22601 * config/mips/mips.c (mips_asan_shadow_offset): New function.
22602 (TARGET_ASAN_SHADOW_OFFSET): Define.
22603 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
22604 true for -fsanitize=address.
22606 2018-04-25 Mark Wielaard <mark@klomp.org>
22608 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
22611 2018-04-25 Jakub Jelinek <jakub@redhat.com>
22613 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
22614 than "alu", remove explicit "memory" and "imm_disp" attributes.
22615 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
22617 PR middle-end/85414
22618 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
22619 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
22620 gen_lowpart_no_emit.
22622 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
22625 * config/i386/i386.c (ix86_expand_builtin): Change memory
22626 operand to XI, extend p0 to Pmode.
22627 * config/i386/i386.md: Change unspec volatile and operand
22628 1 mode to XI, change operand 0 mode to P.
22630 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
22632 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
22633 GET_MODE_MASK before any checking.
22634 (nds32_can_use_bset_p): Likewise.
22635 (nds32_can_use_btgl_p): Likewise.
22637 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
22639 * config/nds32/nds32-doubleword.md: New define_split pattern for
22640 illegal register number.
22642 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
22644 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
22646 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
22648 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
22650 2018-04-25 Richard Biener <rguenther@suse.de>
22652 * lto-streamer.h (LTO_major_version): Bump to 8.
22654 2018-04-25 Jakub Jelinek <jakub@redhat.com>
22656 * BASE-VER: Set to 9.0.0.
22658 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
22660 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
22661 in __abskf2 and __powikf2.
22663 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22666 * config/aarch64/constraints.md (Usg, Usj): New constraints.
22667 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
22668 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
22669 Use the above on operand 2. Reindent.
22670 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
22672 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
22675 * common/config/i386/i386-common.c (ix86_handle_option): Don't
22677 * config/i386/i386.opt (mcet): Removed.
22678 * doc/install.texi: Remove -mcet documentation.
22679 * doc/invoke.texi: Likewise.
22681 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
22684 * doc/install.texi: Remove -mcet from bootstrap-cet.
22686 2018-04-24 Jakub Jelinek <jakub@redhat.com>
22689 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
22690 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
22694 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
22695 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
22696 containing a CONST_VECTOR.
22698 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
22700 * doc/install.texi: Update newlib dependency for nvptx.
22702 2018-04-24 Jakub Jelinek <jakub@redhat.com>
22705 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
22706 instead of INTVAL when shifting x left.
22708 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
22710 PR tree-optimization/85478
22711 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
22712 vect_grouped_store_supported for single element vectors.
22714 2018-04-24 Richard Biener <rguenther@suse.de>
22717 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
22718 load cost increase to the case of non-constant step.
22720 2018-04-24 Jakub Jelinek <jakub@redhat.com>
22723 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
22724 destination if any_malformed_asm.
22726 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
22728 PR middle-end/85496
22729 * expr.c (store_field): In the bitfield case, if the value comes from
22730 a function call and is returned in registers by means of a PARALLEL,
22731 do not change the mode of the temporary unless BLKmode and VOIDmode.
22733 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
22735 PR rtl-optimization/85423
22736 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
22737 dependencies to debug insns when the previous insn is non-debug.
22739 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
22741 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
22742 enums into a single definition.
22743 (fls): Fix predicates and printing.
22746 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
22748 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
22749 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
22750 and short u6 immediate.
22751 (check_if_valid_sleep_operand): Remove.
22752 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
22754 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
22756 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
22757 flag_always_save_lp condition.
22758 * config/nds32/nds32.opt (malways-save-lp): New option.
22760 2018-04-22 Shiva Chen <shiva0217@gmail.com>
22762 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
22763 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
22764 * config/nds32/nds32.h
22765 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
22766 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
22768 2018-04-22 Shiva Chen <shiva0217@gmail.com>
22770 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
22771 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
22773 2018-04-22 Shiva Chen <shiva0217@gmail.com>
22774 Chung-Ju Wu <jasonwucj@gmail.com>
22776 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
22778 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
22779 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
22781 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
22783 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
22785 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
22787 * config/nds32/nds32-protos.h (nds32_data_alignment,
22788 nds32_local_alignment): Declare.
22789 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
22790 nds32_local_alignment): New functions.
22791 (TARGET_CONSTANT_ALIGNMENT): Define.
22792 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
22794 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
22796 * config/nds32/nds32.c
22797 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
22798 (TARGET_MODES_TIEABLE_P): Likewise.
22800 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
22802 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
22803 level Ofast and Og.
22805 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
22806 Chung-Ju Wu <jasonwucj@gmail.com>
22808 * config/nds32/constants.md (unspec_volatile_element): Add enum values
22809 for unaligned access.
22810 * config/nds32/nds32-intrinsic.c: Implementation of expanding
22812 * config/nds32/nds32-intrinsic.md: Likewise.
22813 * config/nds32/nds32_intrinsic.h: Likewise.
22814 * config/nds32/nds32.h (nds32_builtins): Likewise.
22815 * config/nds32/nds32.opt (munaligned-access): New option.
22816 * config/nds32/nds32.c (nds32_asm_file_start): Display
22817 flag_unaligned_access status.
22819 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
22821 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
22822 -mno-relax is present.
22823 * config/riscv/linux.h (LINK_SPEC): Ditto.
22825 2018-04-20 Martin Sebor <msebor@redhat.com>
22828 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
22830 (gimple_fold_builtin_stxcpy_chk): Same.
22831 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
22833 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
22836 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
22837 __powikf2 when long double is IEEE 128-bit.
22839 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
22841 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
22842 step to make sure stack always aligned.
22844 2018-04-20 Carl Love <cel@us.ibm.com>
22847 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
22848 size check for arg0.
22850 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
22851 Tom de Vries <tom@codesourcery.com>
22854 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
22855 Emit insns for calls too.
22856 (nvptx_find_par): Always look for worker-level predecessor insn.
22857 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
22859 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
22860 (nvptx_process_pars): Propagate frames for calls.
22862 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
22865 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
22867 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
22868 (ix86_handle_option): Don't handle OPT_mibt.
22869 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
22871 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
22873 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
22874 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
22875 (ix86_target_macros): Define __CET__ with flag_cf_protection
22876 for -fcf-protection.
22877 * config/i386/i386.c (isa2_opts): Remove -mibt.
22878 * config/i386/i386.h (TARGET_IBT): Removed.
22879 (TARGET_IBT_P): Likewise.
22880 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
22881 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
22882 * config/i386/i386.opt (mcet): Update help message.
22883 (mshstk): Likewise.
22885 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
22886 -mcet as an alias for -mshstk.
22888 2018-04-20 Richard Biener <rguenther@suse.de>
22890 PR middle-end/85475
22891 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
22892 complexity by forcing a single use of the multiply operand.
22894 2018-04-20 Martin Jambor <mjambor@suse.cz>
22897 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
22898 recursion dependency to only apply to non-clones.
22900 2018-04-20 Martin Jambor <mjambor@suse.cz>
22903 * ipa-cp.c (create_specialized_node): Check that clones of
22904 self-recursive edges exist during IPA-CP.
22906 2018-04-19 Toon Moene <toon@moene.org>
22908 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
22911 2018-04-19 Jakub Jelinek <jakub@redhat.com>
22913 PR tree-optimization/85467
22914 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
22915 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
22916 VECTOR_CST element to type.
22918 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
22921 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
22922 * config/i386/i386.md (builtin_setjmp_setup): Removed.
22923 (builtin_longjmp): Likewise.
22924 (save_stack_nonlocal): New pattern.
22925 (restore_stack_nonlocal): Likewise.
22927 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
22930 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
22931 Replace ASM_OUTPUT_LABEL with fprintf.
22933 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
22936 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
22937 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
22938 * config/i386/i386-c.c (ix86_target_macros_internal): Also
22939 define __IBT__ and __SHSTK__ for -fcf-protection.
22940 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
22942 (ix86_trampoline_init): Likewise.
22943 (x86_output_mi_thunk): Likewise.
22944 (ix86_notrack_prefixed_insn_p): Likewise.
22945 (ix86_option_override_internal): Don't disallow -fcf-protection.
22946 * config/i386/i386.md (rdssp<mode>): Also enable for
22948 (incssp<mode>): Likewise.
22949 (nop_endbr): Likewise.
22950 * config/i386/i386.opt (mcet): Change help message to built-in
22953 (mshstk): Likewise.
22954 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
22955 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
22956 enable CET built-in functions.
22958 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
22960 * common/config/i386/i386-common.c
22961 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
22962 OPTION_MASK_ISA_MOVDIRI_UNSET,
22963 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
22964 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
22965 * config.gcc (movdirintrin.h): New header.
22966 * config/i386/cpuid.h (bit_MOVDIRI,
22967 bit_MOVDIR64B): New bits.
22968 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
22970 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
22971 (VOID, PVOID, PCVOID)): New function types.
22972 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
22973 __builtin_ia32_directstoreu_u64,
22974 __builtin_ia32_movdir64b): New builtins.
22975 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
22976 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
22978 (ix86_valid_target_attribute_inner_p): Ditto.
22979 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
22980 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
22981 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
22982 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
22983 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
22984 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
22985 (movdiri<mode>, movdir64b_<mode>): New.
22986 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
22987 * config/i386/immintrin.h: Include movdirintrin.h.
22988 * config/i386/movdirintrin.h: New file.
22989 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
22991 2018-04-19 Richard Biener <rguenther@suse.de>
22993 PR middle-end/85455
22994 * cfg.c (clear_bb_flags): When loop state says we have
22995 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
22997 2018-04-19 Richard Biener <rguenther@suse.de>
22999 PR tree-optimization/84737
23000 * tree-vect-data-refs.c (vect_copy_ref_info): New function
23001 copying restrict info.
23002 (vect_setup_realignment): Use it.
23003 * tree-vectorizer.h (vect_copy_ref_info): Declare.
23004 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
23005 the first DR to all generated stores.
23006 (vectorizable_load): Likewise for loads.
23008 2018-04-19 Jakub Jelinek <jakub@redhat.com>
23010 PR tree-optimization/85446
23011 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
23012 the integral and pointer types to have the same precision.
23014 * doc/install.texi: Document --disable-cet being the default and
23017 2018-04-18 Martin Liska <mliska@suse.cz>
23019 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
23022 2018-04-18 Martin Liska <mliska@suse.cz>
23025 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
23028 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
23029 arguments if they are comparable.
23031 2018-04-18 Martin Liska <mliska@suse.cz>
23034 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
23037 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23040 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
23043 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
23044 ENDBR after calling __morestack.
23046 2018-04-18 David Malcolm <dmalcolm@redhat.com>
23049 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
23050 by using gcc_base_ver to generate a gcc_driver_version, and use
23051 it when generating GCC_DRIVER_NAME.
23052 * configure: Regenerate.
23054 2018-04-18 Jakub Jelinek <jakub@redhat.com>
23057 * config.gcc: Obsolete powerpc*-*-*spe*.
23059 2018-04-17 Jakub Jelinek <jakub@redhat.com>
23062 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
23063 (stabstr_D): Change type of unum from unsigned int to
23064 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
23067 2018-04-17 Jim Wilson <jimw@sifive.com>
23070 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
23071 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
23072 Set arg_pointer_offset after using pretend_args_size.
23074 2018-04-17 Jakub Jelinek <jakub@redhat.com>
23076 PR rtl-optimization/85431
23077 * dse.c (record_store): Ignore zero width stores.
23080 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
23081 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
23082 __builtin_stack_restore rather than after it.
23083 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
23084 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
23085 argument instead of virtual_dynamic_stack_rtx.
23087 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
23089 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
23091 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23092 Add note to error message to explain internal mapping of overloaded
23093 built-in function name to non-overloaded built-in function name.
23094 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
23097 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
23100 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
23101 where the inputs overlap with the output.
23103 2018-04-17 Jakub Jelinek <jakub@redhat.com>
23105 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
23106 (=v, v) alternative and explicit "memory" attribute.
23107 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
23108 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
23110 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
23111 "sselog1" type instead of "sselog".
23112 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
23113 "sselog". Remove explicit "memory" attribute.
23114 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
23115 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
23117 (vec_extract_hi_v32hi): Merge all alternatives into one, use
23118 "sselog1" type instead of "sselog". Remove explicit "memory"
23120 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
23121 use "sselog1" type instead of "sselog". Remove explicit "memory"
23123 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
23124 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
23126 (vec_extract_hi_v64qi): Merge all alternatives into one, use
23127 "sselog1" type instead of "sselog". Remove explicit "memory"
23129 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
23130 use "sselog1" type instead of "sselog". Remove explicit "memory"
23134 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
23136 PR middle-end/85414
23137 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
23140 2018-04-17 Martin Jambor <mjambor@suse.cz>
23143 * ipa-cp.c (create_specialized_node): Call
23144 expand_all_artificial_thunks if necessary.
23146 2018-04-17 Martin Liska <mliska@suse.cz>
23149 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
23150 in message, remote space in between '_G' and '('.
23152 2018-04-17 Jakub Jelinek <jakub@redhat.com>
23155 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
23156 avx512f_vmcmp<mode>3<round_saeonly_name>,
23157 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
23158 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
23159 avx512f_rndscale<mode><round_saeonly_name>,
23160 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
23161 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
23162 Use %<iptr>2 instead of %2 for -masm=intel.
23163 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
23164 avx512f_vcvttss2usi<round_saeonly_name>,
23165 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
23167 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
23168 avx512f_vcvttsd2usi<round_saeonly_name>,
23169 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
23170 Use %q1 instead of %1 for -masm=intel.
23171 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
23172 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
23173 of %3 for -masm=intel.
23174 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
23176 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
23178 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
23180 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
23181 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
23183 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
23185 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
23186 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
23187 %g1 and one with %0 and %1.
23188 (avx512er_vmrcp28<mode><round_saeonly_name>,
23189 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
23190 %1 for -masm=intel.
23191 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
23192 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
23193 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
23194 of %0 and %{%4%} for -masm=intel.
23195 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
23196 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
23197 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
23198 order of %0 and %{%5%}%{z%} for -masm=intel.
23200 2018-04-17 Jan Hubicka <jh@suse.cz>
23203 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
23205 2018-04-17 Martin Liska <mliska@suse.cz>
23208 * multiple_target.c (create_dispatcher_calls): Set apostrophes
23209 for target_clone error message. Make default implementation
23210 clone to be a local declaration.
23211 (separate_attrs): Add new argument and check for an empty
23213 (expand_target_clones): Handle it.
23214 (ipa_target_clone): Make redirection just for target_clones
23217 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
23218 Tom de Vries <tom@codesourcery.com>
23220 PR middle-end/84955
23221 * omp-expand.c (expand_oacc_for): Add dummy false branch for
23222 tiled basic blocks without omp continue statements.
23224 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
23227 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
23228 vec_extract expression as having side effects to make sure it gets
23231 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
23234 * config/i386/i386.c (get_builtin_code_for_version): Check
23237 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
23240 * config.gcc: Support "skylake".
23241 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23243 * config/i386/i386.c (m_SKYLAKE): Define.
23244 (processor_target_table): Add "skylake".
23245 (ix86_option_override_internal): Add "skylake".
23246 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
23247 PROCESSOR_CANNONLAKE.
23248 (get_builtin_code_for_version): Fix priority for
23249 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
23250 PROCESSOR_SKYLAKE-AVX512.
23251 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
23252 (processor_type): Add PROCESSOR_SKYLAKE.
23254 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
23255 Jason Merrill <jason@redhat.com>
23258 * convert.c (convert_to_integer_1): Use direct recursion for
23259 enumeral types and types with a precision less than the number
23260 of bits in their mode.
23262 2018-04-16 Julia Koval <julia.koval@intel.com>
23265 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
23266 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
23268 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
23271 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
23272 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
23273 and -mno-direct-move.
23275 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
23278 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
23279 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
23280 Ensure negative shifts result in {0}.
23282 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
23284 PR rtl-optimization/79916
23285 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
23286 regs (if any) to define how to gnerate SD moves when LRA is in
23289 2018-04-13 Jakub Jelinek <jakub@redhat.com>
23291 PR rtl-optimization/85393
23292 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
23293 * except.c (expand_dw2_landing_pad_for_region): Make static.
23294 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
23295 a label and unconditional jump to old_bb, rather than
23296 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
23299 PR rtl-optimization/85376
23300 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
23301 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
23302 instead of a specific value.
23304 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
23305 Bin Cheng <bin.cheng@arm.com>
23307 PR tree-optimization/82965
23308 PR tree-optimization/83991
23309 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
23310 by_profile_only parameter.
23311 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
23312 information if the loop was predicted to iterate too many times.
23313 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
23315 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
23318 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
23321 2018-04-13 Martin Liska <mliska@suse.cz>
23322 Jakub Jelinek <jakub@redhat.com>
23324 PR middle-end/81657
23325 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
23326 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
23327 * builtins.c (expand_builtin_memory_copy_args): Use
23328 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
23329 handle dest_addr == pc_rtx.
23331 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
23334 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
23335 asked to not generate direct moves.
23336 (fix_trunc<mode>si2_stfiwx): Similar.
23337 (fix_trunc<mode>si2_internal): Similar.
23339 2018-04-12 Jakub Jelinek <jakub@redhat.com>
23342 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
23343 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
23344 lookup if dest in some wider mode is known to be const0_rtx and
23345 if so, record permanent equivalence for it to be ZERO_EXTEND of
23346 the narrower mode destination.
23348 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
23350 * lto-streamer-out.c (output_function): Revert 259346.
23351 * omp-expand.c (expand_oacc_for): Likewise.
23353 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
23355 PR rtl-optimization/85354
23356 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
23357 * sel-sched.c (sel_global_init): ... here.
23359 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
23362 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
23363 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
23364 mode for PE-COFF targets.
23365 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
23366 (i386_pe_asm_lto_end): Likewise.
23367 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
23368 (TARGET_ASM_LTO_END): Likewise.
23369 * config/i386/winnt.c (saved_debug_info_level): New static variable.
23370 (i386_pe_asm_lto_start): New function.
23371 (i386_pe_asm_lto_end): Likewise.
23373 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
23374 Richard Biener <rguenther@suse.de>
23376 PR middle-end/84955
23377 * lto-streamer-out.c (output_function): Fix CFG loop state before
23379 * omp-expand.c (expand_oacc_for): Handle calls to internal
23380 functions like regular functions.
23382 2018-04-12 Richard Biener <rguenther@suse.de>
23385 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
23386 for the early LTO debug to properly generate references to it
23387 during DIE emission. Do not re-use that for the skeleton for
23389 (dwarf2out_early_finish): Likewise.
23391 2018-04-12 Jakub Jelinek <jakub@redhat.com>
23394 * config/i386/sse.md
23395 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
23396 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
23397 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
23398 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
23399 and output is a reg, avoid creating invalid lowpart subreg, but
23400 instead split into a 512-bit move. Don't split if not AVX512VL,
23401 input is xmm16+ reg and output is a mem.
23402 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
23403 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
23404 xmm16+ reg and output is a mem.
23406 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
23408 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
23409 also for flag_dwarf2_cfi_asm.
23411 2018-04-12 Jakub Jelinek <jakub@redhat.com>
23413 PR rtl-optimization/85342
23414 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
23415 a bool scalar var inside of the loop instead. Don't try to update
23416 recog_data.operand after failed apply_change_group.
23418 2018-04-12 Tom de Vries <tom@codesourcery.com>
23421 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
23422 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
23423 array with flexible array member as array without given dimension.
23424 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
23425 argument for undefined param to true.
23427 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
23430 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
23431 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
23432 from PowerPC section.
23433 * config/rs6000/sysv4.opt (mcall-): Improve help text.
23434 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
23435 help text that is too long.
23436 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
23437 help text that is too long.
23438 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
23439 help text that is too long.
23441 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
23443 * config/alpha/alpha.md (stack_probe_internal): Rename
23444 from "probe_stack". Update all callers.
23446 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
23448 PR rtl-optimization/84566
23449 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
23450 sched_macro_fuse_insns.
23452 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
23455 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
23456 (compute_block_dependences): ... from here.
23458 2018-04-11 Jakub Jelinek <jakub@redhat.com>
23460 PR tree-optimization/85331
23461 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
23462 from int to HOST_WIDE_INT.
23464 2018-04-11 Martin Jambor <mjambor@suse.cz>
23467 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
23468 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
23469 not the same as the source val.
23470 (cgraph_edge_brings_value_p): New parameter.
23471 (gather_edges_for_value): Pass destination value to
23472 cgraph_edge_brings_value_p.
23473 (perhaps_add_new_callers): Likewise.
23474 (get_info_about_necessary_edges): Likewise and exclude values brought
23475 only by self-recursive edges.
23476 (create_specialized_node): Redirect only clones of self-calling edges.
23477 (+self_recursive_pass_through_p): New function.
23478 (find_more_scalar_values_for_callers_subset): Use it.
23479 (find_aggregate_values_for_callers_subset): Likewise.
23480 (known_aggs_to_agg_replacement_list): Removed.
23481 (decide_whether_version_node): Re-calculate known constants for all
23482 remaining context clones.
23484 2018-04-11 Richard Biener <rguenther@suse.de>
23487 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
23488 from early DWARF output.
23489 (dwarf2out_early_finish): Output line info unconditionally into
23490 early DWARF and add reference to it.
23492 2018-04-11 Jakub Jelinek <jakub@redhat.com>
23495 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
23496 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
23497 other than V2DFmode using iptr mode attribute.
23498 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
23500 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
23502 PR rtl-optimization/84659
23503 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
23505 2018-04-11 Jakub Jelinek <jakub@redhat.com>
23508 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
23510 (output_loc_list): Pass address of a dummy size variable even in the
23511 locview handling loop.
23512 (index_location_lists): Add comment on why skip_loc_list_entry can't
23515 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
23518 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
23521 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
23524 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
23525 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
23526 and -mstring-compare-inline-limit.
23528 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
23531 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
23532 for stack clash protection in a register whenever we need it to be in
23535 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
23537 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
23538 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
23540 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
23543 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
23545 (mlong-double-): Ditto.
23546 * config/rs6000/sysv4.opt (msdata=): Ditto.
23547 (mtls-size=): Ditto.
23549 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
23551 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
23552 erroneous entries for
23553 "vector int vec_ldl (int, long int *)", and
23554 "vector unsigned int vec_ldl (int, unsigned long int *)".
23555 Add comments and entries for
23556 "vector bool char vec_ldl (int, bool char *)",
23557 "vector bool short vec_ldl (int, bool short *)",
23558 "vector bool int vec_ldl (int, bool int *)",
23559 "vector bool long long vec_ldl (int, bool long long *)",
23560 "vector pixel vec_ldl (int, pixel *)",
23561 "vector long long vec_ldl (int, long long *)",
23562 "vector unsigned long long vec_ldl (int, unsigned long long *)".
23563 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
23564 type tree bool_long_long_type_node and correct definition of
23565 bool_V2DI_type_node to make reference to this new type tree.
23566 (rs6000_mangle_type): Replace erroneous reference to
23567 bool_long_type_node with bool_long_long_type_node.
23568 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
23569 comments to emphasize sign distinctions for char and int types and
23570 replace RS6000_BTI_bool_long constant with
23571 RS6000_BTI_bool_long_long constant. Also add comment to restrict
23572 use of RS6000_BTI_pixel.
23573 (bool_long_type_node): Remove this macro definition.
23574 (bool_long_long_type_node): New macro definition
23576 2018-04-10 Jakub Jelinek <jakub@redhat.com>
23578 PR rtl-optimization/85300
23579 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
23580 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
23581 simplify_unary_operation fails.
23583 2018-04-10 Martin Liska <mliska@suse.cz>
23585 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
23586 cgraph_edge and ipa_ref.
23588 2018-04-10 Jakub Jelinek <jakub@redhat.com>
23592 * config/i386/sse.md
23593 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
23594 computation of the VEC_MERGE selector from mask.
23595 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
23596 Fix decoding of the VEC_MERGE selector into mask.
23598 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
23600 PR tree-optimization/85286
23601 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
23603 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
23605 * final.c (final_1): Set insn_last_address as well as
23606 insn_current_address.
23608 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23611 * explow.c (emit_stack_probe): Call validize_mem on memory location
23612 before passing it to gen_probe_stack. Create address operand and
23613 legitimize it for the probe_stack_address case.
23615 2018-04-09 Jan Hubicka <jh@suse.cz>
23618 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
23619 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
23620 * tree.c (free_lang_data_in_type): Fix handling of binfos;
23622 (free_lang_data): Rebuild type inheritance graph.
23624 2018-04-09 Martin Sebor <msebor@redhat.com>
23626 * invoke.texi (-finline-small-functions): Mention other optimization
23628 (-findirect-inlining, -fpartial-inlining): Same.
23629 (-finline-functions-called-once): Same.
23630 (-freorder-blocks-and-partition): Same.
23632 2018-04-09 Jan Hubicka <jh@suse.cz>
23635 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
23636 jumps; choose last target that matches the criteria (i.e.
23637 no partition changes for non-crossing jumps).
23638 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
23639 support for redirecting crossing jumps to non-crossing.
23641 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
23643 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
23644 also for naked functions.
23646 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
23648 * config/arc/arc.md (add_shift): New pattern.
23649 (add_shift2): Likewise.
23650 (sub_shift): Likewise.
23651 (sub_shift_cmp0_noout): Likewise.
23652 (compare_si_ashiftsi): Likewise.
23653 (xbfu_cmp0_noout): New combine pattern.
23654 (xbfu_cmp0"): Likewise.
23655 (movsi_set_cc_insn): Place the predicable variant first.
23656 (commutative_binary_cmp0_noout): Remove clobber.
23657 (commutative_binary_cmp0): New pattern.
23658 (noncommutative_binary_cmp0): Likewise.
23659 (noncommutative_binary_cmp0_noout): Likewise.
23660 (noncommutative_binary_comparison_result_used): Removed.
23661 (rsub_cmp0): New pattern.
23662 (rsub_cmp0_noout): Likewise.
23663 (extzvsi): Changed, keep only meaningful variants.
23664 (SQH, SEZ): New iterators.
23665 (SQH_postfix): New mode attribute.
23666 (SEZ_prefix): New code attribute.
23667 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
23668 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
23669 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
23670 of numerical value.
23671 (noncommutative_operator): Check the availability of barrel
23674 2018-04-09 Richard Biener <rguenther@suse.de>
23676 PR tree-optimization/85284
23677 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
23678 Only use the niter constraining form of simple_iv when the exit
23679 is always executed.
23681 2018-04-09 Tom de Vries <tom@codesourcery.com>
23684 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
23685 (define_expand "*memory_barrier"): New define_expand.
23686 (define_insn "memory_barrier"): New insn.
23688 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
23690 PR rtl-optimization/80463
23691 PR rtl-optimization/83972
23692 PR rtl-optimization/83480
23694 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
23695 correct producer for the insn.
23696 (tidy_control_flow): Fixup seqnos in case of debug insns.
23698 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
23700 PR rtl-optimization/83913
23702 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
23703 different sched-times when merging exprs.
23705 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
23707 PR rtl-optimization/83962
23709 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
23710 tidy_fallthru_edge and tidy_control_flow.
23712 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
23714 PR rtl-optimization/83530
23716 * sel-sched.c (force_next_insn): New global variable.
23717 (remove_insn_for_debug): When force_next_insn is true, also leave only
23718 next insn in the ready list.
23719 (sel_sched_region): When the region wasn't scheduled, make another pass
23720 over it with force_next_insn set to 1.
23722 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
23724 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
23726 * config/nds32/constants.md (unspec_volatile_element): Add enum values
23727 for interrupt control.
23728 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
23729 functions for interrupt control.
23730 * config/nds32/nds32-intrinsic.md: Likewise.
23731 * config/nds32/nds32_intrinsic.h: Likewise.
23732 * config/nds32/nds32.h (nds32_builtins): Likewise.
23734 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
23736 * config/nds32/nds32.c (nds32_init_machine_status,
23737 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
23738 strict_aligned_p field.
23739 (nds32_expand_to_rtl_hook): New function.
23740 (TARGET_EXPAND_TO_RTL_HOOK): Define.
23741 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
23743 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
23744 Chung-Ju Wu <jasonwucj@gmail.com>
23746 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
23747 * config/nds32/nds32-n7.md: New file.
23748 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
23749 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
23751 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
23752 * config/nds32/nds32.md (pipeline_model): Add n7.
23753 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
23754 * config/nds32/pipelines.md: Include n7 settings.
23756 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
23757 Chung-Ju Wu <jasonwucj@gmail.com>
23759 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
23760 * config/nds32/nds32-e8.md: New file.
23761 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
23762 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
23764 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
23765 * config/nds32/nds32.md (pipeline_model): Add e8.
23766 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
23767 * config/nds32/pipelines.md: Include e8 settings.
23769 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
23770 Chung-Ju Wu <jasonwucj@gmail.com>
23772 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
23773 * config/nds32/nds32-n8.md: New file.
23774 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
23775 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
23777 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
23778 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
23779 * config/nds32/nds32.md (pipeline_model): Add n8.
23780 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
23781 * config/nds32/pipelines.md: Include n8 settings.
23783 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
23784 Chung-Ju Wu <jasonwucj@gmail.com>
23786 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
23787 * config/nds32/nds32-n9-2r1w.md: New file.
23788 * config/nds32/nds32-n9-3r2w.md: New file.
23789 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
23790 nds32_register_ports): New or modify for cpu n9.
23791 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
23793 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
23794 * config/nds32/nds32-utils.c: New file.
23795 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
23796 TARGET_MUL_SLOW): Define.
23797 * config/nds32/nds32.md (pipeline_model): New attribute.
23798 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
23799 New options that support cpu n9.
23800 * config/nds32/pipelines.md: Include n9 settings.
23801 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
23803 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
23805 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
23806 information if necessary.
23807 (output_cond_branch_compare_zero): Likewise.
23808 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
23809 (nds32_target_alignment): Refine for alignment.
23810 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
23811 (FUNCTION_BOUNDARY): Modify.
23812 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
23814 * config/nds32/nds32.opt (malways-align, malign-functions): New.
23816 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
23818 * config/nds32/constants.md (unspec_volatile_element): Add values for
23819 TLB operation and data prefetch.
23820 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
23821 functions for TLB operation and data prefetch.
23822 * config/nds32/nds32-intrinsic.md: Likewise.
23823 * config/nds32/nds32_intrinsic.h: Likewise.
23824 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
23825 (nds32_print_operand): Likewise.
23826 * config/nds32/nds32.h (nds32_builtins): Likewise.
23828 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
23829 Andrew Pinski <pinsika@gcc.gnu.org>
23831 PR middle-end/82976
23832 * match.pd: Use constant_boolean_node of correct type instead of
23833 boolean_true_node or boolean_false_node for simplifying
23834 pointer comparisons to zero.
23836 2018-04-07 Jakub Jelinek <jakub@redhat.com>
23838 PR tree-optimization/80021
23839 * tree.c (verify_type_variant): Make error call in verify_variant_match
23840 translatable and remove final full stop.
23842 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
23844 * config/nds32/constants.md (unspec_volatile_element): Add
23845 UNSPEC_VOLATILE_EH_RETURN.
23846 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
23847 nds32_output_stack_pop): Support dwarf exception handling process.
23848 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
23849 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
23850 exception handling process.
23851 (nds32_compute_stack_frame): Likewise.
23852 (nds32_return_addr_rtx): Likewise.
23853 (nds32_initial_elimination_offset): Likewise.
23854 (nds32_expand_prologue): Likewise.
23855 (nds32_expand_epilogue): Likewise.
23856 (nds32_dynamic_chain_address): New function.
23857 * config/nds32/nds32.h (machine_function): Add fields for dwarf
23858 exception handling.
23859 (DYNAMIC_CHAIN_ADDRESS): Define.
23860 (EH_RETURN_DATA_REGNO): Define.
23861 (EH_RETURN_STACKADJ_RTX): Define.
23862 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
23863 patterns for dwarf exception handling.
23865 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
23867 * config/nds32/nds32.h: Clean up obsolete macros.
23869 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
23871 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
23872 Add enum values for particular instructions.
23873 * config/nds32/nds32-intrinsic.c: Implementation of expanding
23874 particular intrinsic functions.
23875 * config/nds32/nds32-intrinsic.md: Likewise.
23876 * config/nds32/nds32_intrinsic.h: Likewise.
23877 * config/nds32/nds32.h (nds32_builtins): Likewise.
23878 * config/nds32/nds32.md (type): Add pbsad and pbsada.
23879 (btst, ave): New patterns for particular instructions.
23881 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
23883 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
23884 Add enum values for atomic load/store and memory sync.
23885 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
23887 * config/nds32/nds32-intrinsic.md: Likewise.
23888 * config/nds32/nds32_intrinsic.h: Likewise.
23889 * config/nds32/nds32.h (nds32_builtins): Likewise.
23891 2018-04-07 Jakub Jelinek <jakub@redhat.com>
23893 PR tree-optimization/85257
23894 * fold-const.c (native_encode_vector): If not all elts could fit
23895 and off is -1, return 0 rather than offset.
23896 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
23897 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
23898 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
23899 adjust buffer in native_interpret_expr call.
23901 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
23903 * config/nds32/constants.md (unspec_volatile_element): Add cache
23904 control enum values.
23905 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
23906 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
23907 * config/nds32/nds32.c (nds32_cctl_names): New.
23908 (nds32_print_operand): Handle cache control register names.
23909 * config/nds32/nds32.h (nds32_builtins): New enum values.
23910 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
23912 * config/nds32/nds32.md (type): Add mmu.
23913 * config/nds32/pipelines.md (simple_insn): Add mmu.
23915 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
23917 * config/nds32/nds32.md (type): Remove call.
23918 * config/nds32/pipelines.md (simple_insn): Likewise.
23920 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
23922 * config/nds32/constants.md (unspec_volatile_element): Add
23923 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
23924 UNSPEC_VOLATILE_FMFCFG.
23925 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
23926 description for fmfcfg and fmfcsr.
23927 (bdesc_1arg): Add fmtcsr.
23928 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
23929 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
23930 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
23931 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
23932 unspec_fmfcfg): New patterns.
23933 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
23934 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
23935 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
23936 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
23937 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
23938 __nds32__fmfcfg): Define.
23940 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
23942 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
23943 intrinsic register names.
23944 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
23945 intrinsic register enum values and macros.
23947 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
23949 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
23950 for load/store addressing form.
23951 (nds32_print_operand_address): Likewise.
23953 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
23956 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
23957 based on LABEL_REF. Remove useless assertion.
23958 (pic_address_needs_scratch): Fix formatting.
23959 (sparc_legitimize_pic_address): Minor tweaks.
23960 (sparc_delegitimize_address): Adjust assertion accordingly.
23961 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
23962 into symbolic_operand.
23963 (movsi_high_pic_label_ref): Likewise.
23964 (movsi_lo_sum_pic_label_ref): Likewise.
23965 (movdi_pic_label_ref): Likewise.
23966 (movdi_high_pic_label_ref): Likewise.
23967 (movdi_lo_sum_pic_label_ref): Likewise.
23969 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
23971 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
23972 custom LIB_SPEC setup.
23974 2018-04-06 Ruslan Bukin <br@bsdpad.com>
23975 Kito Cheng <kito.cheng@gmail.com>
23977 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
23978 * config/riscv/freebsd.h: New.
23980 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
23982 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
23983 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
23986 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
23987 Kito Cheng <kito.cheng@gmail.com>
23989 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
23990 nds32_output_call, nds32_symbol_binds_local_p): New functions.
23991 * config/nds32/nds32-protos.h (nds32_output_call,
23992 nds32_output_return): Declare.
23993 * config/nds32/nds32.md: Refine all the call and return patterns.
23995 2018-04-06 Jakub Jelinek <jakub@redhat.com>
23998 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
23999 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
24001 PR rtl-optimization/84872
24002 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
24003 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
24004 EDGE_CROSSING edge.
24006 2018-04-06 Tamar Christina <tamar.christina@arm.com>
24008 * expr.c (copy_blkmode_to_reg): Revert 254862.
24009 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
24011 2018-04-06 Richard Biener <rguenther@suse.de>
24013 PR middle-end/85244
24014 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
24015 after seeing a component reference with an adjacent field. Treat
24016 refs to arrays at struct end of external decls similar to
24017 refs to unconstrained commons.
24019 2018-04-06 Jakub Jelinek <jakub@redhat.com>
24022 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
24023 look through SAVE_EXPRs with non-side-effects argument. Adjust
24025 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
24028 2018-04-06 Richard Biener <rguenther@suse.de>
24030 PR middle-end/85180
24031 * alias.c (find_base_term): New wrapper around find_base_term
24032 unwinding CSELIB_VAL_PTR changes.
24033 (find_base_term): Do not restore CSELIB_VAL_PTR during the
24036 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
24038 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
24040 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
24041 constant definitions.
24042 ("nop"): lr 0,0 -> nopr r0
24043 ("nop_lr0", "nop_lr1"): New insn definitions.
24045 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
24047 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
24048 NDS32_V3PUSH_AVAILABLE_P macro.
24050 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
24051 Chung-Ju Wu <jasonwucj@gmail.com>
24053 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
24054 (nds32*-*-*): Add float and fpu_config into supported_defaults.
24055 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
24056 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
24057 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
24058 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
24059 * config/nds32/constraints.md: New constraints and checking for hard
24060 float configuration.
24061 * config/nds32/iterators.md: New mode iterator and attribute for hard
24062 float configuration.
24063 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
24065 * config/nds32/nds32-fpu.md: New file.
24066 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
24067 deal with hard float code generation.
24068 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
24070 (abi_type, float_reg_number): New enum type.
24071 * config/nds32/nds32-predicates.c: New predicates for hard float.
24072 * config/nds32/nds32-protos.h: Declare functions for hard float.
24073 * config/nds32/nds32.c: Implementation for hard float configuration.
24074 * config/nds32/nds32.h: Definitions for hard float configuration.
24075 * config/nds32/nds32.md: Include hard float machine description and
24076 modify patterns for hard float configuration.
24077 * config/nds32/nds32.opt: New options for hard float configuration.
24078 * config/nds32/predicates.md: New predicates for hard float
24081 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
24083 * common/config/nds32/nds32-common.c
24084 (nds32_option_optimization_table): Enable -mreleax-hint by default.
24086 2018-04-05 Jakub Jelinek <jakub@redhat.com>
24088 PR middle-end/85195
24089 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
24090 CONSTRUCTOR_ELT (ctor, ...)->value.
24092 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
24095 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
24097 2018-04-05 Tom de Vries <tom@codesourcery.com>
24100 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
24103 2018-04-05 Shiva Chen <shiva0217@gmail.com>
24104 Kito Cheng <kito.cheng@gmail.com>
24106 * config/nds32/constraints.md (U33): Fine-tune checking condition.
24107 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
24108 * config/nds32/nds32.h (nds32_16bit_address_type): Add
24109 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
24111 2018-04-05 Shiva Chen <shiva0217@gmail.com>
24112 Kito Cheng <kito.cheng@gmail.com>
24114 * config/nds32/constraints.md (Ufe): New memory constraint.
24115 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
24116 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
24117 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
24119 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
24120 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
24122 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
24124 * config/nds32/nds32.md: Use optimize_size in the condition for
24125 alu-shift instructions.
24127 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
24129 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
24131 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
24133 * config/nds32/nds32.md (negsi2): Refine pattern.
24135 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
24136 Chung-Ju Wu <jasonwucj@gmail.com>
24138 * config/nds32/iterators.md (shift_rotate): New code iterator.
24139 (shift): New code attribute.
24140 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
24141 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
24142 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
24143 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
24144 bit-wise operations.
24145 (andsi3, *andsi3): Ditto.
24146 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
24147 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
24148 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
24149 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
24150 nds32_ior_operand, nds32_xor_operand): New predicates.
24152 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
24154 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
24155 (addsi3, subsi3): ... this.
24157 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
24159 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
24161 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
24163 * config/nds32/nds32.md: Adjust indention.
24165 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
24167 * config/nds32/nds32.md (feature): New attribute.
24169 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
24171 * config/nds32/nds32.md (subtype): New attribute.
24173 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
24176 * config/arm/arm-builtins.c (arm_expand_builtin): Change
24177 expansion to perform a bitwise AND of the argument followed by a
24178 boolean negation of the result.
24180 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
24182 PR rtl-optimization/84878
24183 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
24184 the basic block. Assert the use reference is not artificial and that
24185 it has an associated insn.
24187 2018-04-04 Michael Matz <matz@suse.de>
24189 * builtins.c (compute_objsize): Pass correct operand
24190 to array_at_struct_end_p.
24192 2018-04-04 Richard Biener <rguenther@suse.de>
24195 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
24196 from contexts for DINFO_LEVEL_TERSE and below.
24198 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
24200 * config/nds32/nds32-doubleword.md (move_<mode>): Require
24201 resiter_operand condition.
24202 * config/nds32/nds32.md (*move<mode>): Ditto.
24204 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
24205 Monk Chiang <sh.chiang04@gmail.com>
24207 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
24209 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
24211 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
24213 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
24214 Kito Cheng <kito.cheng@gmail.com>
24216 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
24217 nds32_cond_code_str, output_cond_branch,
24218 output_cond_branch_compare_zero, nds32_expand_cbranch,
24219 nds32_expand_cstore, nds32_expand_movcc,
24220 nds32_output_cbranchsi4_equality_zero,
24221 nds32_output_cbranchsi4_equality_reg,
24222 nds32_output_cbranchsi4_equality_reg_or_const_int,
24223 nds32_output_cbranchsi4_greater_less_zero: New functions.
24224 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
24225 nds32_expand_cstore, nds32_expand_movcc,
24226 nds32_output_cbranchsi4_equality_zero,
24227 nds32_output_cbranchsi4_equality_reg,
24228 nds32_output_cbranchsi4_equality_reg_or_const_int,
24229 nds32_output_cbranchsi4_greater_less_zero): Declare.
24230 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
24231 nds32_rimm11s_operand): New predicates.
24232 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
24233 * config/nds32/nds32.md: Rewrite all the branch and conditional move
24236 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
24238 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
24239 * config/nds32/nds32.md: Ditto.
24240 * config/nds32/pipelines.md: Ditto.
24242 2018-04-04 Richard Biener <rguenther@suse.de>
24244 PR tree-optimization/85168
24245 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
24246 propagating abnormals.
24248 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
24250 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
24252 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
24253 Kito Cheng <kito.cheng@gmail.com>
24255 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
24256 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
24257 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
24258 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
24259 * config/nds32/nds32.md (sibcall_internal): New.
24260 (sibcall_register): Remove.
24261 (sibcall_immediate): Remove.
24262 (sibcall_value_internal): New.
24263 (sibcall_value_register): Remove.
24264 (sibcall_value_immediate): Remove.
24265 * config/nds32/predicates.md (nds32_general_register_operand): New.
24266 (nds32_call_address_operand): New.
24268 2018-04-03 Jakub Jelinek <jakub@redhat.com>
24270 PR rtl-optimization/85167
24271 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
24272 bb_defs if *split_p, instead preinitialize it to NULL.
24274 PR tree-optimization/85156
24275 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
24276 evaluating the argument multiple times.
24278 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
24280 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
24282 (_mm_cvtpd_ps): Likewise.
24283 (_mm_cvttpd_epi32): Likewise.
24284 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
24285 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
24286 vector, pixel, and bool following altivec.h include.
24288 2018-04-03 Martin Sebor <msebor@redhat.com>
24290 * doc/extend.texi (Common Function Attributes): Clarify.
24291 (const attribute): Likewise.
24292 (pure attribute): Likewise.
24294 2018-04-03 Jakub Jelinek <jakub@redhat.com>
24297 * config/i386/i386.c (ix86_expand_vector_set): Use
24298 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
24300 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
24302 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
24303 instructions when changing rounding bits to preserve precision bits
24304 in the x87 control word.
24306 2018-04-03 Martin Liska <mliska@suse.cz>
24308 PR tree-optimization/82491
24309 * rtl.h (strip_offset_and_add): Replace += suboffset with
24310 poly_uint64 () + suboffset.
24312 2018-03-29 Martin Liska <mliska@suse.cz>
24313 Martin Jambor <mjambor@suse.cz>
24316 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
24317 param_type is not an integral or pointer type.
24319 2018-04-03 Richard Biener <rguenther@suse.de>
24321 * sese.h (recompute_all_dominators): Remove.
24323 2018-04-02 Martin Sebor <msebor@redhat.com>
24325 * doc/invoke.texi (-Wrestrict): Fix typos.
24327 2018-04-02 Jim Wilson <jimw@sifive.com>
24329 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
24330 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
24331 (<optab>di3, <optab>si3_extend): Likewise.
24332 (<optab>si3_mask, <optab>si3_mask_1): New.
24333 (<optab>di3_mask, <optab>di3_mask_1): New.
24334 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
24335 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
24336 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
24338 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
24340 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
24343 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
24345 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
24346 (nds32_canonicalize_comparison): New function.
24348 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
24349 Kito Cheng <kito.cheng@gmail.com>
24350 Kuan-Lin Chen <kuanlinchentw@gmail.com>
24352 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
24353 * config/nds32/constants.md (unspec_volatile_element): Add
24354 UNSPEC_VOLATILE_RELAX_GROUP.
24355 * config/nds32/nds32-relax-opt.c: New file.
24356 * config/nds32/nds32-predicates.c
24357 (nds32_symbol_load_store_p): New function.
24358 * config/nds32/nds32-protos.h
24359 (nds32_symbol_load_store_p): Declare function.
24360 (make_pass_nds32_relax_opt): Declare new rtl pass function.
24361 * config/nds32/nds32.c
24362 (nds32_register_pass): New function to register pass.
24363 (nds32_register_passes): New function to register passes.
24364 * config/nds32/nds32.md (relax_group): New pattern.
24365 * config/nds32/nds32.opt (mrelax-hint): New option.
24366 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
24368 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
24370 * config/nds32/t-nds32: Modify files dependency.
24372 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
24374 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
24375 (PROFILE_HOOK): Define its implementation.
24377 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
24379 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
24380 type and 32-bit size.
24382 2018-04-01 Jakub Jelinek <jakub@redhat.com>
24384 PR middle-end/85090
24385 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
24386 (V_128_256): New mode iterator.
24387 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
24388 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
24389 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
24391 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
24392 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
24394 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
24397 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
24398 NaN inputs correctly.
24400 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
24403 * config/rs6000/vsx.md (??r): New mode attribute.
24404 (*vsx_mov<mode>_64bit): Use it.
24405 (*vsx_mov<mode>_32bit): Likewise.
24407 2018-03-30 Martin Sebor <msebor@redhat.com>
24409 PR tree-optimization/84818
24410 * builtins.c (check_access): Use warning_n.
24412 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
24415 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
24417 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
24420 2018-03-30 Julia Koval <julia.koval@intel.com>
24423 * x86-tune.def (movx, partial_reg_dependency): Enable for
24426 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
24428 PR inline-asm/84985
24429 * lra-constraints.c (process_alt_operands): Move setting
24430 this_alternative_matches below.
24432 2018-03-29 Martin Liska <mliska@suse.cz>
24435 * doc/invoke.texi: Document how LTO works with debug info.
24436 Describe auto-load support of binutils. Mention 'x86-64'
24437 as valid option value of -march option.
24439 2018-03-29 Jakub Jelinek <jakub@redhat.com>
24441 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
24444 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
24445 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
24446 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
24449 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
24452 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
24453 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
24454 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
24455 for RS6000_BTM_POWERPC64.
24456 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
24457 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
24458 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
24461 (DIVDEU): Likewise.
24463 2018-03-28 Carl Love <cel@us.ibm.com>
24466 2017-09-27 Carl Love <cel@us.ibm.com>
24468 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
24469 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
24470 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
24473 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24475 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
24476 instead of __vector bool.
24477 (_mm_max_pu8): Likewise.
24478 (_mm_min_pi16): Likewise.
24480 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
24483 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
24484 (DIVWEUO): Likewise.
24485 (DIVDEO): Likewise.
24486 (DIVDEUO): Likewise.
24487 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
24488 DIVWEUO and DIVDEUO.
24489 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
24490 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
24491 (div_extend): Likewise.
24492 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
24494 (__builtin_divweuo): Likewise.
24495 (__builtin_divdeo): Likewise.
24496 (__builtin_divdeuo): Likewise.
24498 2018-03-28 Jakub Jelinek <jakub@redhat.com>
24501 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
24502 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
24504 PR tree-optimization/82004
24505 * gimple-match-head.c (optimize_pow_to_exp): New function.
24506 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
24507 Don't fold to exp if optimize_pow_to_exp is false.
24509 2018-03-28 Martin Liska <mliska@suse.cz>
24512 * calls.c (initialize_argument_information): Fix trailing space.
24513 * common.opt: Fix typo and provide better explanation for
24514 -fsanitize-coverage option.
24515 * config/i386/i386.opt: Fix typo.
24517 2018-03-28 Jakub Jelinek <jakub@redhat.com>
24518 Martin Liska <mliska@suse.cz>
24521 * gimplify.c (asan_poison_variable): Don't do the check for
24522 gimplify_omp_ctxp here.
24523 (gimplify_decl_expr): Do it here.
24524 (gimplify_target_expr): Likewise.
24526 2018-03-28 Martin Liska <mliska@suse.cz>
24529 * config/i386/i386.c (ix86_function_arg_advance): Do not call
24530 chkp_type_bounds_count if MPX is not enabled.
24532 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
24534 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
24536 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
24539 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
24540 function to create the function decl for complex long double
24541 multiply and divide for -mabi=ieeelongdouble.
24542 (init_float128_ieee): Call it.
24544 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
24547 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
24548 -fcf-protection=branch -mibt.
24549 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
24551 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
24554 * config/arm/arm.c (arm_valid_symbolic_address): Handle
24555 arm_word_relocations.
24557 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
24560 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
24561 extern array declarations.
24563 2018-03-27 Richard Biener <rguenther@suse.de>
24565 PR middle-end/84067
24566 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
24567 explicit single_use checks.
24569 2018-03-27 Richard Biener <rguenther@suse.de>
24571 PR tree-optimization/85082
24572 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
24575 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
24577 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
24578 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
24579 Turn on fasynchronous-unwind-tables and funwind-tables.
24581 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
24584 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
24585 (*bmi_blsr_<mode>_ccz): Ditto.
24587 2018-03-26 Tom de Vries <tom@codesourcery.com>
24589 PR tree-optimization/85063
24590 * omp-general.c (offloading_function_p): New function. Factor out
24592 * omp-offload.c (pass_omp_target_link::gate): ... here.
24593 * omp-general.h (offloading_function_p): Declare.
24594 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
24595 with attribute omp declare target for offloading functions.
24597 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
24599 PR tree-optimization/84005
24600 * tree-data-ref.h (get_base_for_alignment): Declare.
24601 * tree-data-ref.c (get_base_for_alignment_1): New function.
24602 (get_base_for_alignment): Likewise.
24603 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
24604 get_base_for_alignment to find a suitable base object, instead
24605 of always using drb->base_address.
24607 2018-03-23 Jakub Jelinek <jakub@redhat.com>
24609 PR inline-asm/85022
24610 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
24611 known size by default.
24613 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
24615 PR inline-asm/85030
24616 * lra-constraints.c (process_alt_operands): Don't match BLKmode
24617 and non BLKmode operands.
24619 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24622 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
24623 Clean up attributes.
24625 2018-03-23 Richard Biener <rguenther@suse.de>
24628 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
24629 we are going to emit early debug for LTO.
24631 2018-03-23 Jakub Jelinek <jakub@redhat.com>
24633 PR inline-asm/85034
24634 * function.c (match_asm_constraints_1): Don't optimize if input
24635 doesn't satisfy general_operand predicate for output's mode.
24637 PR inline-asm/85022
24638 * alias.c (write_dependence_p): Don't require for x_canonicalized
24639 non-VOIDmode if x has VOIDmode.
24642 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
24643 just don't try to optimize it rather than assert it never happens.
24645 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
24647 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
24648 macro expansions for definition of ST_INTERNAL_<mode> and
24649 LD_INTERNAL_<mode> builtins.
24650 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
24652 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
24654 (altivec_expand_st_builtin): Likewise.
24655 (altivec_expand_builtin): Remove calls to deleted functions.
24656 (rs6000_address_for_altivec): Delete this function.
24657 * config/rs6000/vector.md: Remove expands for
24658 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
24660 2018-03-22 Sudakshina Das <sudi.das@arm.com>
24663 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
24664 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
24665 re-computing once computed.
24666 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
24667 (arm_init_machine_status): Initialize
24668 machine->static_chain_stack_bytes.
24670 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
24673 * doc/extend.texi: Add four new prototypes for vec_ld.
24674 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
24675 definitions for more logical presentation.
24676 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
24677 entries for V1TI variants of __builtin_altivec_ld builtin.
24678 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
24679 handling of V1TI variant of LVX icode pattern.
24680 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
24681 (rs6000_gimple_fold_builtin): Likewise.
24682 (altivec_init_builtins): Add code to define
24683 __builtin_altivec_lvx_v1ti function.
24685 2018-03-22 Jakub Jelinek <jakub@redhat.com>
24687 PR inline-asm/84941
24688 * function.c (match_asm_constraints_1): Don't do the optimization
24689 if input isn't a REG, SUBREG, MEM or constant.
24691 2018-03-22 Tom de Vries <tom@codesourcery.com>
24693 PR tree-optimization/84956
24694 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
24695 bb_has_abnormal_pred.
24697 2018-03-22 Jakub Jelinek <jakub@redhat.com>
24700 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
24701 DECL_INITIAL (decl) to decl at the end.
24702 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
24703 adjust the comment.
24705 2018-03-21 Joseph Myers <joseph@codesourcery.com>
24707 * doc/extend.texi (__builtin_tgmath): Document when complex
24708 integer types are treated as _Complex _Float64.
24710 2018-03-21 Tom de Vries <tom@codesourcery.com>
24712 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
24714 2018-03-21 Jakub Jelinek <jakub@redhat.com>
24716 PR tree-optimization/84960
24717 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
24718 if it is ENTRY block, move them into single succ of ENTRY in that case.
24720 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
24722 PR tree-optimization/84811
24723 * poly-int.h (poly_span_traits): Remove the T3 parameter and
24724 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
24725 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
24726 (known_subrange_p): Update accordingly. Cast each value involved
24727 in the size comparison, rather than casting the result of the
24730 2018-03-21 Jakub Jelinek <jakub@redhat.com>
24732 PR tree-optimization/84982
24733 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
24734 by flipping the least significant bit rather than all bits from
24735 bitpos to bitpos + bitsize - 1.
24737 2018-03-21 Nathan Sidwell <nathan@acm.org>
24739 * doc/extend.texi (Deprecated Features): Remove mention of
24740 long-deleted deprecations.
24742 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24745 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
24746 * configure: Regenerate.
24748 2018-03-21 Tom de Vries <tom@codesourcery.com>
24750 PR tree-optimization/83126
24751 * tree-parloops.c (num_phis): New function.
24752 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
24754 2018-03-21 Nathan Sidwell <nathan@acm.org>
24756 * doc/extend.texi (Deprecated Features): Update deprecated flags,
24757 mention anon-struct/union members and trailing attributes.
24759 2018-03-21 Bin Cheng <bin.cheng@arm.com>
24761 PR tree-optimization/84969
24762 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
24763 builtin memset partitions if they set different rhs values.
24765 2018-03-21 Jakub Jelinek <jakub@redhat.com>
24767 PR rtl-optimization/84989
24768 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
24769 VEC_DUPLICATE with scalar result mode.
24771 2018-03-21 Martin Liska <mliska@suse.cz>
24774 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
24775 not intended return statement.
24777 2018-03-21 Martin Liska <mliska@suse.cz>
24780 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
24781 (chkp_find_bound_slots_1): Limit number of iterations.
24783 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
24786 * Minor grammar fixes for x86 options.
24788 2018-03-20 Jakub Jelinek <jakub@redhat.com>
24791 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
24792 holding REG_CFA_RESTORE notes, instead turn them into a USE.
24794 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
24797 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
24798 (altivec_lvx_<mode>_1op): Likewise.
24799 (altivec_stvx_<mode>_2op): Likewise.
24800 (altivec_stvx_<mode>_1op): Likewise.
24801 (altivec_lvx_<VM2:mode>): New define_expand.
24802 (altivec_stvx_<VM2:mode>): Likewise.
24803 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
24804 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
24805 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
24806 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
24807 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
24808 (rs6000_gen_lvx): Likewise.
24809 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
24810 (altivec_expand_stv_builtin): Likewise.
24811 (altivec_expand_builtin): Likewise.
24812 * config/rs6000/vector.md: Likewise.
24814 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24817 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
24820 2018-03-20 Richard Biener <rguenther@suse.de>
24823 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
24824 sign-conversions as zero, fall back to standard scalar_stmt
24827 2018-03-20 Martin Liska <mliska@suse.cz>
24830 * predict.c (rebuild_frequencies): Handle case when we have
24831 PROFILE_ABSENT, but flag_guess_branch_prob is false.
24833 2018-03-20 Jakub Jelinek <jakub@redhat.com>
24836 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
24837 flag_section_anchors.
24838 * varasm.c (use_blocks_for_decl_p): Remove hack for
24839 dw2_force_const_mem.
24842 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
24844 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
24845 be created, use lowpart_subreg of operands[0] rather than operands[0]
24847 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
24848 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
24849 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
24850 and n constraint instead of aarch64_shift_imm_di and Usd.
24851 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
24852 (*aarch64_<optab>_reg_minus<mode>3): ... this.
24854 2018-03-20 Sudakshina Das <sudi.das@arm.com>
24857 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
24858 to favor GPR over NEON registers.
24859 (<shift>di3_neon): Likewise.
24861 2018-03-20 Tom de Vries <tom@codesourcery.com>
24864 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
24865 (nvptx_process_pars): Emit bar.sync asap and alap.
24867 2018-03-20 Tom de Vries <tom@codesourcery.com>
24870 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
24871 seen_label if seen_label is already set.
24873 2018-03-20 Jakub Jelinek <jakub@redhat.com>
24876 * config/i386/i386.c (fold_builtin_cpu): For features above 31
24877 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
24878 Use 1U instead of 1. Formatting fixes.
24881 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
24882 instead of TREE_TYPE (s1) for the return value.
24884 2018-03-19 Jakub Jelinek <jakub@redhat.com>
24886 PR tree-optimization/84946
24887 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
24888 bitsize + bitsize in poly_uint64 rather than poly_int64.
24891 * dwarf2asm.c: Include fold-const.c.
24892 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
24893 of decl rather than decl itself.
24895 PR rtl-optimization/84643
24896 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
24898 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
24901 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
24902 calling assemble_variable.
24904 2018-03-19 Sudakshina Das <sudi.das@arm.com>
24907 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
24908 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
24910 2018-03-19 Jim Wilson <jimw@sifive.com>
24913 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
24914 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
24915 (riscv_first_stack_step): Likewise.
24916 (riscv_option_override): Use STACK_BOUNDARY instead of
24917 MIN_STACK_BOUNDARY.
24918 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
24919 MIN_STACK_BOUNDARY.
24920 (BIGGEST_ALIGNMENT): Set to 128.
24921 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
24922 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
24925 2018-03-19 Richard Biener <rguenther@suse.de>
24927 PR tree-optimization/84933
24928 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
24929 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
24931 2018-03-19 Richard Biener <rguenther@suse.de>
24933 PR tree-optimization/84859
24934 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
24935 (cond_if_else_store_replacement): Perform sinking operation on
24936 single-store BBs regardless of MAX_STORES_TO_SINK setting.
24937 Generalize what a BB with a single eligible store is.
24939 2018-03-19 Richard Biener <rguenther@suse.de>
24941 PR tree-optimization/84929
24942 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
24943 chrec_is_positive against non-chrec arg.
24945 2018-03-19 Tamar Christina <tamar.christina@arm.com>
24948 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
24950 2018-03-18 Martin Liska <mliska@suse.cz>
24952 PR rtl-optimization/84635
24953 * regrename.c (build_def_use): Use matches_mode only when
24956 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
24958 PR tree-optimization/84913
24959 * tree-vect-loop.c (vectorizable_reduction): Don't try to
24960 vectorize chains of COND_EXPRs.
24962 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
24964 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
24966 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
24968 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
24970 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
24972 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
24974 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
24975 Kito Cheng <kito.cheng@gmail.com>
24977 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
24978 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
24979 (nds32_adjust_reg_alloc_order): New function.
24980 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
24982 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
24984 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
24985 nds32_print_operand, nds32_print_operand_address): Use
24986 HOST_WIDE_INT_PRINT_DEC instead.
24988 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
24990 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
24992 2018-03-17 Jakub Jelinek <jakub@redhat.com>
24995 * config/i386/i386.c (initial_ix86_tune_features,
24996 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
24997 unsigned long long.
24998 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
24999 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
25000 rather than 1u << ix86_tune. Formatting fix.
25001 (ix86_option_override_internal): Change ix86_arch_mask from
25002 unsigned int to unsigned HOST_WIDE_INT, initialize to
25003 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
25004 (ix86_function_specific_restore): Likewise.
25006 2018-03-16 Jakub Jelinek <jakub@redhat.com>
25009 * postreload.c (reload_combine_recognize_pattern): Perform
25010 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
25011 truncate_int_for_mode the result for the destination's mode.
25014 * hsa-gen.c (mem_type_for_type): Fix comment typo.
25015 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
25017 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
25020 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
25023 * lra-assigns.c (lra_split_hard_reg_for): Don't use
25024 regno_allocno_class_array and sorted_pseudos.
25025 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
25026 insns where regno is used.
25028 2018-03-16 Martin Liska <mliska@suse.cz>
25031 * multiple_target.c (create_dispatcher_calls): Redirect
25032 reference in the symbol table.
25034 2018-03-16 Martin Liska <mliska@suse.cz>
25037 * multiple_target.c (create_dispatcher_calls): Redirect also
25040 2018-03-16 Jakub Jelinek <jakub@redhat.com>
25044 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
25045 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
25046 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
25048 2018-03-16 Julia Koval <julia.koval@intel.com>
25050 * doc/invoke.texi (Skylake Server): Add CLWB.
25051 Cannonlake): Remove CLWB.
25053 2018-03-16 Jakub Jelinek <jakub@redhat.com>
25055 PR tree-optimization/84841
25056 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
25058 (FLOAT_ONE_CONST_TYPE): Define.
25059 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
25060 (sort_by_operand_rank): Put entries with higher constant_type last
25061 rather than first to match comments.
25063 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
25065 * config/nios2/nios2.md (movsi_internal): Fix thinko in
25068 2018-03-15 Jakub Jelinek <jakub@redhat.com>
25071 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
25072 check and use address of target always.
25074 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
25077 * config/i386/i386.c (indirect_thunk_needed): Update comments.
25078 (indirect_thunk_bnd_needed): Likewise.
25079 (indirect_thunks_used): Likewise.
25080 (indirect_thunks_bnd_used): Likewise.
25081 (indirect_return_needed): New.
25082 (indirect_return_bnd_needed): Likewise.
25083 (output_indirect_thunk_function): Add a bool argument for
25085 (output_indirect_thunk_function): Don't generate alias for
25086 function return thunk.
25087 (ix86_code_end): Call output_indirect_thunk_function to generate
25088 function return thunks.
25089 (ix86_output_function_return): Set indirect_return_bnd_needed
25090 and indirect_return_needed instead of indirect_thunk_bnd_needed
25091 and indirect_thunk_needed.
25093 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
25095 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
25096 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
25097 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
25099 2018-03-15 David Malcolm <dmalcolm@redhat.com>
25100 Paul Hua <paul.hua.gm@gmail.com>
25103 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
25105 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
25107 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
25108 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
25109 resp. SFmode cases.
25111 2018-03-15 Tamar Christina <tamar.christina@arm.com>
25114 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
25115 instead of GET_MODE_SIZE when comparing Units.
25117 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
25120 * varasm.c (hash_section): Return an unchangeble hash value
25121 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
25122 Return !aarch64_can_use_per_function_literal_pools_p ().
25124 2018-03-15 Jakub Jelinek <jakub@redhat.com>
25127 * optabs.c (emit_conditional_move): Pass address of cmode's copy
25128 rather than address of cmode as last argument to prepare_cmp_insn.
25130 2018-03-15 Julia Koval <julia.koval@intel.com>
25132 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
25133 F_AVX512VNNI, F_AVX512BITALG): New.
25135 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
25138 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
25139 insn for floating-point loads and stores.
25141 2018-03-14 Carl Love <cel@us.ibm.com>
25143 * config/rs6000/rs6000-c.c: Add macro definitions for
25144 ALTIVEC_BUILTIN_VEC_PERMXOR.
25145 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
25146 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
25147 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
25148 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
25150 * config/doc/extend.texi: Add prototypes for vec_permxor.
25152 2018-03-14 David Malcolm <dmalcolm@redhat.com>
25155 * diagnostic-show-locus.c (class layout_point): Convert m_line
25156 from int to linenum_type.
25157 (line_span::comparator): Use linenum "compare" function when
25158 comparing line numbers.
25159 (test_line_span): New function.
25160 (layout_range::contains_point): Convert param "row" from int to
25162 (layout_range::intersects_line_p): Likewise.
25163 (layout::will_show_line_p): Likewise.
25164 (layout::print_source_line): Likewise.
25165 (layout::should_print_annotation_line_p): Likewise.
25166 (layout::print_annotation_line): Likewise.
25167 (layout::print_leading_fixits): Likewise.
25168 (layout::annotation_line_showed_range_p): Likewise.
25169 (struct line_corrections): Likewise for field m_row.
25170 (line_corrections::line_corrections): Likewise for param "row".
25171 (layout::print_trailing_fixits): Likewise.
25172 (layout::get_state_at_point): Likewise.
25173 (layout::get_x_bound_for_row): Likewise.
25174 (layout::print_line): Likewise.
25175 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
25176 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
25177 * input.c (selftest::test_linenum_comparisons): New function.
25178 (selftest::input_c_tests): Call it.
25179 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
25180 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
25181 * selftest.h (ASSERT_GT): New macro.
25182 (ASSERT_GT_AT): New macro.
25183 (ASSERT_LT): New macro.
25184 (ASSERT_LT_AT): New macro.
25186 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
25188 PR rtl-optimization/84780
25189 * combine.c (distribute_links): Don't make a link based on pc_rtx.
25191 2018-03-14 Martin Liska <mliska@suse.cz>
25193 * tree.c (record_node_allocation_statistics): Use
25194 get_stats_node_kind.
25195 (get_stats_node_kind): New function extracted from
25196 record_node_allocation_statistics.
25197 (free_node): Use get_stats_node_kind.
25199 2018-03-14 Richard Biener <rguenther@suse.de>
25201 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
25202 that the value-set of ANTIC_IN doesn't grow.
25205 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
25207 (BB_VISITED_WITH_VISITED_SUCCS): New define.
25208 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
25210 2018-03-14 Julia Koval <julia.koval@intel.com>
25212 * config.gcc (icelake-client, icelake-server): New.
25214 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
25215 (initial_ix86_arch_features): Ditto.
25216 (PTA_SKYLAKE): Add SGX.
25217 (PTA_ICELAKE): Remove.
25218 (PTA_ICELAKE_CLIENT): New.
25219 (PTA_ICELAKE_SERVER): New.
25220 (ix86_option_override_internal): Split up icelake on icelake client and
25222 (get_builtin_code_for_version): Ditto.
25223 (fold_builtin_cpu): Ditto.
25224 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
25225 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
25226 * config/i386/i386.h (processor_type): Ditto.
25227 * doc/invoke.texi: Ditto.
25229 2018-03-14 Jakub Jelinek <jakub@redhat.com>
25232 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
25233 INTEGER_CST offset, add it together with bitpos / 8 and
25234 sign extend based on POINTER_SIZE.
25238 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
25241 * config/i386/constraints.md (Yc): New register constraint.
25242 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
25243 Use Yc constraint for alternative 2 of operand 0. Remove
25244 preferred_for_speed attribute.
25246 2018-03-14 Richard Biener <rguenther@suse.de>
25248 PR tree-optimization/84830
25249 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
25250 with the old one to avoid oscillations.
25252 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
25255 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
25257 (assign_by_spills): Return a flag of reload assignment failure.
25258 Do not process the reload assignment failures. Do not spill other
25259 reload pseudos if they has the same reg class. Update n if
25261 (lra_assign): Add a return arg. Set up from the result of
25262 assign_by_spills call.
25263 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
25264 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
25265 usage_insns if it is not NULL.
25266 (spill_hard_reg_in_range): New function.
25267 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
25268 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
25269 function prototypes.
25270 (lra_assign): Change prototype.
25271 * lra.c (lra): Add code to deal with fails by splitting hard reg
25274 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
25276 * config/riscv/riscv.opt (mrelax): New option.
25277 * config/riscv/riscv.c (riscv_file_start): Emit ".option
25278 "norelax" when riscv_mrelax is disabled.
25279 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
25281 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25284 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
25285 reassociation for int modes.
25287 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
25289 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
25290 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
25292 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
25293 * config/aarch64/aarch64-sve.md
25294 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
25295 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
25296 (*extend<mode><Vwide>2): Rename to...
25297 (aarch64_sve_extend<mode><Vwide>2): ...this.
25298 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
25299 renaming the old pattern to...
25300 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
25302 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
25303 define_expand, renaming the old pattern to...
25304 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
25305 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
25306 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
25307 account when deciding which SVE instruction the optab should use.
25308 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
25310 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
25312 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
25313 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
25314 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
25315 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
25316 (tlsdesc_small_<mode>): Turn a define_expand and use
25317 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
25318 (tlsdesc_small_advsimd_<mode>): ...this.
25319 (tlsdesc_small_sve_<mode>): New pattern.
25321 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
25323 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
25324 (UNSPEC_UMUL_HIGHPART): New constants.
25325 (MUL_HIGHPART): New int iteraor.
25326 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
25327 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
25329 (*<su>mul<mode>3_highpart): New define_insn.
25331 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
25334 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
25337 2018-03-13 Martin Liska <mliska@suse.cz>
25340 * (sem_item_optimizer::sem_item_optimizer): Initialize new
25342 (sem_item_optimizer::~sem_item_optimizer): Release it.
25343 (sem_item_optimizer::merge_classes): Register variable aliases.
25344 (sem_item_optimizer::fixup_pt_set): New function.
25345 (sem_item_optimizer::fixup_points_to_sets): Likewise.
25346 * ipa-icf.h: Declare new variables and functions.
25348 2018-03-13 Jakub Jelinek <jakub@redhat.com>
25350 PR middle-end/84834
25351 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
25352 integer_pow2p@2 and test integer_pow2p in condition.
25353 (A < 0 ? C : 0): Similarly for @1.
25355 PR middle-end/84831
25356 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
25357 characters starting at p contain '\0' character, don't look beyond
25361 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
25362 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
25365 * reg-stack.c (change_stack): Change update_end var from int to
25366 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
25367 also call set_block_for_insn on the newly added insns and rescan.
25370 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
25371 on the last operand.
25374 * tree.c (stabilize_reference_1): Return save_expr (e) for
25375 STATEMENT_LIST even if it doesn't have side-effects.
25377 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
25379 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
25381 2018-03-12 Renlin Li <renlin.li@arm.com>
25383 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
25384 aarch64_output_scalar_simd_mov_immediate.
25386 2018-03-12 Martin Sebor <msebor@redhat.com>
25388 PR tree-optimization/83456
25389 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
25390 for perfectly overlapping calls to memcpy.
25391 (gimple_fold_builtin_memory_chk): Same.
25392 (gimple_fold_builtin_strcpy): Handle no-warning.
25393 (gimple_fold_builtin_stxcpy_chk): Same.
25394 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
25396 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
25398 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
25399 parameter. Use it for SFmode.
25400 (rs6000_function_arg_advance_1): Adjust.
25401 (rs6000_function_arg): Adjust.
25402 (rs6000_gimplify_va_arg): Pass false for that new parameter.
25404 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
25406 PR rtl-optimization/84169
25407 PR rtl-optimization/84780
25408 * combine.c (can_combine_p): Check for a 2-insn combination whether
25409 the destination register is used between the two insns, too.
25411 2018-03-12 Richard Biener <rguenther@suse.de>
25413 PR tree-optimization/84803
25414 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
25415 for refs DR analysis didn't process.
25417 2018-03-12 Richard Biener <rguenther@suse.de>
25419 PR tree-optimization/84777
25420 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
25421 force-vectorize loops ignore whether we are optimizing for size.
25423 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
25425 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
25426 (TARGET_MD_ASM_ADJUST): Define.
25428 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
25429 Kito Cheng <kito.cheng@gmail.com>
25430 Chung-Ju Wu <jasonwucj@gmail.com>
25432 * config/nds32/nds32.c (nds32_compute_stack_frame,
25433 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
25434 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
25435 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
25436 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
25437 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
25438 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
25439 * config/nds32/nds32.md (prologue, epilogue): Use macro
25440 NDS32_V3PUSH_AVAILABLE_P to do checking.
25442 2018-03-11 Jakub Jelinek <jakub@redhat.com>
25445 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
25446 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
25447 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
25448 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
25449 addition of most attributes on !orig_type_die or the attribute not
25450 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
25452 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
25453 Chung-Ju Wu <jasonwucj@gmail.com>
25455 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
25456 __NDS32_VH__ macro.
25457 * config/nds32/nds32.opt (mvh): New option.
25459 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
25460 Chung-Ju Wu <jasonwucj@gmail.com>
25462 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
25464 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
25465 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
25468 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
25469 Chung-Ju Wu <jasonwucj@gmail.com>
25471 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
25473 * config/nds32/nds32-multiple.md (strlensi): New pattern.
25474 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
25476 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
25477 Kito Cheng <kito.cheng@gmail.com>
25478 Chung-Ju Wu <jasonwucj@gmail.com>
25480 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
25481 UNSPEC_FFMISM and UNSPEC_FLMISM.
25482 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
25483 for ffb, ffmism and flmism.
25484 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
25485 (unspec_ffmism): Ditto.
25486 (unspec_flmism): Ditto.
25487 (nds32_expand_builtin_impl): Check if string extension is available.
25488 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
25489 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
25491 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
25494 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
25497 * lra-assigns.c (assign_by_spills): Return a flag of reload
25498 assignment failure. Do not process the reload assignment
25499 failures. Do not spill other reload pseudos if they has the same
25501 (lra_assign): Add a return arg. Set up from the result of
25502 assign_by_spills call.
25503 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
25504 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
25505 usage_insns if it is not NULL.
25506 (spill_hard_reg_in_range): New function.
25507 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
25508 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
25509 function prototypes.
25510 (lra_assign): Change prototype.
25511 * lra.c (lra): Add code to deal with fails by splitting hard reg
25514 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
25517 * config/i386/i386.opt: Replace Enforcment with Enforcement.
25519 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
25522 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
25523 (dw_val_node): Add val_symbolic_view.
25524 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
25525 (symview_upper_bound): New.
25526 (new_line_info_table): Initialize symviews_since_reset.
25527 (dwarf2out_source_line): Count symviews_since_reset and set
25528 symview_upper_bound.
25529 (dw_val_equal_p): Handle symview.
25530 (add_AT_symview): New.
25531 (print_dw_val): Handle symview.
25532 (attr_checksum, attr_checksum_ordered): Likewise.
25533 (same_dw_val_p, size_of_die): Likewise.
25534 (value_format, output_die): Likewise.
25535 (add_high_low_attributes): Use add_AT_symview for entry_view.
25536 (dwarf2out_finish): Reset symview_upper_bound, clear
25539 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
25542 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
25543 Add strict argument and use it.
25544 (rs6000_split_multireg_move): Update for new strict argument.
25545 (mem_operand_gpr): Disallow all non-offsettable addresses.
25546 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
25548 2018-03-09 Jakub Jelinek <jakub@redhat.com>
25551 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
25552 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
25553 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
25556 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
25557 decl, use remap_type if we want to use the type.
25559 2018-03-09 Martin Sebor <msebor@redhat.com>
25561 PR tree-optimization/84526
25562 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
25564 (builtin_access::generic_overlap): Be prepared to handle non-array
25567 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
25569 PR rtl-optimization/84682
25570 * lra-constraints.c (process_address_1): Check is_address flag
25571 for address constraints.
25572 (process_alt_operands): Likewise.
25573 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
25574 preprocess_constraints.
25575 * recog.h (preprocess_constraints): Add oploc parameter.
25577 * recog.c (preprocess_constraints): Test address_operand for
25578 CT_ADDRESS constraints.
25580 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
25583 * lra-assigns.c (assign_by_spills): Return a flag of reload
25584 assignment failure. Do not process the reload assignment
25585 failures. Do not spill other reload pseudos if they has the same
25587 (lra_assign): Add a return arg. Set up from the result of
25588 assign_by_spills call.
25589 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
25590 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
25591 usage_insns if it is not NULL.
25592 (spill_hard_reg_in_range): New function.
25593 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
25594 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
25595 function prototypes.
25596 (lra_assign): Change prototype.
25597 * lra.c (lra): Add code to deal with fails by splitting hard reg
25600 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25603 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
25604 Accept complain bool parameter. Only emit errors if it is true.
25605 (arm_parse_cpu_option_name): Likewise.
25606 (arm_target_thumb_only): Adjust callers of the above.
25607 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
25608 prototype to take a default true bool parameter.
25609 (arm_parse_arch_option_name): Likewise.
25611 2018-03-09 David Malcolm <dmalcolm@redhat.com>
25612 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
25616 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
25617 * configure: Regenerate.
25618 * configure.ac ("linker --version-script option"): New.
25619 ("linker soname option"): New.
25621 2018-03-09 Richard Biener <rguenther@suse.de>
25623 PR tree-optimization/84775
25624 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
25625 immediate uses of predicate stmts and mark them modified.
25628 PR tree-optimization/84178
25629 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
25631 (version_loop_for_if_conversion): Delay update_ssa call.
25632 (tree_if_conversion): Delay update_ssa until after predicate
25635 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
25638 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
25639 when the function accesses prior frames.
25641 2018-03-08 Jakub Jelinek <jakub@redhat.com>
25644 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
25645 gen_llsym, otherwise call maybe_gen_llsym.
25647 PR inline-asm/84742
25648 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
25649 has ',' character inside of it.
25651 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25654 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
25655 as clobbering CC_REGNUM.
25657 2018-03-08 Richard Biener <rguenther@suse.de>
25659 PR middle-end/84552
25660 * tree-scalar-evolution.c: Include tree-into-ssa.h.
25661 (follow_copies_to_constant): Do not follow SSA names registered
25664 2018-03-08 Richard Biener <rguenther@suse.de>
25666 PR tree-optimization/84178
25667 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
25669 (version_loop_for_if_conversion): Delay update_ssa call.
25670 (tree_if_conversion): Delay update_ssa until after predicate
25673 2018-03-08 David Malcolm <dmalcolm@redhat.com>
25675 PR tree-optimization/84178
25676 * tree-if-conv.c (release_bb_predicate): Remove the
25677 the assertion that the stmts have NULL use_ops.
25678 Discard the statements, asserting that they haven't
25679 yet been added to a BB.
25681 2018-03-08 Richard Biener <rguenther@suse.de>
25683 PR tree-optimization/84746
25684 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
25685 (phi_translate): Pass in destination ANTIC_OUT set.
25686 (phi_translate_1): Likewise. For a simplified result lookup
25687 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
25688 (phi_translate_set): Adjust.
25689 (do_pre_regular_insertion): Likewise.
25690 (do_pre_partial_partial_insertion): Likewise.
25692 2018-03-08 Martin Liska <mliska@suse.cz>
25694 PR gcov-profile/84735
25695 * doc/gcov.texi: Document usage of profile files.
25696 * gcov-io.h: Document changes in the format.
25698 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
25702 * dwarf2out.c (struct dw_line_info_table): Update comments for
25704 (FORCE_RESET_NEXT_VIEW): New.
25705 (FORCE_RESETTING_VIEW_P): New.
25706 (RESETTING_VIEW_P): Check for -1 too.
25707 (ZERO_VIEW_P): Likewise.
25708 (new_line_info_table): Force-reset next view.
25709 (dwarf2out_begin_function): Likewise.
25710 (dwarf2out_source_line): Simplify zero_view_p initialization.
25711 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
25712 view directly. Omit view when omitting .loc at line 0.
25714 2018-03-08 Jakub Jelinek <jakub@redhat.com>
25716 PR tree-optimization/84740
25717 * tree-switch-conversion.c (process_switch): Call build_constructors
25718 only if info.phi_count is non-zero.
25720 PR tree-optimization/84739
25721 * tree-tailcall.c (find_tail_calls): Check call arguments against
25722 DECL_ARGUMENTS (current_function_decl) rather than
25723 DECL_ARGUMENTS (func) when checking for tail recursion.
25725 2018-03-07 Jakub Jelinek <jakub@redhat.com>
25727 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
25728 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
25729 Volker Reichelt's entry and add entries for people that perform
25730 GCC fuzzy testing and report numerous bugs.
25732 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
25735 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
25736 readonly data in sdata, if that is disabled.
25737 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
25738 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
25739 -mreadonly-in-sdata option.
25741 2018-03-07 Martin Sebor <msebor@redhat.com>
25743 PR tree-optimization/84468
25744 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
25745 basic block when looking for nul assignment.
25747 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
25750 * except.h (output_function_exception_table): Adjust prototype.
25751 * except.c (output_function_exception_table): Remove FNNAME parameter
25752 and add SECTION parameter. Ouput one part of the table at a time.
25753 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
25754 the first part of the exception table and emit unwind directives.
25755 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
25756 (i386_pe_seh_cold_init): Likewise.
25757 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
25758 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
25759 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
25760 (ix86_output_call_insn): Emit a nop in one more case for SEH.
25761 * config/i386/winnt.c: Include except.h.
25762 (struct seh_frame_state): Add reg_offset, after_prologue and
25763 in_cold_section fields.
25764 (i386_pe_seh_end_prologue): Set seh->after_prologue.
25765 (i386_pe_seh_cold_init): New function.
25766 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
25767 to seh->in_cold_section.
25768 (seh_emit_push): Record the offset of the push.
25769 (seh_emit_save): Record the offet of the save.
25770 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
25771 Test seh->after_prologue to disregard the epilogue.
25772 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
25773 (i386_pe_end_cold_function): New function.
25775 2018-03-07 Jakub Jelinek <jakub@redhat.com>
25778 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
25779 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
25782 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
25784 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
25785 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
25787 PR middle-end/84723
25788 * multiple_target.c: Include tree-inline.h and intl.h.
25789 (expand_target_clones): Diagnose and fail if node->definition and
25790 !tree_versionable_function_p (node->decl).
25792 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
25794 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
25796 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
25797 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
25798 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
25800 2018-03-06 Jakub Jelinek <jakub@redhat.com>
25803 * combine.c (try_combine): Use reg_or_subregno instead of handling
25804 just paradoxical SUBREGs and REGs.
25806 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
25808 * config/arc/arc.c (arc_finalize_pic): Remove function.
25809 (arc_must_save_register): We use single base PIC register, remove
25810 checks to save/restore the PIC register.
25811 (arc_expand_prologue): Likewise.
25812 * config/arc/arc-protos.h (arc_set_default_type_attributes):
25814 (arc_verify_short): Likewise.
25815 (arc_attr_type): Likewise.
25816 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
25817 (walk_stores): Likewise.
25818 (arc_address_cost): Make it static.
25819 (arc_verify_short): Likewise.
25820 (branch_dest): Likewise.
25821 (arc_attr_type): Likewise.
25822 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
25823 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
25824 (arc_final_prescan_insn): Remove inserting the nops due to
25825 hardware hazards. It is done in reorg step.
25826 (insn_length_variant_t): Remove.
25827 (insn_length_parameters_t): Likewise.
25828 (arc_insn_length_parameters): Likewise.
25829 (arc_get_insn_variants): Likewise.
25830 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
25832 2018-03-06 Jakub Jelinek <jakub@redhat.com>
25834 PR inline-asm/84683
25835 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
25838 PR tree-optimization/84687
25839 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
25841 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
25843 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25845 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
25846 Rename to ppc_speculation_barrier.
25847 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
25848 __builtin_ppc_speculation_barrier.
25850 2018-03-05 Jakub Jelinek <jakub@redhat.com>
25853 * combine.c (combine_simplify_rtx): Don't try to simplify if
25854 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
25857 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
25859 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
25860 to 32 bytes when compiling for POWER9.
25862 2018-03-05 Jakub Jelinek <jakub@redhat.com>
25865 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
25866 regparm >= 3 with no arg reg available also for calls with
25867 flag_force_indirect_call. Pass decl to ix86_function_regparm.
25870 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
25872 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
25874 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
25877 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
25879 2018-03-05 Richard Biener <rguenther@suse.de>
25881 PR tree-optimization/84486
25882 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
25883 When inserting a __builtin_assume_aligned call set the LHS
25884 SSA name alignment info accordingly.
25886 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
25888 PR tree-optimization/84114
25889 * config/aarch64/aarch64.c (aarch64_reassociation_width)
25890 Avoid reassociation of FLOAT_MODE addition.
25892 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
25894 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
25895 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
25896 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
25897 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
25898 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
25899 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
25900 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
25902 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
25903 __builtin_ia32_wbinvd): New builtins.
25904 (SPECIAL_ARGS2): New.
25905 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
25906 (SPECIAL_ARGS2): New.
25907 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
25908 (ix86_valid_target_attribute_inner_p): Ditto.
25909 (ix86_init_mmx_sse_builtins): Add special_args2.
25910 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
25911 TARGET_WBNOINVD_P): New.
25912 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
25913 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
25914 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
25915 * config/i386/immintrin.h (_wbinvd): New intrinsic.
25916 * config/i386/pconfigintrin.h: New file.
25917 * config/i386/wbnoinvdintrin.h: Ditto.
25918 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
25920 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
25922 2018-03-05 Richard Biener <rguenther@suse.de>
25924 PR tree-optimization/84670
25925 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
25927 (BB_VISITED_WITH_VISITED_SUCCS): New define.
25928 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
25929 (compute_antic_aux): Only assert the number of values in ANTIC_IN
25930 doesn't grow if all successors (recursively) were visited at least
25933 2018-03-05 Richard Biener <rguenther@suse.de>
25935 PR tree-optimization/84650
25936 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
25937 if executed in the loop pipeline.
25939 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
25941 * doc/configfiles.texi (Configuration Files): Move info about
25942 conditionalizing $target-protos.h to...
25943 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
25944 differs from $target-protos.h.
25946 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
25947 Chung-Ju Wu <jasonwucj@gmail.com>
25949 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
25950 * config/nds32/nds32-multiple.md (setmemsi): Define.
25951 * config/nds32/nds32-memory-manipulation.c
25952 (nds32_gen_dup_4_byte_to_word_value): New.
25953 (emit_setmem_word_loop): New.
25954 (emit_setmem_byte_loop): New.
25955 (nds32_expand_setmem_loop): New.
25956 (nds32_expand_setmem_loop_v3m): New.
25957 (nds32_expand_setmem_unroll): New.
25958 (nds32_expand_setmem): New.
25960 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
25961 Chung-Ju Wu <jasonwucj@gmail.com>
25963 * config/nds32/nds32-memory-manipulation.c
25964 (nds32_emit_load_store): New.
25965 (nds32_emit_post_inc_load_store): New.
25966 (nds32_emit_mem_move): New.
25967 (nds32_emit_mem_move_block): New.
25968 (nds32_expand_movmemsi_loop_unknown_size): New.
25969 (nds32_expand_movmemsi_loop_known_size): New.
25970 (nds32_expand_movmemsi_loop): New.
25971 (nds32_expand_movmemsi_unroll): New.
25972 (nds32_expand_movmemqi): Rename ...
25973 (nds32_expand_movmemsi): ... to this.
25974 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
25975 (movmemsi): ... to this.
25976 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
25977 (nds32_expand_movmemsi): ... to this.
25979 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
25980 Monk Chiang <sh.chiang04@gmail.com>
25981 Chung-Ju Wu <jasonwucj@gmail.com>
25983 * config/nds32/nds32-protos.h
25984 (nds32_expand_load_multiple): New arguments.
25985 (nds32_expand_store_multiple): Ditto.
25986 (nds32_valid_multiple_load_store): Rename ...
25987 (nds32_valid_multiple_load_store_p): ... to this.
25988 * config/nds32/nds32-memory-manipulation.c
25989 (nds32_expand_load_multiple): Refine implementation.
25990 (nds32_expand_store_multiple): Ditto.
25991 * config/nds32/nds32-multiple.md
25992 (load_multiple): Update nds32_expand_load_multiple interface.
25993 (store_multiple): Update nds32_expand_store_multiple interface.
25994 * config/nds32/nds32-predicates.c
25995 (nds32_valid_multiple_load_store): Rename ...
25996 (nds32_valid_multiple_load_store_p): ... to this and refine
25998 * config/nds32/predicates.md
25999 (nds32_load_multiple_and_update_address_operation): New predicate.
26000 (nds32_store_multiple_and_update_address_operation): New predicate.
26002 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
26003 Chung-Ju Wu <jasonwucj@gmail.com>
26005 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
26006 (combo): New attribute.
26007 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
26009 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
26011 * config/nds32/nds32.opt: Change -mcmodel= default value.
26013 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
26014 Monk Chiang <sh.chiang04@gmail.com>
26015 Chung-Ju Wu <jasonwucj@gmail.com>
26017 * config/nds32/constants.md (unspec_element): New enum.
26018 * config/nds32/constraints.md (Umw): New constraint.
26019 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
26020 * config/nds32/nds32-intrinsic.md: Likewise.
26021 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
26022 (nds32_valid_smw_lwm_base_p): New.
26023 (nds32_output_smw_single_word): New.
26024 (nds32_output_lmw_single_word): New.
26025 (nds32_expand_unaligned_load): New.
26026 (nds32_expand_unaligned_store): New.
26027 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
26028 (nds32_output_smw_single_word): Declare.
26029 (nds32_output_lmw_single_word): Declare.
26030 (nds32_expand_unaligned_load): Declare.
26031 (nds32_expand_unaligned_store): Declare.
26032 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
26033 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
26034 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
26035 NDS32_BUILTIN_UASTORE_DW.
26036 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
26039 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
26040 Kito Cheng <kito.cheng@gmail.com>
26041 Chung-Ju Wu <jasonwucj@gmail.com>
26043 * config/nds32/nds32-intrinsic.c
26044 (nds32_expand_builtin_null_ftype_reg): Delete.
26045 (nds32_expand_builtin_reg_ftype_imm): Ditto.
26046 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
26047 (nds32_read_argument): New.
26048 (nds32_legitimize_target): Ditto.
26049 (nds32_legitimize_argument): Ditto.
26050 (nds32_check_constant_argument): Ditto.
26051 (nds32_expand_unop_builtin): Ditto.
26052 (nds32_expand_unopimm_builtin): Ditto.
26053 (nds32_expand_binop_builtin): Ditto.
26054 (nds32_builtin_decl_impl): Ditto.
26055 (builtin_description): Ditto.
26056 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
26057 (nds32_init_builtins_impl): Ditto.
26058 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
26059 (nds32_builtin_decl): New.
26060 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
26061 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
26063 2018-03-02 Jeff Law <law@redhat.com>
26065 * reorg.c (stop_search_p): Handle DEBUG_INSN.
26066 (redundant_insn, fill_simple_delay_slots): Likewise.
26067 (fill_slots_from_thread): Likewise.
26068 * resource.c (mark_referenced_resources): Likewise.
26069 (mark_set_resources, find_dead_or_set_registers): Likewise.
26071 2018-03-02 Jakub Jelinek <jakub@redhat.com>
26073 * substring-locations.h (format_warning_va): Formatting fix for
26074 ATTRIBUTE_GCC_DIAG.
26075 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
26077 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
26078 * substring-locations.c: Include intl.h.
26079 (format_warning_va): Turned into small wrapper around
26080 format_warning_n_va, renamed to ...
26081 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
26082 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
26084 (format_warning_at_substring_n): New function.
26085 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
26086 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
26087 format_warning_at_substring with just a shorter name instead of
26088 const function pointer.
26089 (fmtwarn_n): New function.
26090 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
26091 appropriate, get rid of all the fmtstr temporaries, move conditionals
26092 with G_() wrapped string literals directly into fmtwarn arguments,
26093 cast dir.len to (int), formatting fixes.
26095 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
26097 * doc/invoke.texi: Remove "Cilk Plus" references.
26099 2018-03-02 Jakub Jelinek <jakub@redhat.com>
26100 Richard Biener <rguenther@suse.de>
26103 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
26104 for error or warning attributes if CALL_FROM_THUNK_P is set.
26107 2018-03-02 Jakub Jelinek <jakub@redhat.com>
26110 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
26111 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
26114 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
26115 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
26117 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
26118 instead of -1U in last predictors element's probability member.
26120 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
26123 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
26124 arguments if they are comparable.
26126 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
26128 PR tree-optimization/84634
26129 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
26130 masks and masked_loop_p with a single loop_masks, making sure it's
26131 null for bb vectorization.
26133 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
26135 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
26136 (vect_analyze_data_ref_access): Use loop->safe_len rather than
26137 loop->force_vectorize to check whether there is no alias.
26139 2018-03-02 Jakub Jelinek <jakub@redhat.com>
26142 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
26144 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
26146 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
26147 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
26148 instead of a loop around prev_real_insn.
26149 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
26152 PR inline-asm/84625
26153 * config/i386/i386.c (ix86_print_operand): Use conditional
26154 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
26157 2018-03-02 Richard Biener <rguenther@suse.de>
26159 PR tree-optimization/84427
26160 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
26161 (bitmap_set_subtract_values): Rewrite to handle multiple
26164 (prune_clobbered_mems): Likewise.
26165 (phi_translate): Take edge instead of pred/phiblock.
26166 (phi_translate_1): Likewise.
26167 (phi_translate_set): Likewise. Insert all translated
26168 exprs for a value into the set, keeping possibly multiple
26169 expressions per value.
26170 (compute_antic_aux): Adjust for phi_translate changes.
26171 When intersecting union the expressions and prune those
26172 not in the final value set, keeping possibly multiple
26173 expressions per value. Do not use value-insertion
26174 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
26175 all expressions. Add verification that the value-sets
26176 only shrink during iteration.
26177 (compute_partial_antic_aux): Adjust for the phi_translate changes.
26178 (do_pre_regular_insertion): Likewise.
26179 (do_pre_partial_partial_insertion): Likewise.
26181 2018-03-02 Richard Biener <rguenther@suse.de>
26184 * config/darwin.c (saved_debug_info_level): New static global.
26185 (darwin_asm_lto_start): Disable debug info generation for LTO out.
26186 (darwin_asm_lto_end): Restore debug info generation settings.
26188 2018-03-01 Martin Liska <mliska@suse.cz>
26191 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
26192 volatile arguments.
26194 2018-03-01 Richard Biener <rguenther@suse.de>
26197 * dwarf2out.c (gen_variable_die): Properly handle late VLA
26198 type annotation with LTO when debug was disabled at compile-time.
26200 2018-03-01 Matthew Fortune <mfortune@gmail.com>
26202 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
26204 (mips_final_postscan_insn): Likewise.
26206 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
26208 PR rtl-optimization/84528
26209 * alias.c (init_alias_target): Add commentary.
26210 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
26211 a unique base value if the frame pointer is not eliminated
26212 to the stack pointer.
26214 2018-03-01 Tom de Vries <tom@codesourcery.com>
26216 PR rtl-optimization/83327
26217 * lra-int.h (hard_regs_spilled_into): Declare.
26218 * lra.c (hard_regs_spilled_into): Define.
26219 (init_reg_info): Init hard_regs_spilled_into.
26220 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
26221 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
26222 (process_bb_lives): Handle hard_regs_spilled_into.
26223 (lra_create_live_ranges_1): Before doing liveness propagation, clear
26224 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
26226 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
26228 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
26229 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
26230 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
26231 * config/rs6000/aix72.h: New file.
26233 2018-02-28 Jakub Jelinek <jakub@redhat.com>
26235 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
26236 instead of warning_at with conditional singular and plural messages
26240 * stor-layout.c (update_alignment_for_field): For
26241 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
26242 && !DECL_PACKED (field), do the alignment update, just use
26243 only desired_align instead of MAX (type_align, desired_align)
26245 (place_field): Don't do known_align < desired_align handling
26246 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
26247 is non-NULL, instead do it after rli->prev_field handling and
26248 only if not within a bitfield word. For DECL_PACKED (field)
26249 use type_align of BITS_PER_UNIT.
26251 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
26253 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
26254 superfluous parentheses and trailing spaces.
26256 2018-02-28 Richard Biener <rguenther@suse.de>
26258 PR tree-optimization/84584
26259 * graphite-scop-detection.c (scop_detection::add_scop): Discard
26260 SCoPs with fake exit edge.
26262 2018-02-28 Martin Liska <mliska@suse.cz>
26265 * timevar.c (timer::print): Fix format to properly print 100%
26268 2018-02-28 Richard Biener <rguenther@suse.de>
26270 PR middle-end/84607
26271 * genmatch.c (capture_info::walk_match): Do not mark
26272 captured expressions without operands as expr_p given
26273 they act more like predicates and should be subject to
26274 "lost tail" side-effect preserving.
26276 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
26278 PR rtl-optimization/81611
26279 * auto-inc-dec.c (attempt_change): Move dead note from
26280 mem_insn if it's the next use of regno
26281 (find_address): Take address use of reg holding
26282 non-incremented value. Add parm to limit search to the named
26284 (merge_in_block): Attempt to use a mem insn that is the next
26285 use of the original regno.
26287 2018-02-27 Martin Sebor <msebor@redhat.com>
26290 * doc/invoke.texi (-Wmissing-attributes): New option.
26291 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
26293 2018-02-27 Martin Sebor <msebor@redhat.com>
26295 PR translation/84207
26296 * diagnostic-core.h (warning_n, error_n, inform_n): Change
26297 n argument to unsigned HOST_WIDE_INT.
26298 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
26299 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
26300 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
26301 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
26303 2018-02-27 Richard Biener <rguenther@suse.de>
26305 PR tree-optimization/84512
26306 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
26307 Do not use the estimate returned from record_stmt_cost for
26308 the scalar iteration cost but sum properly using add_stmt_cost.
26310 2018-02-27 Richard Biener <rguenther@suse.de>
26312 PR tree-optimization/84466
26313 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
26314 Adjust last change to less strictly validate use operands.
26316 2018-02-27 Martin Liska <mliska@suse.cz>
26318 PR gcov-profile/84548
26319 * gcov.c (process_file): Allow partial overlap and consider it
26320 also as group functions.
26321 (output_lines): Properly calculate range of lines for a group.
26323 2018-02-27 Martin Liska <mliska@suse.cz>
26325 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
26326 'ggc' suffixes. Change first column width.
26327 (timer::print): Fix formatting of the column.
26329 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
26331 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
26332 preserve inline entry blocks for the sake of debug inline
26333 entry point markers alone.
26334 (remove_unused_locals): Suggest in comments a better place to
26335 force the preservation of inline entry blocks that are
26336 otherwise unused, but do not preserve them.
26338 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
26340 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
26342 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
26345 * config/i386/constraints.md (Bs): Replace
26346 ix86_indirect_branch_register with
26347 TARGET_INDIRECT_BRANCH_REGISTER.
26349 * config/i386/i386.md (indirect_jump): Likewise.
26350 (tablejump): Likewise.
26351 (*sibcall_memory): Likewise.
26352 (*sibcall_value_memory): Likewise.
26353 Peepholes of indirect call and jump via memory: Likewise.
26354 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
26355 (*sibcall_value_GOT_32): Likewise.
26356 * config/i386/predicates.md (indirect_branch_operand): Likewise.
26357 (GOT_memory_operand): Likewise.
26358 (call_insn_operand): Likewise.
26359 (sibcall_insn_operand): Likewise.
26360 (GOT32_symbol_operand): Likewise.
26361 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
26363 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
26365 PR rtl-optimization/83496
26366 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
26367 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
26368 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
26369 redundant insn, if any.
26370 (relax_delay_slots): Likewise.
26371 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
26373 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
26375 PR tree-optimization/83965
26376 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
26377 that grouped statements are part of a reduction chain. Return
26378 true if the statement is not marked as a reduction itself but
26379 is part of a group.
26380 (vect_recog_dot_prod_pattern): Don't check whether the statement
26381 is part of a group here.
26382 (vect_recog_sad_pattern): Likewise.
26383 (vect_recog_widen_sum_pattern): Likewise.
26385 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
26388 * final.c (rest_of_clean_state): Also look for calls inside sequences.
26390 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
26393 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
26395 (ix86_output_indirect_function_return): New prototype.
26396 (ix86_split_simple_return_pop_internal): Likewise.
26397 * config/i386/i386.c (indirect_return_via_cx): New.
26398 (indirect_return_via_cx_bnd): Likewise.
26399 (indirect_thunk_name): Handle return va CX_REG.
26400 (output_indirect_thunk_function): Create alias for
26401 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
26402 (ix86_output_indirect_jmp): Remove the bool argument.
26403 (ix86_output_indirect_function_return): New function.
26404 (ix86_split_simple_return_pop_internal): Likewise.
26405 * config/i386/i386.md (*indirect_jump): Don't pass false
26406 to ix86_output_indirect_jmp.
26407 (*tablejump_1): Likewise.
26408 (simple_return_pop_internal): Change it to define_insn_and_split.
26409 Call ix86_split_simple_return_pop_internal to split it for
26410 -mfunction-return=.
26411 (simple_return_indirect_internal): Call
26412 ix86_output_indirect_function_return instead of
26413 ix86_output_indirect_jmp.
26415 2018-02-26 Jakub Jelinek <jakub@redhat.com>
26418 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
26419 memset and value initialization afterwards.
26421 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
26423 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
26425 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
26428 * common/config/aarch64/aarch64-common.c
26429 (aarch_option_optimization_table[]): Switch
26430 off fomit-frame-pointer
26432 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
26433 Chung-Ju Wu <jasonwucj@gmail.com>
26435 * config/nds32/nds32-multiple.md (load_multiple): Disallow
26437 (store_multiple): Ditto.
26439 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
26441 * config.gcc: Add --with-cpu support for nds32 target.
26442 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
26443 * config/nds32/nds32.opt: Add -mcpu= option.
26445 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
26447 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
26448 isel=yes): Warn for these deprecated options.
26450 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
26452 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
26453 ISA_2_5_MASKS_EMBEDDED.
26455 2018-02-23 Jakub Jelinek <jakub@redhat.com>
26457 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
26458 p->max as pointers rather than using iterative_hash_expr.
26460 2018-02-23 Carl Love <cel@us.ibm.com>
26462 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
26463 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
26465 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
26466 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
26467 P8V_BUILTIN_VEC_VUNSIGNED2.
26469 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
26472 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
26473 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
26475 * lra-constraints.c (curr_insn_transform): Set up
26476 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
26478 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
26479 * lra-eliminations.c (spill_pseudos): Ditto.
26480 (process_insn_for_elimination): Ditto.
26481 * lra-lives.c (reg_early_clobber_p): Use the new macros.
26482 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
26483 LRA_NON_CLOBBERED_ALT.
26485 2018-02-22 Martin Sebor <msebor@redhat.com>
26487 PR tree-optimization/84480
26488 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
26489 to maybe_diag_stxncpy_trunc. Call it.
26490 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
26491 from gimple_fold_builtin_strcpy. Print inlining stack.
26492 (handle_builtin_stxncpy): Print inlining stack.
26493 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
26495 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
26498 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
26499 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
26500 and -fcheck-pointer-bounds are used together.
26501 (indirect_thunk_prefix): New enum.
26502 (indirect_thunk_need_prefix): New function.
26503 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
26504 "_nt" instead of "_bnd" for NOTRACK prefix.
26505 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
26506 (output_indirect_thunk_function): Likewise.
26508 (ix86_code_end): Update output_indirect_thunk_function calls.
26509 (ix86_output_indirect_branch_via_reg): Replace
26510 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
26511 (ix86_output_indirect_branch_via_push): Likewise.
26512 (ix86_output_function_return): Likewise.
26513 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
26514 incompatible with -fcf-protection=branch and
26515 -fcheck-pointer-bounds.
26517 2018-02-22 Steve Ellcey <sellcey@cavium.com>
26520 * config/aarch64/aarch64.c (aarch64_print_address_internal):
26521 Change gcc_assert call to output_operand_lossage.
26523 2018-02-22 Steve Ellcey <sellcey@cavium.com>
26525 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
26527 2018-02-22 DJ Delorie <dj@redhat.com>
26528 Sebastian Perta <sebastian.perta@renesas.com>
26529 Oleg Endo <olegendo@gcc.gnu.org>
26531 * config/rx/rx.c (rx_rtx_costs): New function.
26532 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
26534 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
26536 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
26538 2018-02-22 Martin Liska <mliska@suse.cz>
26541 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
26542 Add "native" as a possible value.
26544 2018-02-22 Martin Liska <mliska@suse.cz>
26547 * config/i386/i386.c (ix86_option_override_internal):
26548 Add "native" as a possible value for -march and -mtune.
26550 2018-02-22 Jakub Jelinek <jakub@redhat.com>
26553 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
26554 to all type variants.
26556 PR tree-optimization/84503
26557 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
26558 width as info->bitpos + info->bitsize - start.
26559 (merged_store_group::merge_overlapping): Simplify width computation.
26560 (check_no_overlap): New function.
26561 (imm_store_chain_info::try_coalesce_bswap): Compute expected
26562 start + width and last_order of the group, fail if check_no_overlap
26564 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
26565 to group if check_no_overlap fails.
26567 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
26569 * config/rs6000/altivec.md: Delete contraint arguments to
26570 define_expand, define_split, and define_peephole2, and in
26571 define_insn_and_split if always unused.
26572 * config/rs6000/darwin.md: Ditto.
26573 * config/rs6000/dfp.md: Ditto.
26574 * config/rs6000/rs6000.md: Ditto.
26575 * config/rs6000/sync.md: Ditto.
26576 * config/rs6000/vector.md: Ditto.
26577 * config/rs6000/vsx.md: Ditto.
26579 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
26581 * config/rs6000/altivec.md: Write output control strings as braced
26582 blocks instead of double-quoted strings.
26583 * config/rs6000/darwin.md: Ditto.
26584 * config/rs6000/rs6000.md: Ditto.
26585 * config/rs6000/vector.md: Ditto.
26586 * config/rs6000/vsx.md: Ditto.
26588 2018-02-21 Jason Merrill <jason@redhat.com>
26590 PR c++/84314 - ICE with templates and fastcall attribute.
26591 * attribs.c (build_type_attribute_qual_variant): Remove assert.
26593 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
26595 * ipa-cp.c (determine_versionability): Fix comment typos.
26597 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
26600 * ipa-cp.c (determine_versionability): Do not version functions caling
26603 2018-02-21 Martin Liska <mliska@suse.cz>
26606 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
26607 Add "native" as a possible value.
26608 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
26609 the macro when native cpu detection is available.
26611 2018-02-21 Martin Liska <mliska@suse.cz>
26614 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
26615 Add "native" as a possible value.
26616 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
26617 when native cpu detection is available.
26619 2018-02-21 Jakub Jelinek <jakub@redhat.com>
26620 Martin Sebor <msebor@redhat.com>
26622 PR tree-optimization/84478
26623 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
26625 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
26626 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
26627 support which is conservatively correct, for 2 only stay conservative
26628 for maxlen. Formatting and comment capitalization fixes. Add STRICT
26629 argument to the 2 argument get_range_strlen, adjust 6 arg
26630 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
26632 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
26633 (gimple_fold_builtin_strlen): Pass true as last argument to
26636 2018-02-20 Martin Sebor <msebor@redhat.com>
26638 PR middle-end/84095
26639 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
26640 (builtin_memref::set_base_and_offset): Same. Handle inner references.
26641 (builtin_memref::builtin_memref): Factor out parts into
26642 set_base_and_offset and call it.
26644 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
26646 PR middle-end/84406
26647 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
26648 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
26649 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
26650 search at the associated MODE_INT.
26652 2018-02-20 Jeff Law <law@redhat.com>
26654 PR middle-end/82123
26655 PR tree-optimization/81592
26656 PR middle-end/79257
26657 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
26658 for range data rather than using global data.
26659 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
26660 range data rather than using global data.
26661 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
26662 pass it to children as needed.
26663 (struct directive::fmtresult): Similarly.
26664 (struct directive::set_width): Similarly.
26665 (struct directive::set_precision): Similarly.
26666 (format_integer, format_directive, parse_directive): Similarly.
26667 (format_none): Accept unnamed vr_values parameter.
26668 (format_percent, format_floating, format_character): Similarly.
26669 (format_string, format_plain): Similarly.
26670 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
26671 the EVRP range analyzer for range data rather than using global data.
26672 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
26673 gimple-ssa-evrp-analyze.h
26674 (class sprintf_dom_walker): Add after_dom_children member function.
26675 Add evrp_range_analyzer member.
26676 (sprintf_dom_walker::before_dom_children): Call into the EVRP
26677 range analyzer as needed.
26678 (sprintf_dom_walker::after_dom_children): New member function.
26679 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
26681 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
26682 (evrp_range_analyzer::pop_to_marker): Likewise.
26684 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
26686 PR tree-optimization/84419
26687 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
26688 with the required type if its current type is compatible but
26691 2018-02-20 Jakub Jelinek <jakub@redhat.com>
26693 PR middle-end/82004
26694 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
26695 after vectorization.
26697 2018-02-20 Martin Liska <mliska@suse.cz>
26700 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
26701 possible values if we don't have a hint.
26703 2018-02-20 Martin Liska <mliska@suse.cz>
26707 * final.c (shorten_branches): Build align_tab array with one
26709 * opts.c (finish_options): Add alignment option limit check.
26710 (MAX_CODE_ALIGN): Likewise.
26711 (MAX_CODE_ALIGN_VALUE): Likewise.
26712 * doc/invoke.texi: Document maximum allowed option value for
26713 all -falign-* options.
26715 2018-02-19 Jakub Jelinek <jakub@redhat.com>
26718 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
26719 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
26720 * var-tracking.c (emit_note_insn_var_location): Remove all references
26721 to NOTE_INSN_CALL_ARG_LOCATION.
26722 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
26723 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
26724 Use copy_rtx_if_shared.
26725 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
26726 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
26727 (dwarf2out_var_location): Remove handling of
26728 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
26730 * final.c (final_scan_insn): Remove all references to
26731 NOTE_INSN_CALL_ARG_LOCATION.
26732 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
26733 before dumping final insns.
26734 * except.c (emit_note_eh_region_end): Remove all references to
26735 NOTE_INSN_CALL_ARG_LOCATION.
26736 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
26737 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
26738 * config/arc/arc.c (hwloop_optimize): Likewise.
26739 * config/arm/arm.c (create_fix_barrier): Likewise.
26740 * config/s390/s390.c (s390_chunkify_start): Likewise.
26741 * config/sh/sh.c (find_barrier): Likewise.
26742 * config/i386/i386.c (rest_of_insert_endbranch,
26743 ix86_seh_fixup_eh_fallthru): Likewise.
26744 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
26745 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
26746 * config/frv/frv.c (frv_function_prologue): Likewise.
26747 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
26749 (note_outside_basic_block_p): Remove all references to
26750 NOTE_INSN_CALL_ARG_LOCATION.
26751 * gengtype.c (adjust_field_rtx_def): Likewise.
26752 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
26754 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
26755 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
26758 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
26761 PR tree-optimization/84452
26762 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
26763 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
26766 2018-02-19 Martin Liska <mliska@suse.cz>
26769 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
26771 2018-02-19 Martin Liska <mliska@suse.cz>
26772 Richard Sandiford <richard.sandiford@linaro.org>
26774 PR tree-optimization/82491
26775 * gimple-fold.c (get_base_constructor): Make earlier bail out
26778 2018-02-19 Carl Love <cel@us.ibm.com>
26780 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
26781 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
26783 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
26784 P8V_BUILTIN_VEC_NEG.
26786 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
26788 * config/rl78/rl78.md (movdf): New define expand.
26790 2018-02-19 Martin Liska <mliska@suse.cz>
26793 * doc/invoke.texi: Fix typo.
26794 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
26796 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
26798 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
26799 handle rs6000_single_float and rs6000_double_float specially for
26802 2018-02-16 Jeff Law <law@redhat.com>
26804 * config/rx/rx.c (add_pop_cfi_notes): New function.;
26805 (pop_regs): Use it.
26807 2018-02-16 Jakub Jelinek <jakub@redhat.com>
26810 * ipa-inline.c (inline_small_functions): Fix a typo.
26812 2018-02-16 Nathan Sidwell <nathan@acm.org>
26814 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
26816 2018-02-16 Carl Love <cel@us.ibm.com>
26818 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
26819 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
26820 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
26821 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
26822 expansion to P8V_BUILTIN_VEC_FLOAT2.
26824 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
26826 PR rtl-optimization/70023
26827 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
26828 src_regno into account.
26830 2018-02-16 Carl Love <cel@us.ibm.com>
26832 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
26833 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
26834 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
26835 * config/rs6000/rs6000.c: Remove case statements for
26836 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
26837 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
26838 and P9V_BUILTIN_VEC_VINSERT4B.
26839 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
26840 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
26841 * config/rs6000/vsx.md:
26842 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
26845 2018-02-16 Carl Love <cel@us.ibm.com>
26847 * config/rs6000/altivec.h: Add builtin names vec_extract4b
26849 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
26851 * config/rs6000/rs6000-c.c: Add the definitions for
26852 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
26853 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
26854 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
26855 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
26856 definition for insert4b and define insn *insert3b_internal.
26857 * doc/extend.texi: Add documentation for vec_extract4b.
26859 2018-02-16 Nathan Sidwell <nathan@acm.org>
26861 * doc/extend.texi (Backwards Compatibility): Mention friend
26862 injection. Note for-scope is deprecated.
26863 * doc/invoke.texi (-ffriend-injection): Deprecate.
26865 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
26867 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
26868 that moved to I2, also allow destinations that are a paradoxical
26869 subreg (instead of a normal reg).
26871 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
26874 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
26877 2018-02-16 Richard Biener <rguenther@suse.de>
26879 PR tree-optimization/84037
26880 PR tree-optimization/84016
26882 * config/i386/i386.c (ix86_builtin_vectorization_cost):
26883 Adjust vec_construct for the fact we need additional higher latency
26884 128bit inserts for AVX256 and AVX512 vector builds.
26885 (ix86_add_stmt_cost): Scale vector construction cost for
26888 2018-02-16 Richard Biener <rguenther@suse.de>
26890 PR tree-optimization/84417
26891 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
26892 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
26893 (non_rewritable_lvalue_p): Likewise, use poly-ints.
26895 2018-02-16 Martin Liska <mliska@suse.cz>
26898 * internal-fn.def (ASAN_CHECK): Set proper flags.
26899 (ASAN_MARK): Likewise.
26901 2018-02-16 Julia Koval <julia.koval@intel.com>
26903 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
26904 from PTA_CANNONLAKE.
26906 2018-02-16 Jakub Jelinek <jakub@redhat.com>
26909 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
26910 Use ++iter rather than iter++ for std::list iterators.
26911 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
26912 defer deleting them until all nodes in the forest are processed. Do
26913 free even leaf nodes. Change to_process into auto_vec.
26916 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
26917 * vec.h (vec_default_construct): Use memset instead of placement new
26918 if BROKEN_VALUE_INITIALIZATION is defined.
26919 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
26920 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
26923 PR rtl-optimization/83723
26924 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
26925 * lra.c (lra_substitute_pseudo): Likewise. If true, use
26926 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
26928 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
26930 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
26932 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
26934 PR rtl-optimization/81443
26935 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
26936 from inner REGs to paradoxical SUBREGs.
26938 2018-02-16 Richard Biener <rguenther@suse.de>
26940 PR tree-optimization/84399
26941 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
26942 For operands we can analyze at their definition make sure we can
26943 analyze them at each use as well.
26945 2018-02-16 Richard Biener <rguenther@suse.de>
26947 PR tree-optimization/84190
26948 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
26949 volatile accesses if the decl isn't volatile.
26951 2018-02-15 Jason Merrill <jason@redhat.com>
26953 PR c++/84314 - ICE with templates and fastcall attribute.
26954 * attribs.c (build_type_attribute_qual_variant): Don't clobber
26955 TYPE_CANONICAL on an existing type.
26957 2018-02-15 Jakub Jelinek <jakub@redhat.com>
26959 PR tree-optimization/84383
26960 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
26961 dstoff nor call operand_equal_p if dstbase is NULL.
26963 PR tree-optimization/84334
26964 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
26965 also a CONSTANT_CLASS_P, punt.
26967 2018-02-14 Jim Wilson <jimw@sifive.com>
26969 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
26970 first SMALL_OPERAND check. New local min_second_step. Move assert
26971 to where locals are set. Add TARGET_RVC support.
26972 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
26974 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
26976 * doc/invoke.texi: Correct -Wformat-overflow code sample.
26978 2018-02-14 Martin Sebor <msebor@redhat.com>
26980 PR tree-optimization/83698
26981 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
26982 arrays constrain the offset range to their bounds.
26983 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
26984 (builtin_access::overlap): Avoid setting the size of overlap if it's
26986 (maybe_diag_overlap): Also consider arrays when deciding what values
26987 of offsets to include in diagnostics.
26989 2018-02-14 Martin Sebor <msebor@redhat.com>
26992 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
26993 that correspond to the kind of a declaration.
26995 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
26998 * config/pa/pa.md: Load address of PIC label using the linkage table
26999 if the label is nonlocal.
27001 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
27003 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
27004 warning message if user requests -maltivec=be.
27005 * doc/invoke.texi: Document deprecation of -maltivec=be.
27007 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
27010 * config/rs6000/rs6000-c.c: Update definitions for
27011 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
27012 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
27014 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
27017 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
27018 add _get_ssp intrinsics. Remove argument from
27019 __builtin_ia32_rdssp[d|q].
27020 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
27021 * config/i386/i386-builtin.def: Remove argument from
27022 __builtin_ia32_rdssp[d|q].
27023 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
27024 ix86_expand_special_args_builtin for _rdssp[d|q].
27025 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
27026 Clear register before usage.
27027 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
27028 Add documentation for new _get_ssp and _inc_ssp intrinsics.
27030 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
27032 PR tree-optimization/84357
27033 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
27034 operand 1 of an ARRAY_REF too.
27036 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
27039 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
27040 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
27042 (set_of_reg): New struct.
27043 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
27044 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
27045 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
27047 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
27048 Split into bitclr, bitset, bitinvert patterns if appropriate.
27049 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
27050 use rx_fuse_in_memory_bitop.
27051 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
27052 to named insn, correct maximum insn length.
27054 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
27057 * machmode.def: Define a complex mode for PARTIAL_INT.
27058 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
27060 * doc/rtl.texi: Document CSPImode.
27061 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
27063 (msp430_hard_regno_nregs_with_padding): Likewise.
27065 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
27068 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
27070 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
27072 PR rtl-optimization/84169
27073 * combine.c (try_combine): New variable split_i2i3. Set it to true if
27074 we generated a parallel as new i3 and we split that to new i2 and i3
27075 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
27076 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
27077 those to i2, not i1. Partially rewrite this scan code.
27079 2018-02-13 Jakub Jelinek <jakub@redhat.com>
27082 * stor-layout.c (place_field): For variable length fields, adjust
27083 offset_align afterwards not just based on the field's alignment,
27084 but also on the size.
27086 PR middle-end/84309
27087 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
27088 of exps and logs in the use_exp2 case.
27090 2018-02-13 Jeff Law <law@redhat.com>
27092 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
27093 entry for "vector".
27095 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
27098 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
27102 * common.opt (gas-loc-support, gas-locview-support): New.
27103 (ginline-points, ginternal-reset-location-views): New.
27104 * doc/invoke.texi: Document them. Use @itemx where intended.
27105 (gvariable-location-views): Adjust.
27106 * target.def (reset_location_view): New.
27107 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
27108 (TARGET_RESET_LOCATION_VIEW): New.
27109 * doc/tm.texi: Rebuilt.
27110 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
27111 (dwarf2out_default_as_locview_support): New.
27112 (output_asm_line_debug_info): Use option variables.
27113 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
27114 (output_loc_list): Likewise.
27115 (add_high_low_attributes): Check option variables.
27116 Don't output entry view attribute in strict mode.
27117 (gen_inlined_subroutine_die): Check option variables.
27118 (dwarf2out_inline_entry): Likewise.
27119 (init_sections_and_labels): Likewise.
27120 (dwarf2out_early_finish): Likewise.
27121 (maybe_reset_location_view): New, from...
27122 (dwarf2out_var_location): ... here. Call it.
27123 * debug.h (dwarf2out_default_as_loc_support): Declare.
27124 (dwarf2out_default_as_locview_support): Declare.
27125 * hooks.c (hook_int_rtx_insn_0): New.
27126 * hooks.h (hook_int_rtx_insn_0): Declare.
27127 * toplev.c (process_options): Take -gas-loc-support and
27128 -gas-locview-support from dwarf2out. Enable
27129 -gvariable-location-views by default only with locview
27130 assembler support. Enable -ginternal-reset-location-views by
27131 default only if the target defines the corresponding hook.
27132 Enable -ginline-points by default if location views are
27133 enabled; force it disabled if statement frontiers are
27135 * tree-inline.c (expand_call_inline): Check option variables.
27136 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
27138 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
27140 PR tree-optimization/84321
27141 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
27142 handling. Also check whether the anti-range contains any values
27143 that satisfy the mask; switch to a VR_RANGE if not.
27145 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
27148 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
27150 2018-02-13 Martin Jambor <mjambor@suse.cz>
27153 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
27154 of call statements, also set location of a load to a temporary.
27156 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
27158 * config/rl78/rl78.c (add_vector_labels): New function.
27159 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
27160 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
27161 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
27162 which checks that no arguments are passed.
27163 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
27164 * doc/extend.texi: Documentation for the new attribute.
27166 2018-02-13 Andreas Schwab <schwab@suse.de>
27168 * config/riscv/linux.h (CPP_SPEC): Define.
27170 2018-02-13 Jakub Jelinek <jakub@redhat.com>
27173 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
27174 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
27175 OPTION_MASK_ISA_AES as first argument to def_builtin_const
27176 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
27177 instead of OPTION_MASK_ISA_PCLMUL as first argument to
27178 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
27179 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
27180 temporarily for AES and PCLMUL builtins.
27182 PR tree-optimization/84339
27183 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
27184 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
27187 PR middle-end/84309
27188 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
27189 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
27190 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
27192 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
27194 * omp-simd-clone.h: New file.
27195 * omp-simd-clone.c: Include omp-simd-clone.h.
27196 (expand_simd_clones): No longer static.
27197 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
27198 cgraph.h and omp-simd-clone.h.
27199 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
27200 (vect_recog_widen_shift_pattern): Formatting fix.
27201 (vect_pattern_recog_1): Don't check optab for calls.
27204 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
27205 operands[2] into a REG before using gen_lowpart on it.
27207 2018-02-12 Jeff Law <law@redhat.com>
27210 * config/sh/sh.c (find_barrier): Consider a sibling call
27213 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
27214 successfully back substituting a reg.
27216 2018-02-12 Richard Biener <rguenther@suse.de>
27218 PR tree-optimization/84037
27219 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
27220 parameter, move visited init to caller.
27221 (vect_slp_analyze_operations): Separate cost from validity
27222 check, initialize visited once for all instances.
27223 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
27225 * tree-vect-stmts.c (vect_model_simple_cost): Make early
27227 (vect_model_promotion_demotion_cost): Likewise.
27228 (vectorizable_bswap): Guard cost modeling with !slp_node
27229 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
27231 (vectorizable_call): Likewise.
27232 (vectorizable_conversion): Likewise.
27233 (vectorizable_assignment): Likewise.
27234 (vectorizable_shift): Likewise.
27235 (vectorizable_operation): Likewise.
27236 (vectorizable_store): Likewise.
27237 (vectorizable_load): Likewise.
27238 (vectorizable_condition): Likewise.
27239 (vectorizable_comparison): Likewise.
27241 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
27244 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
27245 (ASAN_MARK): Fix fnspec to account for return value, change pointer
27246 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
27248 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
27250 PR middle-end/83665
27251 * params.def (inline-min-speedup): Increase from 8 to 15.
27252 (max-inline-insns-auto): Decrease from 40 to 30.
27253 * ipa-split.c (consider_split): Add some buffer for function to
27254 be considered inlining candidate.
27255 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
27258 2018-02-12 Richard Biener <rguenther@suse.de>
27260 PR tree-optimization/84037
27261 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
27262 matched stmts if we cannot swap the non-matched ones.
27264 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
27266 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
27267 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
27268 _mm_maskz_scalef_round_ss): New intrinsics.
27269 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
27270 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
27271 __builtin_ia32_scalefss_round): Remove.
27272 (__builtin_ia32_scalefsd_mask_round,
27273 __builtin_ia32_scalefss_mask_round): New intrinsics.
27274 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
27275 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
27276 ((match_operand:VF_128 2 "<round_nimm_predicate>"
27277 "<round_constraint>")): Changed to ...
27278 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
27279 "<round_scalar_constraint>")): ... this.
27280 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
27281 %0, %1, %2<round_op3>}"): Changed to ...
27282 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
27283 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
27284 %2<round_scalar_mask_op3>}"): ... this.
27285 * config/i386/subst.md (round_scalar_nimm_predicate): New.
27287 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
27289 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
27290 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
27291 (_mm_maskz_sqrt_round_ss): New intrinsics.
27292 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
27293 (__builtin_ia32_sqrtsd_mask_round)
27294 (__builtin_ia32_sqrtss_mask_round): New builtins.
27295 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
27296 (__builtin_ia32_sqrtss_round): Remove.
27297 (__builtin_ia32_sqrtsd_mask_round)
27298 (__builtin_ia32_sqrtss_mask_round): New builtins.
27299 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
27300 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
27301 ((match_operand:VF_128 1 "vector_operand"
27302 "xBm,<round_constraint>")): Changed to ...
27303 ((match_operand:VF_128 1 "vector_operand"
27304 "xBm,<round_scalar_constraint>")): ... this.
27305 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
27306 %0, %2, %<iptr>1<round_op3>}): Changed to ...
27307 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
27308 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
27309 %<iptr>1<round_scalar_mask_op3>}): ... this.
27310 ((set_attr "prefix" "<round_prefix>")): Changed to ...
27311 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
27313 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
27316 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
27317 Cast vec_cmpeq result to correct type.
27318 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
27319 Cast vec_cmpgt result to correct type.
27321 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
27323 * final.c (final_scan_insn_1): Renamed from...
27324 (final_scan_insn): ... this. New wrapper, to recover
27325 seen from the outermost call in recursive ones.
27326 * config/sparc/sparc.c (output_return): Drop seen from call.
27327 (output_sibcall): Likewise.
27328 * config/visium/visium.c (output_branch): Likewise.
27330 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
27332 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
27335 2018-02-10 Alan Modra <amodra@gmail.com>
27338 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
27339 Specify LR as an input.
27341 2018-02-10 Jakub Jelinek <jakub@redhat.com>
27344 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
27345 remove_member_access_dummy_vars): New functions.
27346 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
27347 lower_omp_1, execute_lower_omp): Use them.
27349 PR rtl-optimization/84308
27350 * shrink-wrap.c (spread_components): Release todo vector.
27352 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
27354 PR rtl-optimization/57193
27355 * ira-color.c (struct allocno_color_data): Add member
27356 conflict_allocno_hard_prefs.
27357 (update_conflict_allocno_hard_prefs): New.
27358 (bucket_allocno_compare_func): Add a preference based on
27359 conflict_allocno_hard_prefs.
27360 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
27361 (color_allocnos): Remove a dead code. Initiate
27362 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
27364 2018-02-09 Jakub Jelinek <jakub@redhat.com>
27367 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
27368 constraint from =wa to wa. Avoid a subreg on the output operand,
27369 instead use a pseudo and subreg it in a move.
27370 (p9_xxbrd_<mode>): Changed to ...
27371 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
27372 (p9_xxbrd_v2df): New expander.
27373 (p9_xxbrw_<mode>): Changed to ...
27374 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
27375 (p9_xxbrw_v4sf): New expander.
27377 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
27379 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
27381 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
27384 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
27385 multiply in 32-bit mode.
27386 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
27387 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
27390 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
27392 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
27393 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
27394 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
27395 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
27397 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
27400 * dwarf2out.c (is_trivial_indirect_ref): New function.
27401 (dwarf2out_late_global_decl): Do not generate a location
27402 attribute for variables that have a non-trivial DECL_VALUE_EXPR
27403 and that are not defined in the current unit.
27405 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
27407 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
27408 instead of a libcall for UNORDERED.
27410 2018-02-09 Tamar Christina <tamar.christina@arm.com>
27413 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
27414 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
27416 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
27419 * config/s390/s390.c (s390_set_current_function): Invoke
27420 s390_indirect_branch_settings also if fndecl didn't change.
27422 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
27424 * config/rs6000/rs6000.md (blockage): Set length to zero.
27426 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
27428 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
27430 2018-02-09 Jakub Jelinek <jakub@redhat.com>
27433 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
27434 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
27438 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
27439 PARALLEL incoming that failed vt_get_decl_and_offset check.
27441 PR middle-end/84237
27442 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
27443 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
27445 (get_variable_section): For decls in named .bss* sections pass true as
27446 second argument to bss_initializer_p.
27448 2018-02-09 Marek Polacek <polacek@redhat.com>
27449 Jakub Jelinek <jakub@redhat.com>
27452 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
27453 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
27454 Sync some changes from cxx_fold_indirect_ref.
27456 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
27458 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
27460 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
27461 (BLOCK_INLINE_ENTRY_LABEL): New.
27462 (dwarf2out_var_location): Disregard inline entry markers.
27463 (inline_entry_data): New struct.
27464 (inline_entry_data_hasher): New hashtable type.
27465 (inline_entry_data_hasher::hash): New.
27466 (inline_entry_data_hasher::equal): New.
27467 (inline_entry_data_table): New variable.
27468 (add_high_low_attributes): Add DW_AT_entry_pc and
27469 DW_AT_GNU_entry_view attributes if a pending entry is found
27470 in inline_entry_data_table. Add old entry_pc attribute only
27471 if debug nonbinding markers are disabled.
27472 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
27473 markers are enabled.
27474 (block_within_block_p, dwarf2out_inline_entry): New.
27475 (dwarf2out_finish): Check that no entries remained in
27476 inline_entry_data_table.
27477 * final.c (reemit_insn_block_notes): Handle inline entry notes.
27478 (final_scan_insn, notice_source_line): Likewise.
27479 (rest_of_clean_state): Skip inline entry markers.
27480 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
27482 * gimple.c (gimple_build_debug_inline_entry): New.
27483 * gimple.h (enum gimple_debug_subcode): Add
27484 GIMPLE_DEBUG_INLINE_ENTRY.
27485 (gimple_build_debug_inline_entry): Declare.
27486 (gimple_debug_inline_entry_p): New.
27487 (gimple_debug_nonbind_marker_p): Adjust.
27488 * insn-notes.def (INLINE_ENTRY): New.
27489 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
27490 inline entry marker notes.
27491 (print_insn): Likewise.
27492 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
27493 (INSN_DEBUG_MARKER_KIND): Likewise.
27494 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
27495 * tree-inline.c (expand_call_inline): Build and insert
27496 debug_inline_entry stmt.
27497 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
27498 inline entry blocks early, if nonbind markers are enabled.
27499 (dump_scope_block): Dump fragment info.
27500 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
27501 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
27502 (gimple_build_debug_inline_entry): New.
27503 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
27504 Enable/disable inline entry points too.
27505 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
27506 (DEBUG_INSN): Describe inline entry markers.
27508 * common.opt (gvariable-location-views): New.
27509 (gvariable-location-views=incompat5): New.
27510 * config.in: Rebuilt.
27511 * configure: Rebuilt.
27512 * configure.ac: Test assembler for view support.
27513 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
27514 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
27515 * dwarf2out.c (var_loc_view): New typedef.
27516 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
27517 (dwarf2out_locviews_in_attribute): New.
27518 (dwarf2out_locviews_in_loclist): New.
27519 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
27520 (enum dw_line_info_opcode): Add LI_adv_address.
27521 (struct dw_line_info_table): Add view.
27522 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
27523 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
27524 (zero_view_p): New variable.
27525 (ZERO_VIEW_P): New macro.
27526 (output_asm_line_debug_info): New.
27527 (struct var_loc_node): Add view.
27528 (add_AT_view_list, AT_loc_list): New.
27529 (add_var_loc_to_decl): Add view param. Test it against last.
27530 (new_loc_list): Add view params. Record them.
27531 (AT_loc_list_ptr): Handle loc and view lists.
27532 (view_list_to_loc_list_val_node): New.
27533 (print_dw_val): Handle dw_val_class_view_list.
27534 (size_of_die): Likewise.
27535 (value_format): Likewise.
27536 (loc_list_has_views): New.
27537 (gen_llsym): Set vl_symbol too.
27538 (maybe_gen_llsym, skip_loc_list_entry): New.
27539 (dwarf2out_maybe_output_loclist_view_pair): New.
27540 (output_loc_list): Output view list or entries too.
27541 (output_view_list_offset): New.
27542 (output_die): Handle dw_val_class_view_list.
27543 (output_dwarf_version): New.
27544 (output_compilation_unit_header): Use it.
27545 (output_skeleton_debug_sections): Likewise.
27546 (output_rnglists, output_line_info): Likewise.
27547 (output_pubnames, output_aranges): Update version comments.
27548 (output_one_line_info_table): Output view numbers in asm comments.
27549 (dw_loc_list): Determine current endview, pass it to new_loc_list.
27550 Call maybe_gen_llsym.
27551 (loc_list_from_tree_1): Adjust.
27552 (add_AT_location_description): Create view list attribute if
27553 needed, check it's absent otherwise.
27554 (convert_cfa_to_fb_loc_list): Adjust.
27555 (maybe_emit_file): Call output_asm_line_debug_info for test.
27556 (dwarf2out_var_location): Reset views as needed. Precompute
27557 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
27558 attribute. Set view.
27559 (new_line_info_table): Reset next view.
27560 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
27561 (dwarf2out_source_line): Likewise. Output view resets and labels to
27562 the assembler, or select appropriate line info opcodes.
27563 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
27564 (optimize_string_length): Catch it. Adjust.
27565 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
27566 dw_val_class_view_list, and remove it if no longer needed.
27567 (hash_loc_list): Hash view numbers.
27568 (loc_list_hasher::equal): Compare them.
27569 (optimize_location_lists): Check whether a view list symbol is
27570 needed, and whether the locview attribute is present, and
27571 whether they match. Remove the locview attribute if no longer
27573 (index_location_lists): Call skip_loc_list_entry for test.
27574 (dwarf2out_finish): Call output_asm_line_debug_info for test.
27575 Use output_dwarf_version.
27576 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
27577 (struct dw_val_node): Add val_view_list.
27578 * final.c (SEEN_NEXT_VIEW): New.
27579 (set_next_view_needed): New.
27580 (clear_next_view_needed): New.
27581 (maybe_output_next_view): New.
27582 (final_start_function): Rename to...
27583 (final_start_function_1): ... this. Take pointer to FIRST,
27584 add SEEN parameter. Emit param bindings in the initial view.
27585 (final_start_function): Reintroduce SEEN-less interface.
27586 (final): Rename to...
27587 (final_1): ... this. Take SEEN parameter. Output final pending
27588 next view at the end.
27589 (final): Reintroduce seen-less interface.
27590 (final_scan_insn): Output pending next view before switching
27591 sections or ending a block. Mark the next view as needed when
27592 outputting variable locations. Notify debug backend of section
27593 changes, and of location view changes.
27594 (rest_of_handle_final): Adjust.
27595 * toplev.c (process_options): Autodetect value for debug variable
27596 location views option. Warn on incompat5 without -gdwarf-5.
27597 * doc/invoke.texi (gvariable-location-views): New.
27598 (gvariable-location-views=incompat5): New.
27599 (gno-variable-location-views): New.
27601 2018-02-08 David Malcolm <dmalcolm@redhat.com>
27603 PR tree-optimization/84136
27604 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
27605 that the result of find_edge is non-NULL.
27607 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
27610 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
27611 storing integer register in SImode. Fix cost of 256 and 512
27612 byte aligned SSE register store.
27614 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
27616 * config/i386/i386.c (ix86_multiplication_cost): Fix
27617 multiplication cost for TARGET_AVX512DQ.
27619 2018-02-08 Marek Polacek <polacek@redhat.com>
27621 PR tree-optimization/84238
27622 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
27625 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
27627 PR tree-optimization/84265
27628 * tree-vect-stmts.c (vectorizable_store): Don't treat
27629 VMAT_CONTIGUOUS accesses as grouped.
27630 (vectorizable_load): Likewise.
27632 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
27634 PR tree-optimization/81635
27635 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
27636 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
27637 (test_round_for_mask): New functions.
27638 (wide_int_cc_tests): Call test_round_for_mask.
27639 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
27640 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
27641 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
27642 range returned by get_range_info.
27644 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
27647 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
27648 * symtab.c: Include builtins.h
27649 (symtab_node::output_to_lto_symbol_table_p): Move here
27650 from lto-streamer-out.c:output_symbol_p.
27651 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
27652 (output_symbol_p): Move all logic to symtab.c
27653 (produce_symtab): Update.
27655 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
27657 * config/s390/s390-opts.h (enum indirect_branch): Define.
27658 * config/s390/s390-protos.h (s390_return_addr_from_memory)
27659 (s390_indirect_branch_via_thunk)
27660 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
27661 (enum s390_indirect_branch_type): Define.
27662 * config/s390/s390.c (struct s390_frame_layout, struct
27663 machine_function): Remove.
27664 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
27665 (indirect_branch_table_label_no, indirect_branch_table_name):
27667 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
27668 (enum s390_indirect_branch_option): Define.
27669 (s390_return_addr_from_memory): New function.
27670 (s390_handle_string_attribute): New function.
27671 (s390_attribute_table): Add new attribute handler.
27672 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
27673 (s390_indirect_branch_via_thunk): New function.
27674 (s390_indirect_branch_via_inline_thunk): New function.
27675 (s390_function_ok_for_sibcall): When jumping via thunk disallow
27676 sibling call optimization for non z10 compiles.
27677 (s390_emit_call): Force indirect branch target to be a single
27678 register. Add r1 clobber for non-z10 compiles.
27679 (s390_emit_epilogue): Emit return jump via return_use expander.
27680 (s390_reorg): Handle JUMP_INSNs as execute targets.
27681 (s390_option_override_internal): Perform validity checks for the
27682 new command line options.
27683 (s390_indirect_branch_attrvalue): New function.
27684 (s390_indirect_branch_settings): New function.
27685 (s390_set_current_function): Invoke s390_indirect_branch_settings.
27686 (s390_output_indirect_thunk_function): New function.
27687 (s390_code_end): Implement target hook.
27688 (s390_case_values_threshold): Implement target hook.
27689 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
27691 * config/s390/s390.h (struct s390_frame_layout)
27692 (struct machine_function): Move here from s390.c.
27693 (TARGET_INDIRECT_BRANCH_NOBP_RET)
27694 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
27695 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
27696 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
27697 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
27698 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
27699 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
27700 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
27701 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
27702 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
27703 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
27704 (mnemonic attribute): Add values which aren't recognized
27706 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
27707 pattern for branch conversion. Fix mnemonic attribute.
27708 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
27709 indirect branch via thunk if requested.
27710 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
27711 ("*indirect_jump"): Disable for branch conversion using out of
27713 ("indirect_jump_via_thunk<mode>_z10")
27714 ("indirect_jump_via_thunk<mode>")
27715 ("indirect_jump_via_inlinethunk<mode>_z10")
27716 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
27717 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
27718 ("casesi_jump_via_inlinethunk<mode>_z10")
27719 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
27720 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
27721 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
27722 ("*indirect2_jump"): Disable for branch conversion.
27723 ("casesi_jump"): Turn into expander and expand patterns for branch
27725 ("return_use"): New expander.
27726 ("*return"): Emit return via thunk and rename it to ...
27727 ("*return<mode>"): ... this one.
27728 * config/s390/s390.opt: Add new options and and enum for the
27731 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
27733 * lra-constraints.c (match_reload): Unconditionally use
27734 gen_lowpart_SUBREG, rather than selecting between that
27735 and equivalent gen_rtx_SUBREG code.
27737 2018-02-08 Richard Biener <rguenther@suse.de>
27739 PR tree-optimization/84233
27740 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
27741 changed flag instead of boguously re-using phi_inserted.
27743 2018-02-08 Martin Jambor <mjambor@suse.cz>
27745 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
27746 static local variables.
27748 2018-02-08 Richard Biener <rguenther@suse.de>
27750 PR tree-optimization/84278
27751 * tree-vect-stmts.c (vectorizable_store): When looking for
27752 smaller vector types to perform grouped strided loads/stores
27753 make sure the mode is supported by the target.
27754 (vectorizable_load): Likewise.
27756 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
27758 * config/aarch64/aarch64.c (aarch64_components_for_bb):
27759 Increase LDP/STP opportunities by adding adjacent callee-saves.
27761 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
27763 PR rtl-optimization/84068
27764 PR rtl-optimization/83459
27765 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
27767 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
27769 PR tree-optimization/84224
27770 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
27771 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
27772 non-zero arguments.
27774 2018-02-07 Iain Sandoe <iain@codesourcery.com>
27777 * config/rs6000/altivec.md (*restore_world): Remove LR use.
27778 * config/rs6000/predicates.md (restore_world_operation): Adjust op
27779 count, remove one USE.
27781 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
27783 * doc/install.texi (Configuration): Document the
27784 --with-long-double-format={ibm,ieee} PowerPC configuration
27788 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
27789 Convert from define_expand to be define_insn_and_split. Rework
27790 float/double/_Float128 conversions to QI/HI/SImode to work with
27791 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
27792 conversions to QI/HImode types did a store and then a load to
27793 truncate the value. For conversions to VSX registers, don't split
27794 the insn, instead emit the code directly. Use the code iterator
27795 any_fix to combine signed and unsigned conversions.
27796 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
27797 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
27798 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
27799 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
27800 (fix_<mode>di2_hw): Likewise.
27801 (fixuns_<mode>di2_hw): Likewise.
27802 (fix_<mode>si2_hw): Likewise.
27803 (fixuns_<mode>si2_hw): Likewise.
27804 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
27805 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
27806 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
27807 fix<uns>_trunc<SFDF:mode>si2_p8.
27808 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
27809 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
27810 (fix<uns>_<mode>_mem): Likewise.
27811 (fctiw<u>z_<mode>_mem): Likewise.
27812 (fix<uns>_<mode>_mem): Likewise.
27813 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
27814 the register allocator from doing a direct move to the GPRs to do
27815 a store, and instead use the ISA 3.0 store byte/half-word from
27816 vector register instruction. For IEEE 128-bit floating point,
27817 also optimize stores of 32-bit ints.
27818 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
27820 2018-02-07 Alan Hayward <alan.hayward@arm.com>
27822 * genextract.c (push_pathstr_operand): New function to support
27824 (walk_rtx): Call push_pathstr_operand.
27825 (print_path): Support [a-zA-Z].
27827 2018-02-07 Richard Biener <rguenther@suse.de>
27829 PR tree-optimization/84037
27830 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
27831 (cse_and_gimplify_to_preheader): Declare.
27832 (vect_get_place_in_interleaving_chain): Likewise.
27833 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
27835 (_loop_vec_info::~_loop_vec_info): Delete it.
27836 (cse_and_gimplify_to_preheader): New function.
27837 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
27838 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
27839 (vectorizable_load): Likewise. For grouped stores always base
27840 the IV on the first element.
27841 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
27842 condition before gimplifying.
27844 2018-02-07 Jakub Jelinek <jakub@redhat.com>
27846 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
27847 *DIV_EXPR and *MOD_EXPR.
27849 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
27852 * config/i386/i386.c (ix86_option_override_internal): Mask out
27853 the CF_SET bit when checking -fcf-protection.
27855 2018-02-07 Tom de Vries <tom@codesourcery.com>
27858 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
27861 2018-02-07 Richard Biener <rguenther@suse.de>
27863 PR tree-optimization/84204
27864 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
27867 PR tree-optimization/84205
27868 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
27869 special-case isl_ast_op_zdiv_r.
27871 PR tree-optimization/84223
27872 * graphite-scop-detection.c (gather_bbs::before_dom_children):
27873 Only add conditions from within the region.
27874 (gather_bbs::after_dom_children): Adjust.
27876 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
27879 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
27880 * config/avr/avr.md: Only post-reload split REG-REG moves if
27881 either register is GENERAL_REG_P.
27883 2018-02-07 Jakub Jelinek <jakub@redhat.com>
27885 PR tree-optimization/84235
27886 * tree-ssa-scopedtables.c
27887 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
27888 if the subtraction is performed in floating point type where NaNs are
27889 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
27890 build 1. Formatting fix.
27892 2018-02-06 Jakub Jelinek <jakub@redhat.com>
27895 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
27896 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
27897 and skip it regardless of bb boundaries. Use CALL_P macro,
27898 don't test INSN_P (insn) together with CALL_P or JUMP_P check
27899 unnecessarily, formatting fix.
27901 2018-02-06 Michael Collison <michael.collison@arm.com>
27903 * config/arm/thumb2.md:
27904 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
27905 (*thumb_mov_notscc): Ditto.
27907 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
27910 * config/rs6000/rs6000.md (su code attribute): Use "u" for
27911 unsigned_fix, not "s".
27913 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
27915 * configure.ac (gcc_fn_eh_frame_ro): New function.
27916 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
27917 correct .eh_frame permissions.
27918 * configure: Regenerate.
27920 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
27922 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
27923 irrelevant options.
27925 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27927 * config/rs6000/rs6000.c (rs6000_option_override_internal):
27928 Display warning message for -mno-speculate-indirect-jumps.
27930 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
27932 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
27934 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
27936 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
27938 PR tree-optimization/84225
27939 * tree-eh.c (find_trapping_overflow): Only call
27940 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
27942 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
27945 * config/i386/i386.c: Reimplement the check of possible options
27946 -mibt/-mshstk conbination. Change error messages.
27947 * doc/invoke.texi: Fix a typo: remove extra '='.
27949 2018-02-06 Marek Polacek <polacek@redhat.com>
27951 PR tree-optimization/84228
27952 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
27954 2018-02-06 Tamar Christina <tamar.christina@arm.com>
27957 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
27958 emitted arch directives.
27959 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
27960 __ARM_FEATURE_COPROC before changing architectures.
27962 2018-02-06 Richard Biener <rguenther@suse.de>
27964 * config/i386/i386.c (print_reg): Fix typo.
27965 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
27967 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
27969 * configure: Regenerate.
27971 2018-02-05 Martin Sebor <msebor@redhat.com>
27973 PR tree-optimization/83369
27974 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
27977 2018-02-05 Martin Liska <mliska@suse.cz>
27979 * doc/invoke.texi: Cherry-pick upstream r323995.
27981 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
27983 * ira.c (ira_init_register_move_cost): Adjust comment.
27985 2018-02-05 Martin Liska <mliska@suse.cz>
27987 PR gcov-profile/84137
27988 * doc/gcov.texi: Fix typo in documentation.
27990 2018-02-05 Martin Liska <mliska@suse.cz>
27992 PR gcov-profile/83879
27993 * doc/gcov.texi: Document necessity of --dynamic-list-data when
27994 using dlopen functionality.
27996 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
27998 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
27999 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
28000 _mm_maskz_range_ss, _mm_mask_range_round_ss,
28001 _mm_maskz_range_round_ss): New intrinsics.
28002 (__builtin_ia32_rangesd128_round)
28003 (__builtin_ia32_rangess128_round): Remove.
28004 (__builtin_ia32_rangesd128_mask_round,
28005 __builtin_ia32_rangess128_mask_round): New builtins.
28006 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
28007 __builtin_ia32_rangess128_round): Remove.
28008 (__builtin_ia32_rangesd128_mask_round,
28009 __builtin_ia32_rangess128_mask_round): New builtins.
28010 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
28011 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
28012 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
28013 "<round_saeonly_constraint>")): Changed to ...
28014 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
28015 "<round_saeonly_scalar_constraint>")): ... this.
28016 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
28017 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
28018 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
28019 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
28020 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
28022 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
28024 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
28026 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
28027 Remove all values except native, 8540 and 8548.
28029 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
28031 * config/i386/i386.c (ix86_output_function_return): Pass
28032 INVALID_REGNUM, instead of -1, as invalid register number to
28033 indirect_thunk_name and output_indirect_thunk.
28035 2018-02-02 Julia Koval <julia.koval@intel.com>
28037 * config.gcc: Add -march=icelake.
28038 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
28039 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
28040 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
28041 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
28042 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
28043 (processor_target_table): Add icelake.
28044 (ix86_option_override_internal): Handle new PTAs.
28045 (get_builtin_code_for_version): Handle icelake.
28046 (M_INTEL_COREI7_ICELAKE): New.
28047 (fold_builtin_cpu): Handle icelake.
28048 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
28049 * doc/invoke.texi: Add -march=icelake.
28051 2018-02-02 Julia Koval <julia.koval@intel.com>
28053 * config/i386/i386.c (ix86_option_override_internal): Change flags type
28054 to wide_int_bitmask.
28055 * wide-int-bitmask.h: New.
28057 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
28060 * config/i386/i386.md: Replace Pmode with word_mode in
28061 builtin_setjmp_setup and builtin_longjmp to support x32.
28063 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
28067 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
28069 (rs6000_supported_cpu_names): New static variable.
28070 (linux_cpu_translation_table): Likewise.
28071 (elf_platform) <cpu>: Define new static variable and use it.
28072 Translate kernel AT_PLATFORM name to canonical name if needed.
28073 Error if platform name is unknown.
28075 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
28078 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
28080 2018-02-01 Jeff Law <law@redhat.com>
28083 * config/i386/i386.c (release_scratch_register_on_entry): Add new
28084 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
28085 the scratch if RELEASE_VIA_POP is false.
28086 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
28087 If we have to save a temporary register, decrement SIZE appropriately.
28088 Pass new arguments to release_scratch_register_on_entry.
28089 (ix86_adjust_stack_and_probe): Likewise.
28090 (ix86_emit_probe_stack_range): Pass new arguments to
28091 release_scratch_register_on_entry.
28093 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
28095 PR rtl-optimization/84157
28096 * combine.c (change_zero_ext): Use REG_P predicate in
28097 front of HARD_REGISTER_P predicate.
28099 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
28101 * config/avr/avr.c (avr_option_override): Move disabling of
28102 -fdelete-null-pointer-checks to...
28103 * common/config/avr/avr-common.c (avr_option_optimization_table):
28106 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
28108 PR tree-optimization/81635
28109 * tree-data-ref.c (split_constant_offset_1): For types that
28110 wrap on overflow, try to use range info to prove that wrapping
28113 2018-02-01 Renlin Li <renlin.li@arm.com>
28116 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
28117 TAILCALL_ADDR_REGS.
28118 (aarch64_register_move_cost): Likewise.
28119 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
28120 TAILCALL_ADDR_REGS.
28121 (REG_CLASS_NAMES): Likewise.
28122 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
28123 TAILCALL_ADDR_REGS. Remove IP registers.
28124 * config/aarch64/aarch64.md (Ucs): Update register constraint.
28126 2018-02-01 Richard Biener <rguenther@suse.de>
28128 * domwalk.h (dom_walker::dom_walker): Add additional constructor
28129 for specifying RPO order and allow NULL for that.
28130 * domwalk.c (dom_walker::dom_walker): Likewise.
28131 (dom_walker::walk): Handle NULL RPO order.
28132 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
28134 (rewrite_update_dom_walker): Likewise.
28135 (mark_def_dom_walker): Likewise.
28137 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
28139 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
28140 (aarch64_maybe_expand_sve_subreg_move): Declare.
28141 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
28142 * config/aarch64/predicates.md (aarch64_any_register_operand): New
28144 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
28145 that are semantically a reverse operation.
28146 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
28147 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
28148 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
28150 (aarch64_can_change_mode_class): For big-endian, forbid changes
28151 between two SVE modes if they have different element sizes.
28153 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
28155 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
28156 the TImode handling for big-endian targets.
28158 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
28160 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
28161 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
28163 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
28164 Remove BSWAP handing for big-endian targets and use the form of
28165 LD1RQ appropariate for the mode.
28167 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
28169 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
28170 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
28171 duplicated element.
28173 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
28176 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
28177 check for operands that need to go through aarch64_sve_reload_be.
28179 2018-02-01 Jakub Jelinek <jakub@redhat.com>
28181 PR tree-optimization/81661
28182 PR tree-optimization/84117
28183 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
28184 * tree-eh.c: Include gimplify.h.
28185 (find_trapping_overflow, replace_trapping_overflow,
28186 rewrite_to_non_trapping_overflow): New functions.
28187 * tree-vect-loop.c: Include tree-eh.h.
28188 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
28189 * tree-data-ref.c: Include tree-eh.h.
28190 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
28192 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
28194 PR rtl-optimization/84123
28195 * combine.c (change_zero_ext): Check if hard register satisfies
28196 can_change_dest_mode before calling gen_lowpart_SUBREG.
28198 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
28201 * ira.c (ira_init_register_move_cost): Remove assert.
28203 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
28205 PR rtl-optimization/84071
28206 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
28207 * doc/tm.texi: Regenerate.
28209 2018-01-31 Richard Biener <rguenther@suse.de>
28211 PR tree-optimization/84132
28212 * tree-data-ref.c (analyze_miv_subscript): Properly
28213 check whether evolution_function_is_affine_multivariate_p
28214 before calling gcd_of_steps_may_divide_p.
28216 2018-01-31 Julia Koval <julia.koval@intel.com>
28219 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
28220 * config/i386/i386.md (rdpid_rex64) New.
28221 (rdpid): Make 32bit only.
28223 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
28226 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
28227 an IDENTIFIER_NODE for FUNCTION_TYPE's.
28229 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
28232 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
28234 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
28236 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
28238 PR rtl-optimization/84071
28239 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
28240 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
28242 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
28244 * config/arc/arc.c (arc_handle_aux_attribute): New function.
28245 (arc_attribute_table): Add 'aux' attribute.
28246 (arc_in_small_data_p): Consider aux like variables.
28247 (arc_is_aux_reg_p): New function.
28248 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
28249 (arc_get_aux_arg): New function.
28250 (prepare_move_operands): Handle aux-register access.
28251 (arc_handle_aux_attribute): New function.
28252 * doc/extend.texi (ARC Variable attributes): Add subsection.
28254 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
28256 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
28257 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
28258 (arc_attribute_table): Add 'uncached' attribute.
28259 (arc_print_operand): Print '.di' flag for uncached memory
28261 (arc_in_small_data_p): Do not consider for small data the uncached
28263 (arc_is_uncached_mem_p): New function.
28264 * config/arc/predicates.md (compact_store_memory_operand): Check
28265 for uncached memory accesses.
28266 (nonvol_nonimm_operand): Likewise.
28267 * doc/extend.texi (ARC Type Attribute): New subsection.
28269 2018-01-31 Jakub Jelinek <jakub@redhat.com>
28272 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
28273 falign-loops=): Add Optimization flag.
28275 2018-01-30 Jeff Law <law@redhat.com>
28278 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
28279 INT_REGISTERS_SAVED. Check it prior to calling
28280 get_scratch_register_on_entry.
28281 (ix86_adjust_stack_and_probe): Similarly.
28282 (ix86_emit_probe_stack_range): Similarly.
28283 (ix86_expand_prologue): Corresponding changes.
28285 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
28288 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
28289 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
28291 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
28294 * lra-constraints.c (curr_insn_transform): Process AND in the
28297 2018-01-30 Jakub Jelinek <jakub@redhat.com>
28299 PR rtl-optimization/83986
28300 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
28301 dependence against last_pending_memory_flush in addition to
28302 pending_jump_insns.
28304 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
28306 PR tree-optimization/81611
28307 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
28310 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
28313 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
28316 2018-01-30 Richard Biener <rguenther@suse.de>
28317 Jakub Jelinek <jakub@redhat.com>
28319 PR tree-optimization/84111
28320 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
28321 inner loops added during recursion, as they don't have up-to-date
28324 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
28327 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
28328 (can_inline_edge_by_limits_p): ... here.
28329 (can_early_inline_edge_p, check_callers,
28330 update_caller_keys, update_callee_keys, recursive_inlining,
28331 add_new_edges_to_heap, speculation_useful_p,
28332 inline_small_functions,
28333 inline_small_functions, flatten_function,
28334 inline_to_all_callers_1): Update.
28336 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
28338 * profile-count.c (profile_count::combine_with_ipa_count): Handle
28341 2018-01-30 Richard Biener <rguenther@suse.de>
28343 PR tree-optimization/83008
28344 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
28345 invariant and constant vector uses in stmts when they need
28346 more than one stmt.
28348 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
28351 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
28352 * configure: Regenerate.
28354 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
28356 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
28358 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
28359 Use gen_rtx_REG rather than gen_lowpart.
28361 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
28363 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
28364 rather than 0 when creating partial subregs.
28366 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
28368 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
28371 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
28374 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
28375 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
28376 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
28377 flags. This restores the settings used before the 2017-07-24.
28378 Turning off pre increment/decrement/modify allows IVOPTS to
28379 optimize DF/SF loops where the index is an int.
28381 2018-01-29 Richard Biener <rguenther@suse.de>
28382 Kelvin Nilsen <kelvin@gcc.gnu.org>
28385 * tree-vect-stmts.c (vectorizable_call): Don't call
28386 targetm.vectorize_builtin_md_vectorized_function if callee is
28389 2018-01-22 Carl Love <cel@us.ibm.com>
28391 * doc/extend.tex: Fix typo in second arg in
28392 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
28394 2018-01-29 Richard Biener <rguenther@suse.de>
28396 PR tree-optimization/84086
28397 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
28398 (flush_ssaname_freelist): When SSA names were released reset
28399 the SCEV hash table.
28401 2018-01-29 Richard Biener <rguenther@suse.de>
28403 PR tree-optimization/84057
28404 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
28405 removed paths when removing edges.
28407 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
28409 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
28410 -mfunction-return=@var{choice}.
28412 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
28414 PR diagnostic/84034
28415 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
28416 Handle CR like TAB.
28417 (layout::print_source_line): Likewise.
28418 (test_get_line_width_without_trailing_whitespace): Add test cases.
28420 2018-01-27 Jakub Jelinek <jakub@redhat.com>
28422 PR middle-end/84040
28423 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
28426 2018-01-26 Jim Wilson <jimw@sifive.com>
28428 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
28430 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
28433 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28435 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
28436 and CMP + SUB-immediate -> SUBS.
28438 2018-01-26 Martin Sebor <msebor@redhat.com>
28440 PR tree-optimization/83896
28441 * tree-ssa-strlen.c (get_string_len): Rename...
28442 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
28443 Avoid assuming length is constant.
28444 (handle_char_store): Use HOST_WIDE_INT for string length.
28446 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
28449 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
28450 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
28452 2018-01-26 Richard Biener <rguenther@suse.de>
28454 PR rtl-optimization/84003
28455 * dse.c (record_store): Only record redundant stores when
28456 the earlier store aliases at least all accesses the later one does.
28458 2018-01-26 Jakub Jelinek <jakub@redhat.com>
28460 PR rtl-optimization/83985
28461 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
28462 REG_CFA_RESTORE insns.
28463 (delete_unmarked_insns): Don't ignore separate shrink wrapping
28464 REG_CFA_RESTORE insns here.
28467 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
28468 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
28470 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
28472 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
28473 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
28474 (arc_init): Likewise.
28475 (arc_override_options): Likewise.
28476 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
28478 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
28480 * config/arc/arc.h (TARGET_DBNZ): Define.
28481 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
28482 properly set the tune attribute.
28483 (dbnz): Use TARGET_DBNZ guard.
28484 * config/arc/arc.opt (mtune): Add core3 option.
28486 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
28488 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
28489 recognize new pic like addresses.
28490 (arc_delegitimize_address): Clean up.
28492 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
28494 * config/arc/arc-arches.def: Option mrf16 valid for all
28496 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
28497 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
28498 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
28499 * config/arc/arc-tables.opt: Regenerate.
28500 * config/arc/arc.c (arc_conditional_register_usage): Handle
28501 reduced register file case.
28502 (arc_file_start): Set must have build attributes.
28503 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
28504 mrf16 option value.
28505 * config/arc/arc.opt (mrf16): Add new option.
28506 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
28507 * config/arc/genmultilib.awk: Handle new mrf16 option.
28508 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
28509 * config/arc/t-multilib: Regenerate.
28510 * doc/invoke.texi (ARC Options): Document mrf16 option.
28512 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
28514 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
28515 * config/arc/arc.c (arc_handle_secure_attribute): New function.
28516 (arc_attribute_table): Add 'secure_call' attribute.
28517 (arc_print_operand): Print secure call operand.
28518 (arc_function_ok_for_sibcall): Don't optimize tail calls when
28520 (arc_is_secure_call_p): New function. * config/arc/arc.md
28521 (call_i): Add support for sjli instruction.
28522 (call_value_i): Likewise.
28523 * config/arc/constraints.md (Csc): New constraint.
28525 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
28526 John Eric Martin <John.Martin@emmicro-us.com>
28528 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
28529 * config/arc/arc.c (_arc_jli_section): New struct.
28530 (arc_jli_section): New type.
28531 (rc_jli_sections): New static variable.
28532 (arc_handle_jli_attribute): New function.
28533 (arc_attribute_table): Add jli_always and jli_fixed attribute.
28534 (arc_file_end): New function.
28535 (TARGET_ASM_FILE_END): Define.
28536 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
28537 (arc_add_jli_section): New function.
28538 (jli_call_scan): Likewise.
28539 (arc_reorg): Call jli_call_scan.
28540 (arc_output_addsi): Remove 'S' from printing asm operand.
28541 (arc_is_jli_call_p): New function.
28542 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
28544 (movhi_insn): Likewise.
28545 (movsi_insn): Likewise.
28546 (movsi_set_cc_insn): Likewise.
28547 (loadqi_update): Likewise.
28548 (load_zeroextendqisi_update): Likewise.
28549 (load_signextendqisi_update): Likewise.
28550 (loadhi_update): Likewise.
28551 (load_zeroextendhisi_update): Likewise.
28552 (load_signextendhisi_update): Likewise.
28553 (loadsi_update): Likewise.
28554 (loadsf_update): Likewise.
28555 (movsicc_insn): Likewise.
28556 (bset_insn): Likewise.
28557 (bxor_insn): Likewise.
28558 (bclr_insn): Likewise.
28559 (bmsk_insn): Likewise.
28560 (bicsi3_insn): Likewise.
28561 (cmpsi_cc_c_insn): Likewise.
28562 (movsi_ne): Likewise.
28563 (movsi_cond_exec): Likewise.
28564 (clrsbsi2): Likewise.
28565 (norm_f): Likewise.
28575 (call_i): Remove 'S' asm letter, add jli instruction.
28576 (call_value_i): Likewise.
28577 * config/arc/arc.op (mjli-always): New option.
28578 * config/arc/constraints.md (Cji): New constraint.
28579 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
28581 (subsf3_fpx): Likewise.
28582 (mulsf3_fpx): Likewise.
28583 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
28585 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
28586 function attrbutes.
28587 * doc/invoke.texi (ARC): Document mjli-always option.
28589 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
28591 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
28592 avoid addition with 0 and use incw and decw where possible.
28594 2018-01-26 Richard Biener <rguenther@suse.de>
28596 PR tree-optimization/81082
28597 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
28598 association if it requires casting to unsigned.
28599 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
28600 from fold_plusminus_mult_expr to catch important cases late when
28601 range info is available.
28603 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
28605 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
28606 * configure.ac (hidden_linkonce): New test.
28607 * configure: Regenerate.
28608 * config.in: Regenerate.
28610 2018-01-26 Julia Koval <julia.koval@intel.com>
28612 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
28613 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
28614 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
28615 _mm_mask_bitshuffle_epi64_mask): Fix type.
28616 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
28617 USI_FTYPE_V4DI_V4DI_USI): Remove.
28618 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
28619 __builtin_ia32_vpshufbitqmb256_mask,
28620 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
28621 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
28622 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
28624 2018-01-26 Alan Modra <amodra@gmail.com>
28627 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
28628 UNSPEC_VBPERMQ. Sort other unspecs.
28630 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
28632 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
28634 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
28636 PR middle-end/83055
28637 * predict.c (drop_profile): Do not push/pop cfun; update also
28639 (handle_missing_profiles): Fix logic looking for zero profiles.
28641 2018-01-25 Jakub Jelinek <jakub@redhat.com>
28643 PR middle-end/83977
28644 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
28645 on functions with #pragma omp declare simd or functions with simd
28647 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
28648 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
28649 Remove trailing \n from warning_at calls.
28651 2018-01-25 Tom de Vries <tom@codesourcery.com>
28654 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
28655 for neutered workers.
28657 2018-01-24 Joseph Myers <joseph@codesourcery.com>
28660 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
28661 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
28663 2018-01-24 Jeff Law <law@redhat.com>
28666 * i386.c (get_probe_interval): Move to earlier point.
28667 (ix86_compute_frame_layout): If -fstack-clash-protection and
28668 the frame is larger than the probe interval, then use pushes
28669 to save registers rather than reg->mem moves.
28670 (ix86_expand_prologue): Remove conditional for int_registers_saved
28673 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
28676 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
28677 min/max for never referenced object.
28679 2018-01-24 Jakub Jelinek <jakub@redhat.com>
28681 PR middle-end/83977
28682 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
28684 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
28685 attributes from DECL_ATTRIBUTES (decl) without affecting
28686 DECL_ATTRIBUTES (current_function_decl).
28687 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
28688 functions with non-NULL DECL_ABSTRACT_ORIGIN.
28690 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
28692 PR tree-optimization/83979
28693 * fold-const.c (fold_comparison): Use constant_boolean_node
28694 instead of boolean_{true,false}_node.
28696 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
28698 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
28701 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
28703 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
28704 Simplify the clause that sets the length attribute.
28705 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
28706 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
28707 clause that sets the length attribute.
28708 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
28710 2018-01-24 Tom de Vries <tom@codesourcery.com>
28713 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
28714 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
28715 Add strict parameter.
28716 (prevent_branch_around_nothing): Insert dummy insn between branch to
28717 label and label with no ptx insn inbetween.
28718 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
28720 2018-01-24 Tom de Vries <tom@codesourcery.com>
28723 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
28724 for neutered threads in warp.
28725 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
28727 2018-01-24 Richard Biener <rguenther@suse.de>
28729 PR tree-optimization/83176
28730 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
28733 2018-01-24 Richard Biener <rguenther@suse.de>
28735 PR tree-optimization/82819
28736 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
28737 code generating pluses that are no-ops in the target precision.
28739 2018-01-24 Richard Biener <rguenther@suse.de>
28741 PR middle-end/84000
28742 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
28744 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
28746 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
28747 to merge probabilities.
28748 * predict.c (probably_never_executed): Also mark as cold functions
28749 with global 0 profile and guessed local profile.
28750 * profile-count.c (profile_probability::combine_with_count): New
28752 * profile-count.h (profile_probability::operator*,
28753 profile_probability::operator*=, profile_probability::operator/,
28754 profile_probability::operator/=): Reduce precision to adjusted
28755 and set value to guessed on contradictory divisions.
28756 (profile_probability::combine_with_freq): Remove.
28757 (profile_probability::combine_wiht_count): Declare.
28758 (profile_count::force_nonzero):: Set to adjusted.
28759 (profile_count::probability_in):: Set quality to adjusted.
28760 * tree-ssa-tail-merge.c (replace_block_by): Use
28761 combine_with_count.
28763 2018-01-23 Andrew Waterman <andrew@sifive.com>
28764 Jim Wilson <jimw@sifive.com>
28766 * config/riscv/riscv.c (riscv_stack_boundary): New.
28767 (riscv_option_override): Set riscv_stack_boundary. Handle
28768 riscv_preferred_stack_boundary_arg.
28769 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
28770 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
28771 (STACK_BOUNDARY): Set to riscv_stack_boundary.
28772 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
28773 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
28774 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
28776 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
28779 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
28780 of struct ix86_frame.
28781 (ix86_expand_epilogue): Likewise. Add a local variable for
28782 the reg_save_offset field in struct ix86_frame.
28784 2018-01-23 Bin Cheng <bin.cheng@arm.com>
28786 PR tree-optimization/82604
28787 * tree-loop-distribution.c (enum partition_kind): New enum item
28788 PKIND_PARTIAL_MEMSET.
28789 (partition_builtin_p): Support above new enum item.
28790 (generate_code_for_partition): Ditto.
28791 (compute_access_range): Differentiate cases that equality can be
28792 proven at all loops, the innermost loops or no loops.
28793 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
28794 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
28795 (finalize_partitions, distribute_loop): Don't fuse partition of
28796 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
28797 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
28798 parloop is enabled.
28800 2018-01-23 Martin Liska <mliska@suse.cz>
28802 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
28803 order to ignore the predictor.
28804 (PRED_POLYMORPHIC_CALL): Likewise.
28805 (PRED_RECURSIVE_CALL): Likewise.
28807 2018-01-23 Martin Liska <mliska@suse.cz>
28809 * tree-profile.c (tree_profiling): Print function header to
28810 aware reader which function we are working on.
28811 * value-prof.c (gimple_find_values_to_profile): Do not print
28812 not interesting value histograms.
28814 2018-01-23 Martin Liska <mliska@suse.cz>
28816 * profile-count.h (enum profile_quality): Add
28817 profile_uninitialized as the first value. Do not number values
28818 as they are zero based.
28819 (profile_count::verify): Update sanity check.
28820 (profile_probability::verify): Likewise.
28822 2018-01-23 Nathan Sidwell <nathan@acm.org>
28824 * doc/invoke.texi (ffor-scope): Deprecate.
28826 2018-01-23 David Malcolm <dmalcolm@redhat.com>
28828 PR tree-optimization/83510
28829 * domwalk.c (set_all_edges_as_executable): New function.
28830 (dom_walker::dom_walker): Convert bool param
28831 "skip_unreachable_blocks" to enum reachability. Move setup of
28832 edge flags to set_all_edges_as_executable and only do it when
28833 reachability is REACHABLE_BLOCKS.
28834 * domwalk.h (enum dom_walker::reachability): New enum.
28835 (dom_walker::dom_walker): Convert bool param
28836 "skip_unreachable_blocks" to enum reachability.
28837 (set_all_edges_as_executable): New decl.
28838 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
28839 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
28841 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
28842 but converting true to REACHABLE_BLOCKS.
28843 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
28845 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
28846 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
28847 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
28849 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
28850 if check_all_array_refs will be called.
28852 2018-01-23 David Malcolm <dmalcolm@redhat.com>
28854 * tree.c (selftest::test_location_wrappers): Add more test
28857 2018-01-23 David Malcolm <dmalcolm@redhat.com>
28859 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
28860 (selftest::test_bit_in_range): Likewise.
28862 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
28865 * doc/sourcebuild.texi (vect_float): Say that the selector
28866 only describes the situation when -funsafe-math-optimizations is on.
28867 (vect_float_strict): Document.
28869 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
28871 PR tree-optimization/83965
28872 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
28873 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
28874 instead of checking only for a reduction.
28875 (vect_recog_widen_sum_pattern): Likewise.
28877 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
28879 * predict.c (probably_never_executed): Only use precise profile info.
28880 (compute_function_frequency): Skip after inlining hack since we now
28881 have quality checking.
28883 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
28885 * profile-count.h (profile_probability::very_unlikely,
28886 profile_probability::unlikely, profile_probability::even): Set
28887 precision to guessed.
28889 2018-01-23 Richard Biener <rguenther@suse.de>
28891 PR tree-optimization/83963
28892 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
28893 Properly terminate dominator walk when crossing the exit edge not
28894 when visiting its source block.
28896 2018-01-23 Jakub Jelinek <jakub@redhat.com>
28899 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
28900 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
28902 2018-01-22 Jakub Jelinek <jakub@redhat.com>
28904 PR tree-optimization/83957
28905 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
28906 semicolon after for body surrounded by braces.
28908 PR tree-optimization/83081
28909 * profile-count.h (profile_probability::split): New method.
28910 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
28911 Use profile_probability::split.
28912 (do_compare_rtx_and_jump): Fix adjustment of probabilities
28913 when splitting a single conditional jump into 2.
28915 2018-01-22 David Malcolm <dmalcolm@redhat.com>
28917 PR tree-optimization/69452
28918 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
28921 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
28923 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
28924 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
28925 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
28927 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
28929 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
28931 * config/rl78/rl78.md (movdi): New define_expand.
28932 * config/rl78/rl78.c (rl78_split_movdi): New function.
28934 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
28937 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
28939 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
28940 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
28941 128-bit to produce an UNSPEC move to get the double word with the
28942 signbit and then a shift directly to do signbit.
28943 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
28944 implementation with a new version that just does either a direct
28945 move or a regular move. Move memory interface to separate insns.
28946 Move insns so they are next to the expander.
28947 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
28948 with signbit move. Split big and little endian case.
28949 (signbit<mode>2_dm_mem_le): Likewise.
28950 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
28951 (signbit<mode>2_dm2): Likewise.
28953 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
28955 * config/rl78/rl78.md (anddi3): New define_expand.
28957 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
28959 * config/rl78/rl78.md (umindi3): New define_expand.
28961 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
28963 * config/rl78/rl78.md (smindi3): New define_expand.
28965 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
28967 * config/rl78/rl78.md (smaxdi3): New define_expand.
28969 2018-01-22 Carl Love <cel@us.ibm.com>
28971 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
28972 LVX_V1TI): Add macro expansion.
28973 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
28974 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
28975 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
28976 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
28977 Change check to determine if the instruction is a byte reversing
28978 entry. Fix typo in comment.
28979 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
28980 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
28981 Add def_builtin calls for new builtins.
28982 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
28983 Add define_insn expansion.
28985 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
28987 * config/rl78/rl78.md (umaxdi3): New define_expand.
28989 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
28991 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
28992 for non-QImode registers.
28994 2018-01-22 Richard Biener <rguenther@suse.de>
28996 PR tree-optimization/83963
28997 * graphite-scop-detection.c (scop_detection::get_sese): Delay
28998 including the loop exit block.
28999 (scop_detection::merge_sese): Likewise.
29000 (scop_detection::add_scop): Do it here instead.
29002 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29004 * doc/sourcebuild.texi (arm_softfloat): Document.
29006 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
29009 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
29010 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
29011 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
29013 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
29014 David Edelsohn <dje.gcc@gmail.com>
29017 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
29018 Change "crset eq" to "crset 2".
29019 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
29020 (*call_indirect_aix<mode>_nospec): Likewise.
29021 (*call_value_indirect_aix<mode>_nospec): Likewise.
29022 (*call_indirect_elfv2<mode>_nospec): Likewise.
29023 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
29024 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
29025 change assembly output from . to $.
29026 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
29027 (indirect_jump<mode>_nospec): Change assembly output from . to $.
29028 (*tablejump<mode>_internal1_nospec): Likewise.
29030 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
29033 * config/sh/sh_optimize_sett_clrt.cc:
29034 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
29036 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
29038 PR tree-optimization/83940
29039 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
29040 offset_dt to vect_constant_def rather than vect_unknown_def_type.
29041 (vect_check_load_store_mask): Add a mask_dt_out parameter and
29042 use it to pass back the definition type.
29043 (vect_check_store_rhs): Likewise rhs_dt_out.
29044 (vect_build_gather_load_calls): Add a mask_dt argument and use
29045 it instead of a call to vect_is_simple_use.
29046 (vectorizable_store): Update calls to vect_check_load_store_mask
29047 and vect_check_store_rhs. Use the dt returned by the latter instead
29048 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
29049 instead of calls to vect_is_simple_use. Pass the scalar rather
29050 than the vector operand to vect_is_simple_use when handling
29051 second and subsequent copies of an rhs value.
29052 (vectorizable_load): Update calls to vect_check_load_store_mask
29053 and vect_build_gather_load_calls. Use the cached mask_dt and
29054 gs_info.offset_dt instead of calls to vect_is_simple_use.
29056 2018-01-20 Jakub Jelinek <jakub@redhat.com>
29058 PR middle-end/83945
29059 * tree-emutls.c: Include gimplify.h.
29060 (lower_emutls_2): New function.
29061 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
29062 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
29063 it before further processing.
29066 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
29067 UINTVAL (trueop1) instead of INTVAL (op1).
29069 2018-01-19 Jakub Jelinek <jakub@redhat.com>
29073 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
29074 INCOMING_FRAME_SP_OFFSET if not defined.
29075 (scan_trace): Add ENTRY argument. If true and
29076 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
29077 emit a note to adjust the CFA offset.
29078 (create_cfi_notes): Adjust scan_trace callers.
29079 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
29080 INCOMING_FRAME_SP_OFFSET in the CIE.
29081 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
29082 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
29084 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
29085 * doc/tm.texi: Regenerated.
29087 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
29089 PR rtl-optimization/83147
29090 * lra-constraints.c (remove_inheritance_pseudos): Use
29091 lra_substitute_pseudo_within_insn.
29093 2018-01-19 Tom de Vries <tom@codesourcery.com>
29094 Cesar Philippidis <cesar@codesourcery.com>
29097 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
29099 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
29102 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
29103 spaces for function labels.
29105 2018-01-19 Martin Liska <mliska@suse.cz>
29107 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
29108 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
29109 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
29110 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
29111 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
29112 (PRED_CONST_RETURN): Change from 69 to 65.
29113 (PRED_NULL_RETURN): Change from 91 to 71.
29114 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
29115 (PRED_LOOP_GUARD): Change from 66 to 73.
29117 2018-01-19 Martin Liska <mliska@suse.cz>
29119 * predict.c (predict_insn_def): Add new assert.
29120 (struct branch_predictor): Change type to signed integer.
29121 (test_prediction_value_range): Amend test to cover
29122 PROB_UNINITIALIZED.
29123 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
29124 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
29125 (PRED_LOOP_ITERATIONS_MAX): Likewise.
29126 (PRED_LOOP_IV_COMPARE): Likewise.
29127 * predict.h (PROB_UNINITIALIZED): Define new constant.
29129 2018-01-19 Martin Liska <mliska@suse.cz>
29131 * predict.c (dump_prediction): Add new format for
29132 analyze_brprob.py script which is enabled with -details
29134 * profile-count.h (precise_p): New function.
29136 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
29138 PR tree-optimization/83922
29139 * tree-vect-loop.c (vect_verify_full_masking): Return false if
29140 there are no statements that need masking.
29141 (vect_active_double_reduction_p): New function.
29142 (vect_analyze_loop_operations): Use it when handling phis that
29143 are not in the loop header.
29145 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
29147 PR tree-optimization/83914
29148 * tree-vect-loop.c (vectorizable_induction): Don't convert
29149 init_expr or apply the peeling adjustment for inductions
29150 that are nested within the vectorized loop.
29152 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29154 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
29157 2018-01-18 Jakub Jelinek <jakub@redhat.com>
29161 * function.h (gimplify_parameters): Add gimple_seq * argument.
29162 * function.c: Include gimple.h and options.h.
29163 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
29164 for the added local temporaries if needed.
29165 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
29166 if there are any parameter cleanups, wrap whole body into a
29167 try/finally with the cleanups.
29169 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
29172 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
29173 Use GET_MODE_CLASS for scalar floating point.
29175 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
29179 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
29180 Fix call of call_cgraph_insertion_hooks.
29182 2018-01-18 Martin Sebor <msebor@redhat.com>
29184 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
29186 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
29189 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
29192 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
29195 * common.opt: (-ffile-prefix-map): New option.
29196 * opts.c (common_handle_option): Defer it.
29197 * opts-global.c (handle_common_deferred_options): Handle it.
29198 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
29199 * file-prefix-map.h: New file.
29200 (remap_debug_filename, add_debug_prefix_map): ...here.
29201 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
29202 * final.c (debug_prefix_map, add_debug_prefix_map
29203 remap_debug_filename): Move to...
29204 * file-prefix-map.c: New file.
29205 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
29206 generalize, get rid of alloca(), use strrchr() instead of strchr().
29207 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
29208 Implement in terms of add_prefix_map().
29209 (remap_macro_filename, remap_debug_filename): Implement in term of
29211 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
29212 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
29213 * dbxout.c: Include file-prefix-map.h.
29214 * varasm.c: Likewise.
29215 * vmsdbgout.c: Likewise.
29216 * xcoffout.c: Likewise.
29217 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
29218 * doc/cppopts.texi (-fmacro-prefix-map): Document.
29219 * doc/invoke.texi (-ffile-prefix-map): Document.
29220 (-fdebug-prefix-map): Update description.
29222 2018-01-18 Martin Liska <mliska@suse.cz>
29224 * config/i386/i386.c (indirect_thunk_name): Document that also
29226 (output_indirect_thunk): Document why both instructions
29227 (pause and lfence) are generated.
29229 2018-01-18 Richard Biener <rguenther@suse.de>
29231 PR tree-optimization/83887
29232 * graphite-scop-detection.c
29233 (scop_detection::get_nearest_dom_with_single_entry): Remove.
29234 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
29235 (scop_detection::merge_sese): Re-implement with a flood-fill
29236 algorithm that properly finds a SESE region if it exists.
29238 2018-01-18 Jakub Jelinek <jakub@redhat.com>
29241 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
29242 pointer_diff optimizations use view_convert instead of convert.
29244 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
29246 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
29247 Generate different code for -mno-speculate-indirect-jumps.
29248 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
29249 (*call_indirect_aix<mode>): Disable for
29250 -mno-speculate-indirect-jumps.
29251 (*call_indirect_aix<mode>_nospec): New define_insn.
29252 (*call_value_indirect_aix<mode>): Disable for
29253 -mno-speculate-indirect-jumps.
29254 (*call_value_indirect_aix<mode>_nospec): New define_insn.
29255 (*sibcall_nonlocal_sysv<mode>): Generate different code for
29256 -mno-speculate-indirect-jumps.
29257 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
29259 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
29261 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
29262 long double type, set the flags for noting the default long double
29263 type, even if we don't pass or return a long double type.
29265 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
29268 * ipa-inline.c (flatten_function): Do not overwrite final inlining
29271 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
29273 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
29274 support for merge[hl].
29275 (fold_mergehl_helper): New helper function.
29276 (tree-vector-builder.h): New #include for tree_vector_builder usage.
29277 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
29278 (altivec_vmrglw_direct): Add xxmrglw insn.
29280 2018-01-17 Andrew Waterman <andrew@sifive.com>
29282 * config/riscv/riscv.c (riscv_conditional_register_usage): If
29283 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
29285 2018-01-17 David Malcolm <dmalcolm@redhat.com>
29288 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
29289 call the lto_location_cache before reading the
29290 DECL_SOURCE_LOCATION of the types.
29292 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
29293 Richard Sandiford <richard.sandiford@linaro.org>
29295 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
29296 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
29297 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
29298 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
29299 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
29301 * config/aarch64/constraints.md (aarch64_movti_operand):
29303 * config/aarch64/predicates.md (Uti): Add new constraint.
29305 2018-01-17 Carl Love <cel@us.ibm.com>
29307 * config/rs6000/vsx.md (define_expand xl_len_r,
29308 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
29309 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
29311 (define_expand, define_insn): Move the shift left from the
29312 define_insn to the define_expand for lxvl and stxvl instructions.
29313 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
29314 and XL_LEN_R definitions to PURE.
29316 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
29318 * config/i386/i386.c (indirect_thunk_name): Declare regno
29319 as unsigned int. Compare regno with INVALID_REGNUM.
29320 (output_indirect_thunk): Ditto.
29321 (output_indirect_thunk_function): Ditto.
29322 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
29323 in the call to output_indirect_thunk_function.
29325 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
29327 PR middle-end/83884
29328 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
29329 rather than the size of inner_type to determine the stack slot size
29330 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
29332 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
29335 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
29338 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
29340 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
29341 endian Linux systems to optionally enable multilibs for selecting
29342 the long double type if the user configured an explicit type.
29343 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
29344 have no long double multilibs if not defined.
29345 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
29346 warn if the user used -mabi={ieee,ibm}longdouble and we built
29347 multilibs for long double.
29348 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
29349 appropriate multilib option.
29350 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
29352 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
29353 for building long double multilibs.
29354 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
29356 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
29358 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
29361 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
29363 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
29366 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
29369 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
29372 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
29374 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
29375 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
29377 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
29379 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
29380 different rtl trees depending on TARGET_64BIT.
29381 (rs6000_gen_lvx): Likewise.
29383 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
29385 * config/visium/visium.md (nop): Tweak comment.
29386 (hazard_nop): Likewise.
29388 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
29390 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
29391 -mspeculate-indirect-jumps.
29392 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
29393 for -mno-speculate-indirect-jumps.
29394 (*call_indirect_elfv2<mode>_nospec): New define_insn.
29395 (*call_value_indirect_elfv2<mode>): Disable for
29396 -mno-speculate-indirect-jumps.
29397 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
29398 (indirect_jump): Emit different RTL for
29399 -mno-speculate-indirect-jumps.
29400 (*indirect_jump<mode>): Disable for
29401 -mno-speculate-indirect-jumps.
29402 (*indirect_jump<mode>_nospec): New define_insn.
29403 (tablejump): Emit different RTL for
29404 -mno-speculate-indirect-jumps.
29405 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
29406 (tablejumpsi_nospec): New define_expand.
29407 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
29408 (tablejumpdi_nospec): New define_expand.
29409 (*tablejump<mode>_internal1): Disable for
29410 -mno-speculate-indirect-jumps.
29411 (*tablejump<mode>_internal1_nospec): New define_insn.
29412 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
29415 2018-01-16 Artyom Skrobov tyomitch@gmail.com
29417 * caller-save.c (insert_save): Drop unnecessary parameter. All
29420 2018-01-16 Jakub Jelinek <jakub@redhat.com>
29421 Richard Biener <rguenth@suse.de>
29424 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
29425 return early, inline manually is_gimple_sizepos. Make sure if we
29426 call gimplify_expr we don't end up with a gimple constant.
29427 * tree.c (variably_modified_type_p): Don't return true for
29428 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
29429 * gimplify.h (is_gimple_sizepos): Remove.
29431 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
29433 PR tree-optimization/83857
29434 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
29435 vectorizable_live_operation for pure SLP statements.
29436 (vectorizable_live_operation): Handle PHIs.
29438 2018-01-16 Richard Biener <rguenther@suse.de>
29440 PR tree-optimization/83867
29441 * tree-vect-stmts.c (vect_transform_stmt): Precompute
29442 nested_in_vect_loop_p since the scalar stmt may get invalidated.
29444 2018-01-16 Jakub Jelinek <jakub@redhat.com>
29447 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
29448 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
29449 If off is not INTEGER_CST, issue a may not be aligned warning
29450 rather than isn't aligned. Use isn%'t rather than isn't.
29451 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
29453 <case MULT_EXPR>: Improve the case when bottom and one of the
29454 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
29455 operand, in that case check if the other operand is multiple of
29456 bottom divided by the INTEGER_CST operand.
29458 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
29461 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
29462 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
29463 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
29464 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
29465 * config/pa/pa.c (pa_function_arg_advance): Likewise.
29466 (pa_function_arg, pa_arg_partial_bytes): Likewise.
29467 (pa_function_arg_size): New function.
29469 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
29471 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
29472 in a separate statement.
29474 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
29476 PR tree-optimization/83847
29477 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
29478 group gathers and scatters.
29480 2018-01-16 Jakub Jelinek <jakub@redhat.com>
29482 PR rtl-optimization/86620
29483 * params.def (max-sched-ready-insns): Bump minimum value to 1.
29485 PR rtl-optimization/83213
29486 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
29487 to last if both are JUMP_INSNs.
29489 PR tree-optimization/83843
29490 * gimple-ssa-store-merging.c
29491 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
29492 store_immediate_info for bswap/nop orig_stores.
29494 2018-01-15 Andrew Waterman <andrew@sifive.com>
29496 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
29498 <UDIV>: Increase cost if !TARGET_DIV.
29500 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
29502 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
29503 (define_attr "cr_logical_3op"): New.
29504 (cceq_ior_compare): Adjust.
29505 (cceq_ior_compare_complement): Adjust.
29506 (*cceq_rev_compare): Adjust.
29507 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
29508 (is_cracked_insn): Adjust.
29509 (insn_must_be_first_in_group): Adjust.
29510 * config/rs6000/40x.md: Adjust.
29511 * config/rs6000/440.md: Adjust.
29512 * config/rs6000/476.md: Adjust.
29513 * config/rs6000/601.md: Adjust.
29514 * config/rs6000/603.md: Adjust.
29515 * config/rs6000/6xx.md: Adjust.
29516 * config/rs6000/7450.md: Adjust.
29517 * config/rs6000/7xx.md: Adjust.
29518 * config/rs6000/8540.md: Adjust.
29519 * config/rs6000/cell.md: Adjust.
29520 * config/rs6000/e300c2c3.md: Adjust.
29521 * config/rs6000/e500mc.md: Adjust.
29522 * config/rs6000/e500mc64.md: Adjust.
29523 * config/rs6000/e5500.md: Adjust.
29524 * config/rs6000/e6500.md: Adjust.
29525 * config/rs6000/mpc.md: Adjust.
29526 * config/rs6000/power4.md: Adjust.
29527 * config/rs6000/power5.md: Adjust.
29528 * config/rs6000/power6.md: Adjust.
29529 * config/rs6000/power7.md: Adjust.
29530 * config/rs6000/power8.md: Adjust.
29531 * config/rs6000/power9.md: Adjust.
29532 * config/rs6000/rs64.md: Adjust.
29533 * config/rs6000/titan.md: Adjust.
29535 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
29537 * config/i386/predicates.md (indirect_branch_operand): Rewrite
29538 ix86_indirect_branch_register logic.
29540 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
29542 * config/i386/constraints.md (Bs): Update
29543 ix86_indirect_branch_register check. Don't check
29544 ix86_indirect_branch_register with GOT_memory_operand.
29546 * config/i386/predicates.md (GOT_memory_operand): Don't check
29547 ix86_indirect_branch_register here.
29548 (GOT32_symbol_operand): Likewise.
29550 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
29552 * config/i386/predicates.md (constant_call_address_operand):
29553 Rewrite ix86_indirect_branch_register logic.
29554 (sibcall_insn_operand): Likewise.
29556 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
29558 * config/i386/constraints.md (Bs): Replace
29559 ix86_indirect_branch_thunk_register with
29560 ix86_indirect_branch_register.
29562 * config/i386/i386.md (indirect_jump): Likewise.
29563 (tablejump): Likewise.
29564 (*sibcall_memory): Likewise.
29565 (*sibcall_value_memory): Likewise.
29566 Peepholes of indirect call and jump via memory: Likewise.
29567 * config/i386/i386.opt: Likewise.
29568 * config/i386/predicates.md (indirect_branch_operand): Likewise.
29569 (GOT_memory_operand): Likewise.
29570 (call_insn_operand): Likewise.
29571 (sibcall_insn_operand): Likewise.
29572 (GOT32_symbol_operand): Likewise.
29574 2018-01-15 Jakub Jelinek <jakub@redhat.com>
29576 PR middle-end/83837
29577 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
29578 type rather than type addr's type points to.
29579 (expand_omp_atomic_mutex): Likewise.
29580 (expand_omp_atomic): Likewise.
29582 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
29585 * config/i386/i386.c (output_indirect_thunk_function): Use
29586 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
29587 for __x86_return_thunk.
29589 2018-01-15 Richard Biener <rguenther@suse.de>
29591 PR middle-end/83850
29592 * expmed.c (extract_bit_field_1): Fix typo.
29594 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29597 * config/arm/iterators.md (VF): New mode iterator.
29598 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
29599 Remove integer-related logic from pattern.
29600 (neon_vabd<mode>_3): Likewise.
29602 2018-01-15 Jakub Jelinek <jakub@redhat.com>
29604 PR middle-end/82694
29605 * common.opt (fstrict-overflow): No longer an alias.
29606 (fwrapv-pointer): New option.
29607 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
29608 also for pointer types based on flag_wrapv_pointer.
29609 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
29610 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
29611 opts->x_flag_wrapv got set.
29612 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
29613 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
29614 POINTER_TYPE_OVERFLOW_UNDEFINED.
29615 * match.pd: Likewise in address comparison pattern.
29616 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
29618 2018-01-15 Richard Biener <rguenther@suse.de>
29621 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
29622 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
29623 Reset type names to their identifier if their TYPE_DECL doesn't
29624 have linkage (and thus is used for ODR and devirt).
29625 (save_debug_info_for_decl): Remove.
29626 (save_debug_info_for_type): Likewise.
29627 (add_tree_to_fld_list): Adjust.
29628 * tree-pretty-print.c (dump_generic_node): Make dumping of
29629 type names more robust.
29631 2018-01-15 Richard Biener <rguenther@suse.de>
29633 * BASE-VER: Bump to 8.0.1.
29635 2018-01-14 Martin Sebor <msebor@redhat.com>
29638 * builtins.c (check_access): Avoid warning when the no-warning bit
29641 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
29643 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
29644 * ira-color (allocno_hard_regs_compare): Likewise.
29646 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
29649 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
29650 Use .pushsection/.popsection.
29652 2018-01-14 Martin Sebor <msebor@redhat.com>
29655 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
29657 2018-01-14 Jakub Jelinek <jakub@redhat.com>
29659 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
29660 entry from extra_headers.
29661 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
29662 extra_headers, make the list bitwise identical to the i?86-*-* one.
29664 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
29666 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
29667 -mcmodel=large with -mindirect-branch=thunk,
29668 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
29669 -mfunction-return=thunk-extern.
29670 * doc/invoke.texi: Document -mcmodel=large is incompatible with
29671 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
29672 -mfunction-return=thunk and -mfunction-return=thunk-extern.
29674 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
29676 * config/i386/i386.c (print_reg): Print the name of the full
29677 integer register without '%'.
29678 (ix86_print_operand): Handle 'V'.
29679 * doc/extend.texi: Document 'V' modifier.
29681 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
29683 * config/i386/constraints.md (Bs): Disallow memory operand for
29684 -mindirect-branch-register.
29686 * config/i386/predicates.md (indirect_branch_operand): Likewise.
29687 (GOT_memory_operand): Likewise.
29688 (call_insn_operand): Likewise.
29689 (sibcall_insn_operand): Likewise.
29690 (GOT32_symbol_operand): Likewise.
29691 * config/i386/i386.md (indirect_jump): Call convert_memory_address
29692 for -mindirect-branch-register.
29693 (tablejump): Likewise.
29694 (*sibcall_memory): Likewise.
29695 (*sibcall_value_memory): Likewise.
29696 Disallow peepholes of indirect call and jump via memory for
29697 -mindirect-branch-register.
29698 (*call_pop): Replace m with Bw.
29699 (*call_value_pop): Likewise.
29700 (*sibcall_pop_memory): Replace m with Bs.
29701 * config/i386/i386.opt (mindirect-branch-register): New option.
29702 * doc/invoke.texi: Document -mindirect-branch-register option.
29704 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
29706 * config/i386/i386-protos.h (ix86_output_function_return): New.
29707 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
29708 set function_return_type.
29709 (indirect_thunk_name): Add ret_p to indicate thunk for function
29711 (output_indirect_thunk_function): Pass false to
29712 indirect_thunk_name.
29713 (ix86_output_indirect_branch_via_reg): Likewise.
29714 (ix86_output_indirect_branch_via_push): Likewise.
29715 (output_indirect_thunk_function): Create alias for function
29716 return thunk if regno < 0.
29717 (ix86_output_function_return): New function.
29718 (ix86_handle_fndecl_attribute): Handle function_return.
29719 (ix86_attribute_table): Add function_return.
29720 * config/i386/i386.h (machine_function): Add
29721 function_return_type.
29722 * config/i386/i386.md (simple_return_internal): Use
29723 ix86_output_function_return.
29724 (simple_return_internal_long): Likewise.
29725 * config/i386/i386.opt (mfunction-return=): New option.
29726 (indirect_branch): Mention -mfunction-return=.
29727 * doc/extend.texi: Document function_return function attribute.
29728 * doc/invoke.texi: Document -mfunction-return= option.
29730 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
29732 * config/i386/i386-opts.h (indirect_branch): New.
29733 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
29734 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
29735 with local indirect jump when converting indirect call and jump.
29736 (ix86_set_indirect_branch_type): New.
29737 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
29738 (indirectlabelno): New.
29739 (indirect_thunk_needed): Likewise.
29740 (indirect_thunk_bnd_needed): Likewise.
29741 (indirect_thunks_used): Likewise.
29742 (indirect_thunks_bnd_used): Likewise.
29743 (INDIRECT_LABEL): Likewise.
29744 (indirect_thunk_name): Likewise.
29745 (output_indirect_thunk): Likewise.
29746 (output_indirect_thunk_function): Likewise.
29747 (ix86_output_indirect_branch_via_reg): Likewise.
29748 (ix86_output_indirect_branch_via_push): Likewise.
29749 (ix86_output_indirect_branch): Likewise.
29750 (ix86_output_indirect_jmp): Likewise.
29751 (ix86_code_end): Call output_indirect_thunk_function if needed.
29752 (ix86_output_call_insn): Call ix86_output_indirect_branch if
29754 (ix86_handle_fndecl_attribute): Handle indirect_branch.
29755 (ix86_attribute_table): Add indirect_branch.
29756 * config/i386/i386.h (machine_function): Add indirect_branch_type
29757 and has_local_indirect_jump.
29758 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
29760 (tablejump): Likewise.
29761 (*indirect_jump): Use ix86_output_indirect_jmp.
29762 (*tablejump_1): Likewise.
29763 (simple_return_indirect_internal): Likewise.
29764 * config/i386/i386.opt (mindirect-branch=): New option.
29765 (indirect_branch): New.
29768 (thunk-inline): Likewise.
29769 (thunk-extern): Likewise.
29770 * doc/extend.texi: Document indirect_branch function attribute.
29771 * doc/invoke.texi: Document -mindirect-branch= option.
29773 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
29776 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
29778 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
29780 * ipa-inline.c (want_inline_small_function_p): Return false if
29781 inlining has already failed with CIF_FINAL_ERROR.
29782 (update_caller_keys): Call want_inline_small_function_p before
29784 (update_callee_keys): Likewise.
29786 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
29788 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
29790 (rs6000_quadword_masked_address_p): Likewise.
29791 (quad_aligned_load_p): Likewise.
29792 (quad_aligned_store_p): Likewise.
29793 (const_load_sequence_p): Add comment to describe the outer-most loop.
29794 (mimic_memory_attributes_and_flags): New function.
29795 (rs6000_gen_stvx): Likewise.
29796 (replace_swapped_aligned_store): Likewise.
29797 (rs6000_gen_lvx): Likewise.
29798 (replace_swapped_aligned_load): Likewise.
29799 (replace_swapped_load_constant): Capitalize argument name in
29800 comment describing this function.
29801 (rs6000_analyze_swaps): Add a third pass to search for vector loads
29802 and stores that access quad-word aligned addresses and replace
29803 with stvx or lvx instructions when appropriate.
29804 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
29805 New function prototype.
29806 (rs6000_quadword_masked_address_p): Likewise.
29807 (rs6000_gen_lvx): Likewise.
29808 (rs6000_gen_stvx): Likewise.
29809 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
29810 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
29811 when memory address is aligned.
29812 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
29813 this split to select lvx instruction when memory address is aligned.
29814 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
29815 instruction when memory address is aligned.
29816 (*vsx_le_perm_load_v16qi): Likewise.
29817 (four unnamed splitters): Modify to select the stvx instruction
29818 when memory is aligned.
29820 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
29822 * predict.c (determine_unlikely_bbs): Handle correctly BBs
29823 which appears in the queue multiple times.
29825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29826 Alan Hayward <alan.hayward@arm.com>
29827 David Sherwood <david.sherwood@arm.com>
29829 * tree-vectorizer.h (vec_lower_bound): New structure.
29830 (_loop_vec_info): Add check_nonzero and lower_bounds.
29831 (LOOP_VINFO_CHECK_NONZERO): New macro.
29832 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
29833 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
29834 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
29835 fields. Make seg_len the distance travelled, not including the
29837 (dr_direction_indicator): Declare.
29838 (dr_zero_step_indicator): Likewise.
29839 (dr_known_forward_stride_p): Likewise.
29840 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
29842 (runtime_alias_check_p): Allow runtime alias checks with
29844 (operator ==): Compare access_size and align.
29845 (prune_runtime_alias_test_list): Rework for new distinction between
29846 the access_size and seg_len.
29847 (create_intersect_range_checks_index): Likewise. Cope with polynomial
29849 (get_segment_min_max): New function.
29850 (create_intersect_range_checks): Use it.
29851 (dr_step_indicator): New function.
29852 (dr_direction_indicator): Likewise.
29853 (dr_zero_step_indicator): Likewise.
29854 (dr_known_forward_stride_p): Likewise.
29855 * tree-loop-distribution.c (data_ref_segment_size): Return
29856 DR_STEP * (niters - 1).
29857 (compute_alias_check_pairs): Update call to the dr_with_seg_len
29859 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
29860 (vect_preserves_scalar_order_p): New function, split out from...
29861 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
29862 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
29863 (vect_vfa_access_size): New function.
29864 (vect_vfa_align): Likewise.
29865 (vect_compile_time_alias): Take access_size_a and access_b arguments.
29866 (dump_lower_bound): New function.
29867 (vect_check_lower_bound): Likewise.
29868 (vect_small_gap_p): Likewise.
29869 (vectorizable_with_step_bound_p): Likewise.
29870 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
29871 depencies if the vectorization factor is 1. Convert the checks
29872 for nonzero steps into checks on the bounds of DR_STEP. Try using
29873 a bunds check for variable steps if the minimum required step is
29874 relatively small. Update calls to the dr_with_seg_len
29875 constructor and to vect_compile_time_alias.
29876 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
29878 (vect_loop_versioning): Call it.
29879 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
29881 (vect_estimate_min_profitable_iters): Account for any bounds checks.
29883 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29884 Alan Hayward <alan.hayward@arm.com>
29885 David Sherwood <david.sherwood@arm.com>
29887 * doc/sourcebuild.texi (vect_scatter_store): Document.
29888 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
29890 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
29892 * genopinit.c (main): Add supports_vec_scatter_store and
29893 supports_vec_scatter_store_cached to target_optabs.
29894 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
29895 IFN_MASK_SCATTER_STORE.
29896 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
29898 * internal-fn.h (internal_store_fn_p): Declare.
29899 (internal_fn_stored_value_index): Likewise.
29900 * internal-fn.c (scatter_store_direct): New macro.
29901 (expand_scatter_store_optab_fn): New function.
29902 (direct_scatter_store_optab_supported_p): New macro.
29903 (internal_store_fn_p): New function.
29904 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
29905 IFN_MASK_SCATTER_STORE.
29906 (internal_fn_mask_index): Likewise.
29907 (internal_fn_stored_value_index): New function.
29908 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
29909 for scatter stores.
29910 * optabs-query.h (supports_vec_scatter_store_p): Declare.
29911 * optabs-query.c (supports_vec_scatter_store_p): New function.
29912 * tree-vectorizer.h (vect_get_store_rhs): Declare.
29913 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
29914 true for scatter stores.
29915 (vect_gather_scatter_fn_p): Handle scatter stores too.
29916 (vect_check_gather_scatter): Consider using scatter stores if
29917 supports_vec_scatter_store_p.
29918 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
29919 scatter stores too.
29920 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
29921 internal_fn_stored_value_index.
29922 (check_load_store_masking): Handle scatter stores too.
29923 (vect_get_store_rhs): Make public.
29924 (vectorizable_call): Use internal_store_fn_p.
29925 (vectorizable_store): Handle scatter store internal functions.
29926 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
29927 when deciding whether the end of the group has been reached.
29928 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
29929 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
29930 (mask_scatter_store<mode>): New insns.
29932 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29933 Alan Hayward <alan.hayward@arm.com>
29934 David Sherwood <david.sherwood@arm.com>
29936 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
29937 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
29938 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
29940 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
29941 Use vect_truncate_gather_scatter_offset if we can't treat the
29942 operation as a normal gather load or scatter store.
29943 (get_group_load_store_type): Take the gather_scatter_info
29944 as argument. Try using a gather load or scatter store for
29945 single-element groups.
29946 (get_load_store_type): Update calls to get_group_load_store_type
29947 and vect_use_strided_gather_scatters_p.
29949 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29950 Alan Hayward <alan.hayward@arm.com>
29951 David Sherwood <david.sherwood@arm.com>
29953 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
29954 optional tree argument.
29955 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
29957 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
29958 but continue to use the current value as a fallback.
29959 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
29960 to compare the updates.
29961 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
29962 (get_load_store_type): Use it when handling a strided access.
29963 (vect_get_strided_load_store_ops): New function.
29964 (vect_get_data_ptr_increment): Likewise.
29965 (vectorizable_load): Handle strided gather loads. Always pass
29966 a step to vect_create_data_ref_ptr and bump_vector_ptr.
29968 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29969 Alan Hayward <alan.hayward@arm.com>
29970 David Sherwood <david.sherwood@arm.com>
29972 * doc/md.texi (gather_load@var{m}): Document.
29973 (mask_gather_load@var{m}): Likewise.
29974 * genopinit.c (main): Add supports_vec_gather_load and
29975 supports_vec_gather_load_cached to target_optabs.
29976 * optabs-tree.c (init_tree_optimization_optabs): Use
29977 ggc_cleared_alloc to allocate target_optabs.
29978 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
29979 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
29981 * internal-fn.h (internal_load_fn_p): Declare.
29982 (internal_gather_scatter_fn_p): Likewise.
29983 (internal_fn_mask_index): Likewise.
29984 (internal_gather_scatter_fn_supported_p): Likewise.
29985 * internal-fn.c (gather_load_direct): New macro.
29986 (expand_gather_load_optab_fn): New function.
29987 (direct_gather_load_optab_supported_p): New macro.
29988 (direct_internal_fn_optab): New function.
29989 (internal_load_fn_p): Likewise.
29990 (internal_gather_scatter_fn_p): Likewise.
29991 (internal_fn_mask_index): Likewise.
29992 (internal_gather_scatter_fn_supported_p): Likewise.
29993 * optabs-query.c (supports_at_least_one_mode_p): New function.
29994 (supports_vec_gather_load_p): Likewise.
29995 * optabs-query.h (supports_vec_gather_load_p): Declare.
29996 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
29997 and memory_type field.
29998 (NUM_PATTERNS): Bump to 15.
29999 * tree-vect-data-refs.c: Include internal-fn.h.
30000 (vect_gather_scatter_fn_p): New function.
30001 (vect_describe_gather_scatter_call): Likewise.
30002 (vect_check_gather_scatter): Try using internal functions for
30003 gather loads. Recognize existing calls to a gather load function.
30004 (vect_analyze_data_refs): Consider using gather loads if
30005 supports_vec_gather_load_p.
30006 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
30007 (vect_get_gather_scatter_offset_type): Likewise.
30008 (vect_convert_mask_for_vectype): Likewise.
30009 (vect_add_conversion_to_patterm): Likewise.
30010 (vect_try_gather_scatter_pattern): Likewise.
30011 (vect_recog_gather_scatter_pattern): New pattern recognizer.
30012 (vect_vect_recog_func_ptrs): Add it.
30013 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
30014 internal_fn_mask_index and internal_gather_scatter_fn_p.
30015 (check_load_store_masking): Take the gather_scatter_info as an
30016 argument and handle gather loads.
30017 (vect_get_gather_scatter_ops): New function.
30018 (vectorizable_call): Check internal_load_fn_p.
30019 (vectorizable_load): Likewise. Handle gather load internal
30021 (vectorizable_store): Update call to check_load_store_masking.
30022 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
30023 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
30024 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
30025 (aarch64_gather_scale_operand_d): New predicates.
30026 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
30027 (mask_gather_load<mode>): New insns.
30029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30030 Alan Hayward <alan.hayward@arm.com>
30031 David Sherwood <david.sherwood@arm.com>
30033 * optabs.def (fold_left_plus_optab): New optab.
30034 * doc/md.texi (fold_left_plus_@var{m}): Document.
30035 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
30036 * internal-fn.c (fold_left_direct): Define.
30037 (expand_fold_left_optab_fn): Likewise.
30038 (direct_fold_left_optab_supported_p): Likewise.
30039 * fold-const-call.c (fold_const_fold_left): New function.
30040 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
30041 * tree-parloops.c (valid_reduction_p): New function.
30042 (gather_scalar_reductions): Use it.
30043 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
30044 (vect_finish_replace_stmt): Declare.
30045 * tree-vect-loop.c (fold_left_reduction_fn): New function.
30046 (needs_fold_left_reduction_p): New function, split out from...
30047 (vect_is_simple_reduction): ...here. Accept reductions that
30048 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
30049 (vect_force_simple_reduction): Also store the reduction type in
30050 the assignment's STMT_VINFO_REDUC_TYPE.
30051 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
30052 (merge_with_identity): New function.
30053 (vect_expand_fold_left): Likewise.
30054 (vectorize_fold_left_reduction): Likewise.
30055 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
30056 scalar phi in place for it. Check for target support and reject
30057 cases that would reassociate the operation. Defer the transform
30058 phase to vectorize_fold_left_reduction.
30059 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
30060 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
30061 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
30063 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30065 * tree-if-conv.c (predicate_mem_writes): Remove redundant
30066 call to ifc_temp_var.
30068 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30069 Alan Hayward <alan.hayward@arm.com>
30070 David Sherwood <david.sherwood@arm.com>
30072 * target.def (legitimize_address_displacement): Take the original
30073 offset as a poly_int.
30074 * targhooks.h (default_legitimize_address_displacement): Update
30076 * targhooks.c (default_legitimize_address_displacement): Likewise.
30077 * doc/tm.texi: Regenerate.
30078 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
30079 as an argument, moving assert of ad->disp == ad->disp_term to...
30080 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
30081 Try calling targetm.legitimize_address_displacement before expanding
30082 the address rather than afterwards, and adjust for the new interface.
30083 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
30084 Match the new hook interface. Handle SVE addresses.
30085 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
30086 new hook interface.
30088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30090 * Makefile.in (OBJS): Add early-remat.o.
30091 * target.def (select_early_remat_modes): New hook.
30092 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
30093 * doc/tm.texi: Regenerate.
30094 * targhooks.h (default_select_early_remat_modes): Declare.
30095 * targhooks.c (default_select_early_remat_modes): New function.
30096 * timevar.def (TV_EARLY_REMAT): New timevar.
30097 * passes.def (pass_early_remat): New pass.
30098 * tree-pass.h (make_pass_early_remat): Declare.
30099 * early-remat.c: New file.
30100 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
30102 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
30104 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30105 Alan Hayward <alan.hayward@arm.com>
30106 David Sherwood <david.sherwood@arm.com>
30108 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
30109 vfm1 with a bound_epilog parameter.
30110 (vect_do_peeling): Update calls accordingly, and move the prologue
30111 call earlier in the function. Treat the base bound_epilog as 0 for
30112 fully-masked loops and retain vf - 1 for other loops. Add 1 to
30113 this base when peeling for gaps.
30114 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
30115 with fully-masked loops.
30116 (vect_estimate_min_profitable_iters): Handle the single peeled
30117 iteration in that case.
30119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30120 Alan Hayward <alan.hayward@arm.com>
30121 David Sherwood <david.sherwood@arm.com>
30123 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
30124 single-element interleaving even if the size is not a power of 2.
30125 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
30126 accesses for single-element interleaving if the group size is
30129 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30130 Alan Hayward <alan.hayward@arm.com>
30131 David Sherwood <david.sherwood@arm.com>
30133 * doc/md.texi (fold_extract_last_@var{m}): Document.
30134 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
30135 * optabs.def (fold_extract_last_optab): New optab.
30136 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
30137 * internal-fn.c (fold_extract_direct): New macro.
30138 (expand_fold_extract_optab_fn): Likewise.
30139 (direct_fold_extract_optab_supported_p): Likewise.
30140 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
30141 * tree-vect-loop.c (vect_model_reduction_cost): Handle
30142 EXTRACT_LAST_REDUCTION.
30143 (get_initial_def_for_reduction): Do not create an initial vector
30144 for EXTRACT_LAST_REDUCTION reductions.
30145 (vectorizable_reduction): Leave the scalar phi in place for
30146 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
30147 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
30148 epilogue code for EXTRACT_LAST_REDUCTION and defer the
30149 transform phase to vectorizable_condition.
30150 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
30152 (vect_finish_stmt_generation): ...here.
30153 (vect_finish_replace_stmt): New function.
30154 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
30155 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
30157 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
30159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30160 Alan Hayward <alan.hayward@arm.com>
30161 David Sherwood <david.sherwood@arm.com>
30163 * doc/md.texi (extract_last_@var{m}): Document.
30164 * optabs.def (extract_last_optab): New optab.
30165 * internal-fn.def (EXTRACT_LAST): New internal function.
30166 * internal-fn.c (cond_unary_direct): New macro.
30167 (expand_cond_unary_optab_fn): Likewise.
30168 (direct_cond_unary_optab_supported_p): Likewise.
30169 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
30170 loops using EXTRACT_LAST.
30171 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
30172 (extract_last_<mode>): ...this optab.
30173 (vec_extract<mode><Vel>): Update accordingly.
30175 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30176 Alan Hayward <alan.hayward@arm.com>
30177 David Sherwood <david.sherwood@arm.com>
30179 * target.def (empty_mask_is_expensive): New hook.
30180 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
30181 * doc/tm.texi: Regenerate.
30182 * targhooks.h (default_empty_mask_is_expensive): Declare.
30183 * targhooks.c (default_empty_mask_is_expensive): New function.
30184 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
30185 if the target says that empty masks are expensive.
30186 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
30188 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
30190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30191 Alan Hayward <alan.hayward@arm.com>
30192 David Sherwood <david.sherwood@arm.com>
30194 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
30195 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
30196 (vect_use_loop_mask_for_alignment_p): New function.
30197 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
30198 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
30199 niters_skip argument. Make sure that the first niters_skip elements
30200 of the first iteration are inactive.
30201 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
30202 Update call to vect_set_loop_masks_directly.
30203 (get_misalign_in_elems): New function, split out from...
30204 (vect_gen_prolog_loop_niters): ...here.
30205 (vect_update_init_of_dr): Take a code argument that specifies whether
30206 the adjustment should be added or subtracted.
30207 (vect_update_init_of_drs): Likewise.
30208 (vect_prepare_for_masked_peels): New function.
30209 (vect_do_peeling): Skip prologue peeling if we're using a mask
30210 instead. Update call to vect_update_inits_of_drs.
30211 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
30213 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
30214 alignment. Do not include the number of peeled iterations in
30215 the minimum threshold in that case.
30216 (vectorizable_induction): Adjust the start value down by
30217 LOOP_VINFO_MASK_SKIP_NITERS iterations.
30218 (vect_transform_loop): Call vect_prepare_for_masked_peels.
30219 Take the number of skipped iterations into account when calculating
30221 * tree-vect-stmts.c (vect_gen_while_not): New function.
30223 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30224 Alan Hayward <alan.hayward@arm.com>
30225 David Sherwood <david.sherwood@arm.com>
30227 * doc/sourcebuild.texi (vect_fully_masked): Document.
30228 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
30229 default value to 0.
30230 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
30232 (vect_analyze_loop_2): ...here. Don't check the vectorization
30233 factor against the number of loop iterations if the loop is
30236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30237 Alan Hayward <alan.hayward@arm.com>
30238 David Sherwood <david.sherwood@arm.com>
30240 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
30241 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
30242 (dump_groups): Update accordingly.
30243 (iv_use::mem_type): New member variable.
30244 (address_p): New function.
30245 (record_use): Add a mem_type argument and initialize the new
30247 (record_group_use): Add a mem_type argument. Use address_p.
30248 Remove obsolete null checks of base_object. Update call to record_use.
30249 (find_interesting_uses_op): Update call to record_group_use.
30250 (find_interesting_uses_cond): Likewise.
30251 (find_interesting_uses_address): Likewise.
30252 (get_mem_type_for_internal_fn): New function.
30253 (find_address_like_use): Likewise.
30254 (find_interesting_uses_stmt): Try find_address_like_use before
30255 calling find_interesting_uses_op.
30256 (addr_offset_valid_p): Use the iv mem_type field as the type
30257 of the addressed memory.
30258 (add_autoinc_candidates): Likewise.
30259 (get_address_cost): Likewise.
30260 (split_small_address_groups_p): Use address_p.
30261 (split_address_groups): Likewise.
30262 (add_iv_candidate_for_use): Likewise.
30263 (autoinc_possible_for_pair): Likewise.
30264 (rewrite_groups): Likewise.
30265 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
30266 (determine_group_iv_cost): Update after split of USE_ADDRESS.
30267 (get_alias_ptr_type_for_ptr_address): New function.
30268 (rewrite_use_address): Rewrite address uses in calls that were
30269 identified by find_address_like_use.
30271 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30272 Alan Hayward <alan.hayward@arm.com>
30273 David Sherwood <david.sherwood@arm.com>
30275 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
30277 * gimple-expr.h (is_gimple_addressable: Likewise.
30278 * gimple-expr.c (is_gimple_address): Likewise.
30279 * internal-fn.c (expand_call_mem_ref): New function.
30280 (expand_mask_load_optab_fn): Use it.
30281 (expand_mask_store_optab_fn): Likewise.
30283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30284 Alan Hayward <alan.hayward@arm.com>
30285 David Sherwood <david.sherwood@arm.com>
30287 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
30288 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
30289 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
30290 (cond_umax@var{mode}): Document.
30291 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
30292 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
30293 (cond_umin_optab, cond_umax_optab): New optabs.
30294 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
30295 (COND_IOR, COND_XOR): New internal functions.
30296 * internal-fn.h (get_conditional_internal_fn): Declare.
30297 * internal-fn.c (cond_binary_direct): New macro.
30298 (expand_cond_binary_optab_fn): Likewise.
30299 (direct_cond_binary_optab_supported_p): Likewise.
30300 (get_conditional_internal_fn): New function.
30301 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
30302 Cope with reduction statements that are vectorized as calls rather
30304 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
30305 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
30306 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
30307 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
30308 (UNSPEC_COND_EOR): New unspecs.
30309 (optab): Add mappings for them.
30310 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
30311 (sve_int_op, sve_fp_op): New int attributes.
30313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30314 Alan Hayward <alan.hayward@arm.com>
30315 David Sherwood <david.sherwood@arm.com>
30317 * optabs.def (while_ult_optab): New optab.
30318 * doc/md.texi (while_ult@var{m}@var{n}): Document.
30319 * internal-fn.def (WHILE_ULT): New internal function.
30320 * internal-fn.h (direct_internal_fn_supported_p): New override
30321 that takes two types as argument.
30322 * internal-fn.c (while_direct): New macro.
30323 (expand_while_optab_fn): New function.
30324 (convert_optab_supported_p): Likewise.
30325 (direct_while_optab_supported_p): New macro.
30326 * wide-int.h (wi::udiv_ceil): New function.
30327 * tree-vectorizer.h (rgroup_masks): New structure.
30328 (vec_loop_masks): New typedef.
30329 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
30330 and fully_masked_p.
30331 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
30332 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
30333 (vect_max_vf): New function.
30334 (slpeel_make_loop_iterate_ntimes): Delete.
30335 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
30336 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
30337 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
30338 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
30339 internal-fn.h, stor-layout.h and optabs-query.h.
30340 (vect_set_loop_mask): New function.
30341 (add_preheader_seq): Likewise.
30342 (add_header_seq): Likewise.
30343 (interleave_supported_p): Likewise.
30344 (vect_maybe_permute_loop_masks): Likewise.
30345 (vect_set_loop_masks_directly): Likewise.
30346 (vect_set_loop_condition_masked): Likewise.
30347 (vect_set_loop_condition_unmasked): New function, split out from
30348 slpeel_make_loop_iterate_ntimes.
30349 (slpeel_make_loop_iterate_ntimes): Rename to..
30350 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
30351 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
30352 (vect_do_peeling): Update call accordingly.
30353 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
30355 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
30356 mask_compare_type, can_fully_mask_p and fully_masked_p.
30357 (release_vec_loop_masks): New function.
30358 (_loop_vec_info): Use it to free the loop masks.
30359 (can_produce_all_loop_masks_p): New function.
30360 (vect_get_max_nscalars_per_iter): Likewise.
30361 (vect_verify_full_masking): Likewise.
30362 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
30363 retries, and free the mask rgroups before retrying. Check loop-wide
30364 reasons for disallowing fully-masked loops. Make the final decision
30365 about whether use a fully-masked loop or not.
30366 (vect_estimate_min_profitable_iters): Do not assume that peeling
30367 for the number of iterations will be needed for fully-masked loops.
30368 (vectorizable_reduction): Disable fully-masked loops.
30369 (vectorizable_live_operation): Likewise.
30370 (vect_halve_mask_nunits): New function.
30371 (vect_double_mask_nunits): Likewise.
30372 (vect_record_loop_mask): Likewise.
30373 (vect_get_loop_mask): Likewise.
30374 (vect_transform_loop): Handle the case in which the final loop
30375 iteration might handle a partial vector. Call vect_set_loop_condition
30376 instead of slpeel_make_loop_iterate_ntimes.
30377 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
30378 (check_load_store_masking): New function.
30379 (prepare_load_store_mask): Likewise.
30380 (vectorizable_store): Handle fully-masked loops.
30381 (vectorizable_load): Likewise.
30382 (supportable_widening_operation): Use vect_halve_mask_nunits for
30384 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
30385 (vect_gen_while): New function.
30386 * config/aarch64/aarch64.md (umax<mode>3): New expander.
30387 (aarch64_uqdec<mode>): New insn.
30389 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30390 Alan Hayward <alan.hayward@arm.com>
30391 David Sherwood <david.sherwood@arm.com>
30393 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
30394 (reduc_xor_scal_optab): New optabs.
30395 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
30396 (reduc_xor_scal_@var{m}): Document.
30397 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
30398 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
30399 internal functions.
30400 * fold-const-call.c (fold_const_call): Handle them.
30401 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
30402 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
30403 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
30404 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
30405 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
30406 (UNSPEC_XORV): New unspecs.
30407 (optab): Add entries for them.
30408 (BITWISEV): New int iterator.
30409 (bit_reduc_op): New int attributes.
30411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30412 Alan Hayward <alan.hayward@arm.com>
30413 David Sherwood <david.sherwood@arm.com>
30415 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
30416 * internal-fn.def (VEC_SHL_INSERT): New internal function.
30417 * optabs.def (vec_shl_insert_optab): New optab.
30418 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
30419 (duplicate_and_interleave): Likewise.
30420 * tree-vect-loop.c: Include internal-fn.h.
30421 (neutral_op_for_slp_reduction): New function, split out from
30422 get_initial_defs_for_reduction.
30423 (get_initial_def_for_reduction): Handle option 2 for variable-length
30424 vectors by loading the neutral value into a vector and then shifting
30425 the initial value into element 0.
30426 (get_initial_defs_for_reduction): Replace the code argument with
30427 the neutral value calculated by neutral_op_for_slp_reduction.
30428 Use gimple_build_vector for constant-length vectors.
30429 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
30430 but the first group_size elements have a neutral value.
30431 Use duplicate_and_interleave otherwise.
30432 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
30433 Update call to get_initial_defs_for_reduction. Handle SLP
30434 reductions for variable-length vectors by creating one vector
30435 result for each scalar result, with the elements associated
30436 with other scalar results stubbed out with the neutral value.
30437 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
30438 Require IFN_VEC_SHL_INSERT for double reductions on
30439 variable-length vectors, or SLP reductions that have
30440 a neutral value. Require can_duplicate_and_interleave_p
30441 support for variable-length unchained SLP reductions if there
30442 is no neutral value, such as for MIN/MAX reductions. Also require
30443 the number of vector elements to be a multiple of the number of
30444 SLP statements when doing variable-length unchained SLP reductions.
30445 Update call to vect_create_epilog_for_reduction.
30446 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
30447 and remove initial values.
30448 (duplicate_and_interleave): Make public.
30449 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
30450 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
30452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30453 Alan Hayward <alan.hayward@arm.com>
30454 David Sherwood <david.sherwood@arm.com>
30456 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
30457 (can_duplicate_and_interleave_p): New function.
30458 (vect_get_and_check_slp_defs): Take the vector of statements
30459 rather than just the current one. Remove excess parentheses.
30460 Restriction rejectinon of vect_constant_def and vect_external_def
30461 for variable-length vectors to boolean types, or types for which
30462 can_duplicate_and_interleave_p is false.
30463 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
30464 (duplicate_and_interleave): New function.
30465 (vect_get_constant_vectors): Use gimple_build_vector for
30466 constant-length vectors and suitable variable-length constant
30467 vectors. Use duplicate_and_interleave for other variable-length
30468 vectors. Don't defer the update when inserting new statements.
30470 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30471 Alan Hayward <alan.hayward@arm.com>
30472 David Sherwood <david.sherwood@arm.com>
30474 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
30475 min_profitable_iters doesn't go negative.
30477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30478 Alan Hayward <alan.hayward@arm.com>
30479 David Sherwood <david.sherwood@arm.com>
30481 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
30482 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
30483 * optabs.def (vec_mask_load_lanes_optab): New optab.
30484 (vec_mask_store_lanes_optab): Likewise.
30485 * internal-fn.def (MASK_LOAD_LANES): New internal function.
30486 (MASK_STORE_LANES): Likewise.
30487 * internal-fn.c (mask_load_lanes_direct): New macro.
30488 (mask_store_lanes_direct): Likewise.
30489 (expand_mask_load_optab_fn): Handle masked operations.
30490 (expand_mask_load_lanes_optab_fn): New macro.
30491 (expand_mask_store_optab_fn): Handle masked operations.
30492 (expand_mask_store_lanes_optab_fn): New macro.
30493 (direct_mask_load_lanes_optab_supported_p): Likewise.
30494 (direct_mask_store_lanes_optab_supported_p): Likewise.
30495 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
30497 (vect_load_lanes_supported): Likewise.
30498 * tree-vect-data-refs.c (strip_conversion): New function.
30499 (can_group_stmts_p): Likewise.
30500 (vect_analyze_data_ref_accesses): Use it instead of checking
30501 for a pair of assignments.
30502 (vect_store_lanes_supported): Take a masked_p parameter.
30503 (vect_load_lanes_supported): Likewise.
30504 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
30505 vect_store_lanes_supported and vect_load_lanes_supported.
30506 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
30507 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
30508 parameter. Don't allow gaps for masked accesses.
30509 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
30510 and vect_load_lanes_supported.
30511 (get_load_store_type): Take a masked_p parameter and update
30512 call to get_group_load_store_type.
30513 (vectorizable_store): Update call to get_load_store_type.
30514 Handle IFN_MASK_STORE_LANES.
30515 (vectorizable_load): Update call to get_load_store_type.
30516 Handle IFN_MASK_LOAD_LANES.
30518 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30519 Alan Hayward <alan.hayward@arm.com>
30520 David Sherwood <david.sherwood@arm.com>
30522 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
30524 * config/aarch64/aarch64-protos.h
30525 (aarch64_sve_struct_memory_operand_p): Declare.
30526 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
30527 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
30528 (VPRED, vpred): Handle SVE structure modes.
30529 * config/aarch64/constraints.md (Utx): New constraint.
30530 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
30531 (aarch64_sve_struct_nonimmediate_operand): New predicates.
30532 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
30533 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
30534 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
30535 structure modes. Split into pieces after RA.
30536 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
30537 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
30539 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
30540 SVE structure modes.
30541 (aarch64_classify_address): Likewise.
30542 (sizetochar): Move earlier in file.
30543 (aarch64_print_operand): Handle SVE register lists.
30544 (aarch64_array_mode): New function.
30545 (aarch64_sve_struct_memory_operand_p): Likewise.
30546 (TARGET_ARRAY_MODE): Redefine.
30548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30549 Alan Hayward <alan.hayward@arm.com>
30550 David Sherwood <david.sherwood@arm.com>
30552 * target.def (array_mode): New target hook.
30553 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
30554 * doc/tm.texi: Regenerate.
30555 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
30556 * hooks.c (hook_optmode_mode_uhwi_none): New function.
30557 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
30558 targetm.array_mode.
30559 * stor-layout.c (mode_for_array): Likewise. Support polynomial
30562 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30563 Alan Hayward <alan.hayward@arm.com>
30564 David Sherwood <david.sherwood@arm.com>
30566 * fold-const.c (fold_binary_loc): Check the argument types
30567 rather than the result type when testing for a vector operation.
30569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30571 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
30572 * doc/tm.texi: Regenerate.
30574 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30575 Alan Hayward <alan.hayward@arm.com>
30576 David Sherwood <david.sherwood@arm.com>
30578 * doc/invoke.texi (-msve-vector-bits=): Document new option.
30579 (sve): Document new AArch64 extension.
30580 * doc/md.texi (w): Extend the description of the AArch64
30581 constraint to include SVE vectors.
30582 (Upl, Upa): Document new AArch64 predicate constraints.
30583 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
30585 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
30586 (msve-vector-bits=): New option.
30587 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
30588 SVE when these are disabled.
30589 (sve): New extension.
30590 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
30591 modes. Adjust their number of units based on aarch64_sve_vg.
30592 (MAX_BITSIZE_MODE_ANY_MODE): Define.
30593 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
30594 aarch64_addr_query_type.
30595 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
30596 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
30597 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
30598 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
30599 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
30600 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
30601 (aarch64_simd_imm_zero_p): Delete.
30602 (aarch64_check_zero_based_sve_index_immediate): Declare.
30603 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
30604 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
30605 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
30606 (aarch64_sve_float_mul_immediate_p): Likewise.
30607 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
30608 rather than an rtx.
30609 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
30610 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
30611 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
30612 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
30613 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
30614 (aarch64_regmode_natural_size): Likewise.
30615 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
30616 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
30618 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
30619 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
30620 for VG and the SVE predicate registers.
30621 (V_ALIASES): Add a "z"-prefixed alias.
30622 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
30623 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
30624 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
30625 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
30626 (REG_CLASS_NAMES): Add entries for them.
30627 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
30628 and the predicate registers.
30629 (aarch64_sve_vg): Declare.
30630 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
30631 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
30632 (REGMODE_NATURAL_SIZE): Define.
30633 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
30635 * config/aarch64/aarch64.c: Include cfgrtl.h.
30636 (simd_immediate_info): Add a constructor for series vectors,
30637 and an associated step field.
30638 (aarch64_sve_vg): New variable.
30639 (aarch64_dbx_register_number): Handle VG and the predicate registers.
30640 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
30641 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
30642 (VEC_ANY_DATA, VEC_STRUCT): New constants.
30643 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
30644 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
30645 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
30646 (aarch64_get_mask_mode): New functions.
30647 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
30648 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
30649 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
30650 predicate modes and predicate registers. Explicitly restrict
30651 GPRs to modes of 16 bytes or smaller. Only allow FP registers
30652 to store a vector mode if it is recognized by
30653 aarch64_classify_vector_mode.
30654 (aarch64_regmode_natural_size): New function.
30655 (aarch64_hard_regno_caller_save_mode): Return the original mode
30657 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
30658 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
30659 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
30660 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
30662 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
30663 does not overlap dest if the function is frame-related. Handle
30665 (aarch64_split_add_offset): New function.
30666 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
30667 them aarch64_add_offset.
30668 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
30669 and update call to aarch64_sub_sp.
30670 (aarch64_add_cfa_expression): New function.
30671 (aarch64_expand_prologue): Pass extra temporary registers to the
30672 functions above. Handle the case in which we need to emit new
30673 DW_CFA_expressions for registers that were originally saved
30674 relative to the stack pointer, but now have to be expressed
30675 relative to the frame pointer.
30676 (aarch64_output_mi_thunk): Pass extra temporary registers to the
30678 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
30679 IP0 and IP1 values for SVE frames.
30680 (aarch64_expand_vec_series): New function.
30681 (aarch64_expand_sve_widened_duplicate): Likewise.
30682 (aarch64_expand_sve_const_vector): Likewise.
30683 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
30684 Handle SVE constants. Use emit_move_insn to move a force_const_mem
30685 into the register, rather than emitting a SET directly.
30686 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
30687 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
30688 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
30689 (offset_9bit_signed_scaled_p): New functions.
30690 (aarch64_replicate_bitmask_imm): New function.
30691 (aarch64_bitmask_imm): Use it.
30692 (aarch64_cannot_force_const_mem): Reject expressions involving
30693 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
30694 (aarch64_classify_index): Handle SVE indices, by requiring
30695 a plain register index with a scale that matches the element size.
30696 (aarch64_classify_address): Handle SVE addresses. Assert that
30697 the mode of the address is VOIDmode or an integer mode.
30698 Update call to aarch64_classify_symbol.
30699 (aarch64_classify_symbolic_expression): Update call to
30700 aarch64_classify_symbol.
30701 (aarch64_const_vec_all_in_range_p): New function.
30702 (aarch64_print_vector_float_operand): Likewise.
30703 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
30704 "vN" for FP registers with SVE modes. Handle (const ...) vectors
30705 and the FP immediates 1.0 and 0.5.
30706 (aarch64_print_address_internal): Handle SVE addresses.
30707 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
30708 (aarch64_regno_regclass): Handle predicate registers.
30709 (aarch64_secondary_reload): Handle big-endian reloads of SVE
30711 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
30712 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
30713 (aarch64_convert_sve_vector_bits): New function.
30714 (aarch64_override_options): Use it to handle -msve-vector-bits=.
30715 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
30716 rather than an rtx.
30717 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
30718 Handle SVE vector and predicate modes. Accept VL-based constants
30719 that need only one temporary register, and VL offsets that require
30720 no temporary registers.
30721 (aarch64_conditional_register_usage): Mark the predicate registers
30722 as fixed if SVE isn't available.
30723 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
30724 Return true for SVE vector and predicate modes.
30725 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
30726 rather than an unsigned int. Handle SVE modes.
30727 (aarch64_preferred_simd_mode): Update call accordingly. Handle
30729 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
30731 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
30732 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
30733 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
30734 (aarch64_sve_float_mul_immediate_p): New functions.
30735 (aarch64_sve_valid_immediate): New function.
30736 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
30737 Explicitly reject structure modes. Check for INDEX constants.
30738 Handle PTRUE and PFALSE constants.
30739 (aarch64_check_zero_based_sve_index_immediate): New function.
30740 (aarch64_simd_imm_zero_p): Delete.
30741 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
30742 vector modes. Accept constants in the range of CNT[BHWD].
30743 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
30744 ask for an Advanced SIMD mode.
30745 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
30746 (aarch64_simd_vector_alignment): Handle SVE predicates.
30747 (aarch64_vectorize_preferred_vector_alignment): New function.
30748 (aarch64_simd_vector_alignment_reachable): Use it instead of
30750 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
30751 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
30753 (MAX_VECT_LEN): Delete.
30754 (expand_vec_perm_d): Add a vec_flags field.
30755 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
30756 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
30757 (aarch64_evpc_ext): Don't apply a big-endian lane correction
30759 (aarch64_evpc_rev): Rename to...
30760 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
30761 (aarch64_evpc_rev_global): New function.
30762 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
30763 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
30765 (aarch64_evpc_sve_tbl): New function.
30766 (aarch64_expand_vec_perm_const_1): Update after rename of
30767 aarch64_evpc_rev. Handle SVE permutes too, trying
30768 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
30769 than aarch64_evpc_tbl.
30770 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
30771 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
30772 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
30773 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
30774 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
30775 (aarch64_expand_sve_vcond): New functions.
30776 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
30777 of aarch64_vector_mode_p.
30778 (aarch64_dwarf_poly_indeterminate_value): New function.
30779 (aarch64_compute_pressure_classes): Likewise.
30780 (aarch64_can_change_mode_class): Likewise.
30781 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
30782 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
30783 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
30784 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
30785 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
30786 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
30787 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
30788 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
30790 (Dn, Dl, Dr): Accept const as well as const_vector.
30791 (Dz): Likewise. Compare against CONST0_RTX.
30792 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
30793 of "vector" where appropriate.
30794 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
30795 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
30796 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
30797 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
30798 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
30799 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
30800 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
30801 (v_int_equiv): Extend to SVE modes.
30802 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
30804 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
30805 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
30806 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
30807 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
30808 (SVE_COND_FP_CMP): New int iterators.
30809 (perm_hilo): Handle the new unpack unspecs.
30810 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
30812 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
30813 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
30814 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
30815 (aarch64_equality_operator, aarch64_constant_vector_operand)
30816 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
30817 (aarch64_sve_nonimmediate_operand): Likewise.
30818 (aarch64_sve_general_operand): Likewise.
30819 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
30820 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
30821 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
30822 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
30823 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
30824 (aarch64_sve_float_arith_immediate): Likewise.
30825 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
30826 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
30827 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
30828 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
30829 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
30830 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
30831 (aarch64_sve_float_arith_operand): Likewise.
30832 (aarch64_sve_float_arith_with_sub_operand): Likewise.
30833 (aarch64_sve_float_mul_operand): Likewise.
30834 (aarch64_sve_vec_perm_operand): Likewise.
30835 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
30836 (aarch64_mov_operand): Accept const_poly_int and const_vector.
30837 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
30838 as well as const_vector.
30839 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
30840 in file. Use CONST0_RTX and CONSTM1_RTX.
30841 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
30842 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
30843 Use aarch64_simd_imm_zero.
30844 * config/aarch64/aarch64-sve.md: New file.
30845 * config/aarch64/aarch64.md: Include it.
30846 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
30847 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
30848 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
30849 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
30850 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
30851 (sve): New attribute.
30852 (enabled): Disable instructions with the sve attribute unless
30854 (movqi, movhi): Pass CONST_POLY_INT operaneds through
30855 aarch64_expand_mov_immediate.
30856 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
30857 CNT[BHSD] immediates.
30858 (movti): Split CONST_POLY_INT moves into two halves.
30859 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
30860 Split additions that need a temporary here if the destination
30861 is the stack pointer.
30862 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
30863 (*add<mode>3_poly_1): New instruction.
30864 (set_clobber_cc): New expander.
30866 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30868 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
30869 parameter and use it instead of GET_MODE_SIZE (innermode). Use
30870 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
30871 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
30872 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
30873 Change innermode from fixed_mode_size to machine_mode.
30874 (simplify_subreg): Update call accordingly. Handle a constant-sized
30875 subreg of a variable-length CONST_VECTOR.
30877 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
30878 Alan Hayward <alan.hayward@arm.com>
30879 David Sherwood <david.sherwood@arm.com>
30881 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
30882 (add_offset_to_base): New function, split out from...
30883 (create_mem_ref): ...here. When handling a scale other than 1,
30884 check first whether the address is valid without the offset.
30885 Add it into the base if so, leaving the index and scale as-is.
30887 2018-01-12 Jakub Jelinek <jakub@redhat.com>
30890 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
30891 fold_for_warn before checking if arg2 is INTEGER_CST.
30893 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
30895 * config/rs6000/predicates.md (load_multiple_operation): Delete.
30896 (store_multiple_operation): Delete.
30897 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
30898 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
30899 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
30900 guarded by TARGET_STRING.
30901 (rs6000_output_load_multiple): Delete.
30902 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30903 OPTION_MASK_STRING / TARGET_STRING handling.
30904 (print_operand) <'N', 'O'>: Add comment that these are unused now.
30905 (const rs6000_opt_masks) <"string">: Change mask to 0.
30906 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
30907 (MASK_STRING): Delete.
30908 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
30910 (load_multiple): Delete.
30917 (store_multiple): Delete.
30924 (movmemsi_8reg): Delete.
30925 (corresponding unnamed define_insn): Delete.
30926 (movmemsi_6reg): Delete.
30927 (corresponding unnamed define_insn): Delete.
30928 (movmemsi_4reg): Delete.
30929 (corresponding unnamed define_insn): Delete.
30930 (movmemsi_2reg): Delete.
30931 (corresponding unnamed define_insn): Delete.
30932 (movmemsi_1reg): Delete.
30933 (corresponding unnamed define_insn): Delete.
30934 * config/rs6000/rs6000.opt (mno-string): New.
30935 (mstring): Replace by deprecation warning stub.
30936 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
30938 2018-01-12 Jakub Jelinek <jakub@redhat.com>
30940 * regrename.c (regrename_do_replace): If replacing the same
30941 reg multiple times, try to reuse last created gen_raw_REG.
30944 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
30945 main to workaround a bug in GDB.
30947 2018-01-12 Tom de Vries <tom@codesourcery.com>
30950 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
30952 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
30954 PR rtl-optimization/80481
30955 * ira-color.c (get_cap_member): New function.
30956 (allocnos_conflict_by_live_ranges_p): Use it.
30957 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
30958 (setup_slot_coalesced_allocno_live_ranges): Ditto.
30960 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
30963 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
30964 (*saddl_se_1): Ditto.
30965 (*ssubsi_1): Ditto.
30966 (*ssubl_se_1): Ditto.
30968 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
30970 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
30971 rather than wi::to_widest for DR_INITs.
30972 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
30973 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
30974 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
30976 (vect_analyze_group_access_1): Note that here.
30978 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
30980 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
30981 polynomial type sizes.
30983 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
30985 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
30986 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
30987 (gimple_add_tmp_var): Likewise.
30989 2018-01-12 Martin Liska <mliska@suse.cz>
30991 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
30992 (gimple_alloc_sizes): Likewise.
30993 (dump_gimple_statistics): Use PRIu64 in printf format.
30994 * gimple.h: Change uint64_t to int.
30996 2018-01-12 Martin Liska <mliska@suse.cz>
30998 * tree-core.h: Use uint64_t instead of int.
30999 * tree.c (tree_node_counts): Likewise.
31000 (tree_node_sizes): Likewise.
31001 (dump_tree_statistics): Use PRIu64 in printf format.
31003 2018-01-12 Martin Liska <mliska@suse.cz>
31005 * Makefile.in: As qsort_chk is implemented in vec.c, add
31006 vec.o to linkage of gencfn-macros.
31007 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
31008 passing the info to record_node_allocation_statistics.
31009 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
31011 * ggc-common.c (struct ggc_usage): Add operator== and use
31012 it in operator< and compare function.
31013 * mem-stats.h (struct mem_usage): Likewise.
31014 * vec.c (struct vec_usage): Remove operator< and compare
31015 function. Can be simply inherited.
31017 2018-01-12 Martin Jambor <mjambor@suse.cz>
31020 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
31021 * tree-ssa-math-opts.c: Include domwalk.h.
31022 (convert_mult_to_fma_1): New function.
31023 (fma_transformation_info): New type.
31024 (fma_deferring_state): Likewise.
31025 (cancel_fma_deferring): New function.
31026 (result_of_phi): Likewise.
31027 (last_fma_candidate_feeds_initial_phi): Likewise.
31028 (convert_mult_to_fma): Added deferring logic, split actual
31029 transformation to convert_mult_to_fma_1.
31030 (math_opts_dom_walker): New type.
31031 (math_opts_dom_walker::after_dom_children): New method, body moved
31032 here from pass_optimize_widening_mul::execute, added deferring logic
31034 (pass_optimize_widening_mul::execute): Moved most of code to
31035 math_opts_dom_walker::after_dom_children.
31036 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
31037 * config/i386/i386.c (ix86_option_override_internal): Added
31038 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
31040 2018-01-12 Richard Biener <rguenther@suse.de>
31043 * dwarf2out.c (gen_variable_die): Do not reset old_die for
31044 inline instance vars.
31046 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
31049 * config/rx/rx.c (rx_is_restricted_memory_address):
31050 Handle SUBREG case.
31052 2018-01-12 Richard Biener <rguenther@suse.de>
31054 PR tree-optimization/80846
31055 * target.def (split_reduction): New target hook.
31056 * targhooks.c (default_split_reduction): New function.
31057 * targhooks.h (default_split_reduction): Declare.
31058 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
31059 target requests first reduce vectors by combining low and high
31061 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
31062 (get_vectype_for_scalar_type_and_size): Export.
31063 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
31064 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
31065 * doc/tm.texi: Regenerate.
31066 * config/i386/i386.c (ix86_split_reduction): Implement
31067 TARGET_VECTORIZE_SPLIT_REDUCTION.
31069 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
31072 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
31073 in PIC mode except for TARGET_VXWORKS_RTP.
31074 * config/sparc/sparc.c: Include cfgrtl.h.
31075 (TARGET_INIT_PIC_REG): Define.
31076 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
31077 (sparc_pic_register_p): New predicate.
31078 (sparc_legitimate_address_p): Use it.
31079 (sparc_legitimize_pic_address): Likewise.
31080 (sparc_delegitimize_address): Likewise.
31081 (sparc_mode_dependent_address_p): Likewise.
31082 (gen_load_pcrel_sym): Remove 4th parameter.
31083 (load_got_register): Adjust call to above. Remove obsolete stuff.
31084 (sparc_expand_prologue): Do not call load_got_register here.
31085 (sparc_flat_expand_prologue): Likewise.
31086 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
31087 (sparc_use_pseudo_pic_reg): New function.
31088 (sparc_init_pic_reg): Likewise.
31089 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
31090 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
31092 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
31094 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
31095 Add item for branch_cost.
31097 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
31099 PR rtl-optimization/83565
31100 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
31101 not extend the result to a larger mode for rotate operations.
31102 (num_sign_bit_copies1): Likewise.
31104 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
31107 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
31109 Use values-Xc.o for -pedantic.
31110 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
31112 2018-01-12 Martin Liska <mliska@suse.cz>
31115 * ipa-devirt.c (final_warning_record::grow_type_warnings):
31117 (possible_polymorphic_call_targets): Use it.
31118 (ipa_devirt): Likewise.
31120 2018-01-12 Martin Liska <mliska@suse.cz>
31122 * profile-count.h (enum profile_quality): Use 0 as invalid
31123 enum value of profile_quality.
31125 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
31127 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
31128 -mext-string options.
31130 2018-01-12 Richard Biener <rguenther@suse.de>
31132 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
31133 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
31134 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
31136 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
31138 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
31140 * configure.ac (--with-long-double-format): Add support for the
31141 configuration option to change the default long double format on
31143 * config.gcc (powerpc*-linux*-*): Likewise.
31144 * configure: Regenerate.
31145 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
31146 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
31147 used without modification.
31149 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
31151 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
31152 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
31153 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
31154 MISC_BUILTIN_SPEC_BARRIER.
31155 (rs6000_init_builtins): Likewise.
31156 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
31158 (speculation_barrier): New define_insn.
31159 * doc/extend.texi: Document __builtin_speculation_barrier.
31161 2018-01-11 Jakub Jelinek <jakub@redhat.com>
31164 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
31165 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
31166 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
31168 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
31169 integral modes instead of "ss" and "sd".
31170 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
31171 vectors with 32-bit and 64-bit elements.
31172 (vecdupssescalarmodesuffix): New mode attribute.
31173 (vec_dup<mode>): Use it.
31175 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
31178 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
31179 frame if argument is passed on stack.
31181 2018-01-11 Jakub Jelinek <jakub@redhat.com>
31184 * ree.c (combine_reaching_defs): Optimize also
31185 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
31186 reg2=any_extend(exp); reg1=reg2;, formatting fix.
31188 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
31190 PR middle-end/83189
31191 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
31193 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
31195 PR middle-end/83718
31196 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
31197 after they are computed.
31199 2018-01-11 Bin Cheng <bin.cheng@arm.com>
31201 PR tree-optimization/83695
31202 * gimple-loop-linterchange.cc
31203 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
31204 reset cached scev information after interchange.
31205 (pass_linterchange::execute): Remove call to scev_reset_htab.
31207 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31209 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
31210 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
31211 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
31212 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
31213 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
31214 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
31215 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
31216 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
31217 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
31218 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
31219 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
31220 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
31221 (V_lane_reg): Likewise.
31222 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
31224 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
31225 (vfmal_lane_low<mode>_intrinsic,
31226 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
31227 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
31228 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
31229 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
31230 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
31231 vfmsl_lane_high<mode>_intrinsic): New define_insns.
31233 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31235 * config/arm/arm-cpus.in (fp16fml): New feature.
31236 (ALL_SIMD): Add fp16fml.
31237 (armv8.2-a): Add fp16fml as an option.
31238 (armv8.3-a): Likewise.
31239 (armv8.4-a): Add fp16fml as part of fp16.
31240 * config/arm/arm.h (TARGET_FP16FML): Define.
31241 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
31243 * config/arm/arm-modes.def (V2HF): Define.
31244 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
31245 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
31246 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
31247 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
31248 vfmsl_low, vfmsl_high): New set of builtins.
31249 * config/arm/iterators.md (PLUSMINUS): New code iterator.
31250 (vfml_op): New code attribute.
31251 (VFMLHALVES): New int iterator.
31252 (VFML, VFMLSEL): New mode attributes.
31253 (V_reg): Define mapping for V2HF.
31254 (V_hi, V_lo): New mode attributes.
31255 (VF_constraint): Likewise.
31256 (vfml_half, vfml_half_selector): New int attributes.
31257 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
31259 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
31260 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
31262 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
31263 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
31264 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
31265 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
31267 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
31268 Document new effective target and option set.
31270 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31272 * config/arm/arm-cpus.in (armv8_4): New feature.
31273 (ARMv8_4a): New fgroup.
31274 (armv8.4-a): New arch.
31275 * config/arm/arm-tables.opt: Regenerate.
31276 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
31277 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
31278 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
31279 Add matching rules for -march=armv8.4-a and extensions.
31280 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
31282 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
31285 * config/rx/rx.md (BW): New mode attribute.
31286 (sync_lock_test_and_setsi): Add mode suffix to insn output.
31288 2018-01-11 Richard Biener <rguenther@suse.de>
31290 PR tree-optimization/83435
31291 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
31292 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
31293 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
31295 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
31296 Alan Hayward <alan.hayward@arm.com>
31297 David Sherwood <david.sherwood@arm.com>
31299 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
31301 (aarch64_classify_address): Initialize it. Track polynomial offsets.
31302 (aarch64_print_address_internal): Use it to check for a zero offset.
31304 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
31305 Alan Hayward <alan.hayward@arm.com>
31306 David Sherwood <david.sherwood@arm.com>
31308 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
31309 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
31310 Return a poly_int64 rather than a HOST_WIDE_INT.
31311 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
31312 rather than a HOST_WIDE_INT.
31313 * config/aarch64/aarch64.h (aarch64_frame): Protect with
31314 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
31315 hard_fp_offset, frame_size, initial_adjust, callee_offset and
31316 final_offset from HOST_WIDE_INT to poly_int64.
31317 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
31318 to_constant when getting the number of units in an Advanced SIMD
31320 (aarch64_builtin_vectorized_function): Check for a constant number
31322 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
31324 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
31325 attribute instead of GET_MODE_NUNITS.
31326 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
31327 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
31328 GET_MODE_SIZE for fixed-size registers.
31329 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
31330 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
31331 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
31332 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
31333 (aarch64_print_operand, aarch64_print_address_internal)
31334 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
31335 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
31336 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
31337 Handle polynomial GET_MODE_SIZE.
31338 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
31339 wider than SImode without modification.
31340 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
31341 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
31342 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
31343 passing and returning SVE modes.
31344 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
31345 rather than GEN_INT.
31346 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
31347 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
31348 (aarch64_allocate_and_probe_stack_space): Likewise.
31349 (aarch64_layout_frame): Cope with polynomial offsets.
31350 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
31351 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
31352 polynomial offsets.
31353 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
31354 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
31355 poly_int64 rather than a HOST_WIDE_INT.
31356 (aarch64_get_separate_components, aarch64_process_components)
31357 (aarch64_expand_prologue, aarch64_expand_epilogue)
31358 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
31359 (aarch64_anchor_offset): New function, split out from...
31360 (aarch64_legitimize_address): ...here.
31361 (aarch64_builtin_vectorization_cost): Handle polynomial
31362 TYPE_VECTOR_SUBPARTS.
31363 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
31365 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
31366 number of elements from the PARALLEL rather than the mode.
31367 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
31368 rather than GET_MODE_BITSIZE.
31369 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
31370 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
31371 (aarch64_expand_vec_perm_const_1): Handle polynomial
31372 d->perm.length () and d->perm elements.
31373 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
31374 Apply to_constant to d->perm elements.
31375 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
31376 polynomial CONST_VECTOR_NUNITS.
31377 (aarch64_move_pointer): Take amount as a poly_int64 rather
31379 (aarch64_progress_pointer): Avoid temporary variable.
31380 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
31381 the mode attribute instead of GET_MODE.
31383 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
31384 Alan Hayward <alan.hayward@arm.com>
31385 David Sherwood <david.sherwood@arm.com>
31387 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
31388 x exists before using it.
31389 (aarch64_add_constant_internal): Rename to...
31390 (aarch64_add_offset_1): ...this. Replace regnum with separate
31391 src and dest rtxes. Handle the case in which they're different,
31392 including when the offset is zero. Replace scratchreg with an rtx.
31393 Use 2 additions if there is no spare register into which we can
31394 move a 16-bit constant.
31395 (aarch64_add_constant): Delete.
31396 (aarch64_add_offset): Replace reg with separate src and dest
31397 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
31398 Use aarch64_add_offset_1.
31399 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
31400 an rtx rather than an int. Take the delta as a poly_int64
31401 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
31402 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
31403 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
31404 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
31405 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
31406 and aarch64_add_sp.
31407 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
31408 aarch64_add_constant.
31410 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
31412 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
31413 Use scalar_float_mode.
31415 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
31417 * config/aarch64/aarch64-simd.md
31418 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
31419 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
31420 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
31421 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
31422 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
31423 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
31424 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
31425 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
31426 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
31427 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
31429 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
31432 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
31433 targ_options->x_arm_arch_string is non NULL.
31435 2018-01-11 Tamar Christina <tamar.christina@arm.com>
31437 * config/aarch64/aarch64.h
31438 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
31440 2018-01-11 Sudakshina Das <sudi.das@arm.com>
31443 * expmed.c (emit_store_flag_force): Swap if const op0
31444 and change VOIDmode to mode of op0.
31446 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
31448 PR rtl-optimization/83761
31449 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
31450 than bytes to mode_for_size.
31452 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
31454 PR middle-end/83189
31455 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
31456 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
31459 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
31461 PR middle-end/83575
31462 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
31463 when in layout mode.
31464 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
31465 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
31468 2018-01-10 Michael Collison <michael.collison@arm.com>
31470 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
31471 * config/aarch64/aarch64-option-extension.def: Add
31472 AARCH64_OPT_EXTENSION of 'fp16fml'.
31473 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
31474 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
31475 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
31476 * config/aarch64/constraints.md (Ui7): New constraint.
31477 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
31478 (VFMLA_SEL_W): Ditto.
31481 (VFMLA16_LOW): New int iterator.
31482 (VFMLA16_HIGH): Ditto.
31483 (UNSPEC_FMLAL): New unspec.
31484 (UNSPEC_FMLSL): Ditto.
31485 (UNSPEC_FMLAL2): Ditto.
31486 (UNSPEC_FMLSL2): Ditto.
31487 (f16mac): New code attribute.
31488 * config/aarch64/aarch64-simd-builtins.def
31489 (aarch64_fmlal_lowv2sf): Ditto.
31490 (aarch64_fmlsl_lowv2sf): Ditto.
31491 (aarch64_fmlalq_lowv4sf): Ditto.
31492 (aarch64_fmlslq_lowv4sf): Ditto.
31493 (aarch64_fmlal_highv2sf): Ditto.
31494 (aarch64_fmlsl_highv2sf): Ditto.
31495 (aarch64_fmlalq_highv4sf): Ditto.
31496 (aarch64_fmlslq_highv4sf): Ditto.
31497 (aarch64_fmlal_lane_lowv2sf): Ditto.
31498 (aarch64_fmlsl_lane_lowv2sf): Ditto.
31499 (aarch64_fmlal_laneq_lowv2sf): Ditto.
31500 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
31501 (aarch64_fmlalq_lane_lowv4sf): Ditto.
31502 (aarch64_fmlsl_lane_lowv4sf): Ditto.
31503 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
31504 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
31505 (aarch64_fmlal_lane_highv2sf): Ditto.
31506 (aarch64_fmlsl_lane_highv2sf): Ditto.
31507 (aarch64_fmlal_laneq_highv2sf): Ditto.
31508 (aarch64_fmlsl_laneq_highv2sf): Ditto.
31509 (aarch64_fmlalq_lane_highv4sf): Ditto.
31510 (aarch64_fmlsl_lane_highv4sf): Ditto.
31511 (aarch64_fmlalq_laneq_highv4sf): Ditto.
31512 (aarch64_fmlsl_laneq_highv4sf): Ditto.
31513 * config/aarch64/aarch64-simd.md:
31514 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
31515 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
31516 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
31517 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
31518 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
31519 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
31520 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
31521 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
31522 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
31523 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
31524 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
31525 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
31526 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
31527 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
31528 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
31529 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
31530 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
31531 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
31532 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
31533 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
31534 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
31535 (vfmlsl_low_u32): Ditto.
31536 (vfmlalq_low_u32): Ditto.
31537 (vfmlslq_low_u32): Ditto.
31538 (vfmlal_high_u32): Ditto.
31539 (vfmlsl_high_u32): Ditto.
31540 (vfmlalq_high_u32): Ditto.
31541 (vfmlslq_high_u32): Ditto.
31542 (vfmlal_lane_low_u32): Ditto.
31543 (vfmlsl_lane_low_u32): Ditto.
31544 (vfmlal_laneq_low_u32): Ditto.
31545 (vfmlsl_laneq_low_u32): Ditto.
31546 (vfmlalq_lane_low_u32): Ditto.
31547 (vfmlslq_lane_low_u32): Ditto.
31548 (vfmlalq_laneq_low_u32): Ditto.
31549 (vfmlslq_laneq_low_u32): Ditto.
31550 (vfmlal_lane_high_u32): Ditto.
31551 (vfmlsl_lane_high_u32): Ditto.
31552 (vfmlal_laneq_high_u32): Ditto.
31553 (vfmlsl_laneq_high_u32): Ditto.
31554 (vfmlalq_lane_high_u32): Ditto.
31555 (vfmlslq_lane_high_u32): Ditto.
31556 (vfmlalq_laneq_high_u32): Ditto.
31557 (vfmlslq_laneq_high_u32): Ditto.
31558 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
31559 (AARCH64_FL_FOR_ARCH8_4): New.
31560 (AARCH64_ISA_F16FML): New ISA flag.
31561 (TARGET_F16FML): New feature flag for fp16fml.
31562 (doc/invoke.texi): Document new fp16fml option.
31564 2018-01-10 Michael Collison <michael.collison@arm.com>
31566 * config/aarch64/aarch64-builtins.c:
31567 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
31568 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
31569 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
31570 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
31571 (AARCH64_ISA_SHA3): New ISA flag.
31572 (TARGET_SHA3): New feature flag for sha3.
31573 * config/aarch64/iterators.md (sha512_op): New int attribute.
31574 (CRYPTO_SHA512): New int iterator.
31575 (UNSPEC_SHA512H): New unspec.
31576 (UNSPEC_SHA512H2): Ditto.
31577 (UNSPEC_SHA512SU0): Ditto.
31578 (UNSPEC_SHA512SU1): Ditto.
31579 * config/aarch64/aarch64-simd-builtins.def
31580 (aarch64_crypto_sha512hqv2di): New builtin.
31581 (aarch64_crypto_sha512h2qv2di): Ditto.
31582 (aarch64_crypto_sha512su0qv2di): Ditto.
31583 (aarch64_crypto_sha512su1qv2di): Ditto.
31584 (aarch64_eor3qv8hi): Ditto.
31585 (aarch64_rax1qv2di): Ditto.
31586 (aarch64_xarqv2di): Ditto.
31587 (aarch64_bcaxqv8hi): Ditto.
31588 * config/aarch64/aarch64-simd.md:
31589 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
31590 (aarch64_crypto_sha512su0qv2di): Ditto.
31591 (aarch64_crypto_sha512su1qv2di): Ditto.
31592 (aarch64_eor3qv8hi): Ditto.
31593 (aarch64_rax1qv2di): Ditto.
31594 (aarch64_xarqv2di): Ditto.
31595 (aarch64_bcaxqv8hi): Ditto.
31596 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
31597 (vsha512h2q_u64): Ditto.
31598 (vsha512su0q_u64): Ditto.
31599 (vsha512su1q_u64): Ditto.
31600 (veor3q_u16): Ditto.
31601 (vrax1q_u64): Ditto.
31602 (vxarq_u64): Ditto.
31603 (vbcaxq_u16): Ditto.
31604 * config/arm/types.md (crypto_sha512): New type attribute.
31605 (crypto_sha3): Ditto.
31606 (doc/invoke.texi): Document new sha3 option.
31608 2018-01-10 Michael Collison <michael.collison@arm.com>
31610 * config/aarch64/aarch64-builtins.c:
31611 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
31612 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
31613 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
31614 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
31615 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
31616 (AARCH64_ISA_SM4): New ISA flag.
31617 (TARGET_SM4): New feature flag for sm4.
31618 * config/aarch64/aarch64-simd-builtins.def
31619 (aarch64_sm3ss1qv4si): Ditto.
31620 (aarch64_sm3tt1aq4si): Ditto.
31621 (aarch64_sm3tt1bq4si): Ditto.
31622 (aarch64_sm3tt2aq4si): Ditto.
31623 (aarch64_sm3tt2bq4si): Ditto.
31624 (aarch64_sm3partw1qv4si): Ditto.
31625 (aarch64_sm3partw2qv4si): Ditto.
31626 (aarch64_sm4eqv4si): Ditto.
31627 (aarch64_sm4ekeyqv4si): Ditto.
31628 * config/aarch64/aarch64-simd.md:
31629 (aarch64_sm3ss1qv4si): Ditto.
31630 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
31631 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
31632 (aarch64_sm4eqv4si): Ditto.
31633 (aarch64_sm4ekeyqv4si): Ditto.
31634 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
31635 (sm3part_op): Ditto.
31636 (CRYPTO_SM3TT): Ditto.
31637 (CRYPTO_SM3PART): Ditto.
31638 (UNSPEC_SM3SS1): New unspec.
31639 (UNSPEC_SM3TT1A): Ditto.
31640 (UNSPEC_SM3TT1B): Ditto.
31641 (UNSPEC_SM3TT2A): Ditto.
31642 (UNSPEC_SM3TT2B): Ditto.
31643 (UNSPEC_SM3PARTW1): Ditto.
31644 (UNSPEC_SM3PARTW2): Ditto.
31645 (UNSPEC_SM4E): Ditto.
31646 (UNSPEC_SM4EKEY): Ditto.
31647 * config/aarch64/constraints.md (Ui2): New constraint.
31648 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
31649 * config/arm/types.md (crypto_sm3): New type attribute.
31650 (crypto_sm4): Ditto.
31651 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
31652 (vsm3tt1aq_u32): Ditto.
31653 (vsm3tt1bq_u32): Ditto.
31654 (vsm3tt2aq_u32): Ditto.
31655 (vsm3tt2bq_u32): Ditto.
31656 (vsm3partw1q_u32): Ditto.
31657 (vsm3partw2q_u32): Ditto.
31658 (vsm4eq_u32): Ditto.
31659 (vsm4ekeyq_u32): Ditto.
31660 (doc/invoke.texi): Document new sm4 option.
31662 2018-01-10 Michael Collison <michael.collison@arm.com>
31664 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
31665 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
31666 (AARCH64_FL_FOR_ARCH8_4): New.
31667 (AARCH64_FL_V8_4): New flag.
31668 (doc/invoke.texi): Document new armv8.4-a option.
31670 2018-01-10 Michael Collison <michael.collison@arm.com>
31672 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
31673 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
31674 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
31675 * config/aarch64/aarch64-option-extension.def: Add
31676 AARCH64_OPT_EXTENSION of 'sha2'.
31677 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
31678 (crypto): Disable sha2 and aes if crypto disabled.
31679 (crypto): Enable aes and sha2 if enabled.
31680 (simd): Disable sha2 and aes if simd disabled.
31681 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
31683 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
31684 (TARGET_SHA2): New feature flag for sha2.
31685 (TARGET_AES): New feature flag for aes.
31686 * config/aarch64/aarch64-simd.md:
31687 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
31688 conditional on TARGET_AES.
31689 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
31690 (aarch64_crypto_sha1hsi): Make pattern conditional
31692 (aarch64_crypto_sha1hv4si): Ditto.
31693 (aarch64_be_crypto_sha1hv4si): Ditto.
31694 (aarch64_crypto_sha1su1v4si): Ditto.
31695 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
31696 (aarch64_crypto_sha1su0v4si): Ditto.
31697 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
31698 (aarch64_crypto_sha256su0v4si): Ditto.
31699 (aarch64_crypto_sha256su1v4si): Ditto.
31700 (doc/invoke.texi): Document new aes and sha2 options.
31702 2018-01-10 Martin Sebor <msebor@redhat.com>
31704 PR tree-optimization/83781
31705 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
31708 2018-01-11 Martin Sebor <msebor@gmail.com>
31709 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
31711 PR tree-optimization/83501
31712 PR tree-optimization/81703
31714 * tree-ssa-strlen.c (get_string_cst): Rename...
31715 (get_string_len): ...to this. Handle global constants.
31716 (handle_char_store): Adjust.
31718 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
31719 Jim Wilson <jimw@sifive.com>
31721 * config/riscv/riscv-protos.h (riscv_output_return): New.
31722 * config/riscv/riscv.c (struct machine_function): New naked_p field.
31723 (riscv_attribute_table, riscv_output_return),
31724 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
31725 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
31726 (riscv_compute_frame_info): Only compute frame->mask if not a naked
31728 (riscv_expand_prologue): Add early return for naked function.
31729 (riscv_expand_epilogue): Likewise.
31730 (riscv_function_ok_for_sibcall): Return false for naked function.
31731 (riscv_set_current_function): New.
31732 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
31733 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
31734 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
31735 * doc/extend.texi (RISC-V Function Attributes): New.
31737 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
31739 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
31740 check for 128-bit long double before checking TCmode.
31741 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
31742 128-bit long doubles before checking TFmode or TCmode.
31743 (FLOAT128_IBM_P): Likewise.
31745 2018-01-10 Martin Sebor <msebor@redhat.com>
31747 PR tree-optimization/83671
31748 * builtins.c (c_strlen): Unconditionally return zero for the empty
31750 Use -Warray-bounds for warnings.
31751 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
31752 for non-constant array indices with COMPONENT_REF, arrays of
31753 arrays, and pointers to arrays.
31754 (gimple_fold_builtin_strlen): Determine and set length range for
31755 non-constant character arrays.
31757 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
31759 PR middle-end/81897
31760 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
31763 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
31765 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
31767 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
31770 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
31771 VECTOR_MEM_ALTIVEC_OR_VSX_P.
31772 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
31773 indexed_or_indirect_operand predicate.
31774 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
31775 (*vsx_le_perm_load_v8hi): Likewise.
31776 (*vsx_le_perm_load_v16qi): Likewise.
31777 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
31778 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
31779 (*vsx_le_perm_store_v8hi): Likewise.
31780 (*vsx_le_perm_store_v16qi): Likewise.
31781 (eight unnamed splitters): Likewise.
31783 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
31785 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
31786 * config/rs6000/emmintrin.h: Likewise.
31787 * config/rs6000/mmintrin.h: Likewise.
31788 * config/rs6000/xmmintrin.h: Likewise.
31790 2018-01-10 David Malcolm <dmalcolm@redhat.com>
31793 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
31795 * tree.c (tree_nop_conversion): Return true for location wrapper
31797 (maybe_wrap_with_location): New function.
31798 (selftest::check_strip_nops): New function.
31799 (selftest::test_location_wrappers): New function.
31800 (selftest::tree_c_tests): Call it.
31801 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
31802 (maybe_wrap_with_location): New decl.
31803 (EXPR_LOCATION_WRAPPER_P): New macro.
31804 (location_wrapper_p): New inline function.
31805 (tree_strip_any_location_wrapper): New inline function.
31807 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
31810 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
31811 stack_realign_offset for the largest alignment of stack slot
31813 (ix86_find_max_used_stack_alignment): New function.
31814 (ix86_finalize_stack_frame_flags): Use it. Set
31815 max_used_stack_alignment if we don't realign stack.
31816 * config/i386/i386.h (machine_function): Add
31817 max_used_stack_alignment.
31819 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
31821 * config/arm/arm.opt (-mbranch-cost): New option.
31822 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
31825 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
31828 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
31829 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
31831 2018-01-10 Richard Biener <rguenther@suse.de>
31834 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
31835 early out so it also covers the case where we have a non-NULL
31838 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
31840 PR tree-optimization/83753
31841 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
31842 for non-strided grouped accesses if the number of elements is 1.
31844 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
31847 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
31848 * i386.h (TARGET_USE_GATHER): Define.
31849 * x86-tune.def (X86_TUNE_USE_GATHER): New.
31851 2018-01-10 Martin Liska <mliska@suse.cz>
31854 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
31855 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
31857 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
31858 CLEANUP_NO_PARTITIONING is not set.
31860 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
31862 * doc/rtl.texi: Remove documentation of (const ...) wrappers
31863 for vectors, as a partial revert of r254296.
31864 * rtl.h (const_vec_p): Delete.
31865 (const_vec_duplicate_p): Don't test for vector CONSTs.
31866 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
31867 * expmed.c (make_tree): Likewise.
31870 * common.md (E, F): Use CONSTANT_P instead of checking for
31872 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
31873 checking for CONST_VECTOR.
31875 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
31877 PR middle-end/83575
31878 * predict.c (force_edge_cold): Handle in more sane way edges
31879 with no prediction.
31881 2018-01-09 Carl Love <cel@us.ibm.com>
31883 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
31885 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
31886 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
31887 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
31888 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
31889 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
31890 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
31891 * config/rs6000/rs6000-protos.h: Add extern defition for
31892 rs6000_generate_float2_double_code.
31893 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
31895 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
31896 (float2_v2df): Add define_expand.
31898 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
31901 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
31902 op_mode in the force_to_mode call.
31904 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
31906 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
31907 instead of checking each element individually.
31908 (aarch64_evpc_uzp): Likewise.
31909 (aarch64_evpc_zip): Likewise.
31910 (aarch64_evpc_ext): Likewise.
31911 (aarch64_evpc_rev): Likewise.
31912 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
31913 instead of checking each element individually. Return true without
31915 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
31916 whether all selected elements come from the same input, instead of
31917 checking each element individually. Remove calls to gen_rtx_REG,
31918 start_sequence and end_sequence and instead assert that no rtl is
31921 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
31923 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
31924 order of HIGH and CONST checks.
31926 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
31928 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
31929 if the destination isn't an SSA_NAME.
31931 2018-01-09 Richard Biener <rguenther@suse.de>
31933 PR tree-optimization/83668
31934 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
31936 (canonicalize_loop_form): ... here, renamed from ...
31937 (canonicalize_loop_closed_ssa_form): ... this and amended to
31938 swap successor edges for loop exit blocks to make us use
31939 the RPO order we need for initial schedule generation.
31941 2018-01-09 Joseph Myers <joseph@codesourcery.com>
31943 PR tree-optimization/64811
31944 * match.pd: When optimizing comparisons with Inf, avoid
31945 introducing or losing exceptions from comparisons with NaN.
31947 2018-01-09 Martin Liska <mliska@suse.cz>
31950 * asan.c (shadow_mem_size): Add gcc_assert.
31952 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
31954 Don't save registers in main().
31957 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
31958 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
31959 * config/avr/avr.c (avr_set_current_function): Don't error if
31960 naked, OS_task or OS_main are specified at the same time.
31961 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
31963 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
31965 * common/config/avr/avr-common.c (avr_option_optimization_table):
31966 Switch on -mmain-is-OS_task for optimizing compilations.
31968 2018-01-09 Richard Biener <rguenther@suse.de>
31970 PR tree-optimization/83572
31971 * graphite.c: Include cfganal.h.
31972 (graphite_transform_loops): Connect infinite loops to exit
31973 and remove fake edges at the end.
31975 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
31977 * ipa-inline.c (edge_badness): Revert accidental checkin.
31979 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
31982 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
31983 symbols; not inline clones.
31985 2018-01-09 Jakub Jelinek <jakub@redhat.com>
31988 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
31989 hard registers. Formatting fixes.
31991 PR preprocessor/83722
31992 * gcc.c (try_generate_repro): Pass
31993 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
31994 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
31997 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
31998 Kito Cheng <kito.cheng@gmail.com>
32000 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
32001 (riscv_leaf_function_p): Delete.
32002 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
32004 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
32006 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
32008 (do_ifelse): New function.
32009 (do_isel): New function.
32010 (do_sub3): New function.
32011 (do_add3): New function.
32012 (do_load_mask_compare): New function.
32013 (do_overlap_load_compare): New function.
32014 (expand_compare_loop): New function.
32015 (expand_block_compare): Call expand_compare_loop() when appropriate.
32016 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
32017 option description.
32018 (-mblock-compare-inline-loop-limit): New option.
32020 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
32023 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
32024 Reverse order of second and third operands in first alternative.
32025 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
32026 of first and second elements in UNSPEC_VPERMR vector.
32027 (altivec_expand_vec_perm_le): Likewise.
32029 2018-01-08 Jeff Law <law@redhat.com>
32031 PR rtl-optimizatin/81308
32032 * tree-switch-conversion.c (cfg_altered): New file scoped static.
32033 (process_switch): If group_case_labels makes a change, then set
32035 (pass_convert_switch::execute): If a switch is converted, then
32036 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
32038 PR rtl-optimization/81308
32039 * recog.c (split_all_insns): Conditionally cleanup the CFG after
32042 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
32044 PR target/83663 - Revert r255946
32045 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
32046 generation for cases where splatting a value is not useful.
32047 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
32048 across a vec_duplicate and a paradoxical subreg forming a vector
32049 mode to a vec_concat.
32051 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32053 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
32054 -march=armv8.3-a variants.
32055 * config/arm/t-multilib: Likewise.
32056 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
32058 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
32060 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
32062 (cceq_ior_compare_complement): Give it a name so I can use it, and
32063 change boolean_or_operator predicate to boolean_operator so it can
32064 be used to generate a crand.
32065 (eqne): New code iterator.
32066 (bd/bd_neg): New code_attrs.
32067 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
32068 a single define_insn.
32069 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
32070 decrement (bdnzt/bdnzf/bdzt/bdzf).
32071 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
32072 with the new names of the branch decrement patterns, and added the
32073 names of the branch decrement conditional patterns.
32075 2018-01-08 Richard Biener <rguenther@suse.de>
32077 PR tree-optimization/83563
32078 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
32081 2018-01-08 Richard Biener <rguenther@suse.de>
32083 PR middle-end/83713
32084 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
32086 2018-01-08 Richard Biener <rguenther@suse.de>
32088 PR tree-optimization/83685
32089 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
32090 references to abnormals.
32092 2018-01-08 Richard Biener <rguenther@suse.de>
32095 * dwarf2out.c (output_indirect_strings): Handle empty
32096 skeleton_debug_str_hash.
32097 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
32099 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
32101 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
32102 (emit_store_direct): Likewise.
32103 (arc_trampoline_adjust_address): Likewise.
32104 (arc_asm_trampoline_template): New function.
32105 (arc_initialize_trampoline): Use asm_trampoline_template.
32106 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
32107 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
32108 * config/arc/arc.md (flush_icache): Delete pattern.
32110 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
32112 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
32113 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
32116 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
32119 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
32120 by not USED_FOR_TARGET.
32121 (make_pass_resolve_sw_modes): Likewise.
32123 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
32125 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
32128 2018-01-08 Richard Biener <rguenther@suse.de>
32130 PR middle-end/83580
32131 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
32133 2018-01-08 Richard Biener <rguenther@suse.de>
32135 PR middle-end/83517
32136 * match.pd ((t * 2) / 2) -> t): Add missing :c.
32138 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
32140 PR middle-end/81897
32141 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
32142 basic blocks with a small number of successors.
32143 (convert_control_dep_chain_into_preds): Improve handling of
32145 (dump_predicates): Split apart into...
32146 (dump_pred_chain): ...here...
32147 (dump_pred_info): ...and here.
32148 (can_one_predicate_be_invalidated_p): Add debugging printfs.
32149 (can_chain_union_be_invalidated_p): Improve check for invalidation
32151 (uninit_uses_cannot_happen): Avoid unnecessary if
32152 convert_control_dep_chain_into_preds yielded nothing.
32154 2018-01-06 Martin Sebor <msebor@redhat.com>
32156 PR tree-optimization/83640
32157 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
32158 subtracting negative offset from size.
32159 (builtin_access::overlap): Adjust offset bounds of the access to fall
32160 within the size of the object if possible.
32162 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
32164 PR rtl-optimization/83699
32165 * expmed.c (extract_bit_field_1): Restrict the vector usage of
32166 extract_bit_field_as_subreg to cases in which the extracted
32167 value is also a vector.
32169 * lra-constraints.c (process_alt_operands): Test for the equivalence
32170 substitutions when detecting a possible reload cycle.
32172 2018-01-06 Jakub Jelinek <jakub@redhat.com>
32175 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
32176 by default if flag_selective_schedling{,2}. Formatting fixes.
32178 PR rtl-optimization/83682
32179 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
32180 if it has non-VECTOR_MODE element mode.
32181 (vec_duplicate_p): Likewise.
32183 PR middle-end/83694
32184 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
32185 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
32187 2018-01-05 Jakub Jelinek <jakub@redhat.com>
32190 * config/i386/i386-builtin.def
32191 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
32192 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
32193 Require also OPTION_MASK_ISA_AVX512F in addition to
32194 OPTION_MASK_ISA_GFNI.
32195 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
32196 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
32197 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
32198 to OPTION_MASK_ISA_GFNI.
32199 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
32200 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
32201 OPTION_MASK_ISA_AVX512BW.
32202 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
32203 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
32204 addition to OPTION_MASK_ISA_GFNI.
32205 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
32206 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
32207 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
32208 to OPTION_MASK_ISA_GFNI.
32209 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
32210 a requirement for all ISAs rather than any of them with a few
32212 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
32214 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
32215 bitmasks to be enabled with 3 exceptions, instead of requiring any
32216 enabled ISA with lots of exceptions.
32217 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
32218 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
32219 Change avx512bw in isa attribute to avx512f.
32220 * config/i386/sgxintrin.h: Add license boilerplate.
32221 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
32222 to __AVX512F__ and __AVX512VL to __AVX512VL__.
32223 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
32224 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
32226 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
32227 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
32228 temporarily sse2 rather than sse if not enabled already.
32231 * config/i386/sse.md (VI248_VLBW): Rename to ...
32232 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
32233 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
32234 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
32235 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
32236 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
32237 mode iterator instead of VI248_VLBW.
32239 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
32241 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
32242 (record_modified): Skip clobbers; add debug output.
32243 (param_change_prob): Use sreal frequencies.
32245 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
32247 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
32248 punt for user-aligned variables.
32250 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
32252 * tree-chrec.c (chrec_contains_symbols): Return true for
32255 2018-01-05 Sudakshina Das <sudi.das@arm.com>
32258 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
32259 of (x|y) == x for BICS pattern.
32261 2018-01-05 Jakub Jelinek <jakub@redhat.com>
32263 PR tree-optimization/83605
32264 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
32265 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
32268 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
32270 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
32271 * config/epiphany/rtems.h: New file.
32273 2018-01-04 Jakub Jelinek <jakub@redhat.com>
32274 Uros Bizjak <ubizjak@gmail.com>
32277 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
32278 QIreg_operand instead of register_operand predicate.
32279 * config/i386/i386.c (ix86_rop_should_change_byte_p,
32280 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
32281 comments instead of -fmitigate[-_]rop.
32283 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
32286 * cgraphunit.c (symbol_table::compile): Switch to text_section
32287 before calling assembly_start debug hook.
32288 * run-rtl-passes.c (run_rtl_passes): Likewise.
32291 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
32293 * tree-vrp.c (extract_range_from_binary_expr_1): Check
32294 range_int_cst_p rather than !symbolic_range_p before calling
32295 extract_range_from_multiplicative_op_1.
32297 2018-01-04 Jeff Law <law@redhat.com>
32299 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
32300 redundant test in assertion.
32302 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
32304 * doc/rtl.texi: Document machine_mode wrapper classes.
32306 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
32308 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
32309 using tree_to_uhwi.
32311 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
32313 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
32314 the VEC_PERM_EXPR fold to fail.
32316 2018-01-04 Jakub Jelinek <jakub@redhat.com>
32319 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
32320 to switched_sections.
32322 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
32325 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
32326 test for d.testing.
32328 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
32331 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
32332 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
32334 2018-01-04 Jakub Jelinek <jakub@redhat.com>
32337 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
32338 is BLKmode and bitpos not zero or mode change is needed.
32340 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
32343 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
32346 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
32349 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
32350 instead of MULT rtx. Update all corresponding splitters.
32351 (*saddl_se): Ditto.
32352 (*ssub<modesuffix>): Ditto.
32353 (*ssubl_se): Ditto.
32354 (*cmp_sadd_di): Update split patterns.
32355 (*cmp_sadd_si): Ditto.
32356 (*cmp_sadd_sidi): Ditto.
32357 (*cmp_ssub_di): Ditto.
32358 (*cmp_ssub_si): Ditto.
32359 (*cmp_ssub_sidi): Ditto.
32360 * config/alpha/predicates.md (const23_operand): New predicate.
32361 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
32362 Look for ASHIFT, not MULT inner operand.
32363 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
32365 2018-01-04 Martin Liska <mliska@suse.cz>
32367 PR gcov-profile/83669
32368 * gcov.c (output_intermediate_file): Add version to intermediate
32370 * doc/gcov.texi: Document new field 'version' in intermediate
32371 file format. Fix location of '-k' option of gcov command.
32373 2018-01-04 Martin Liska <mliska@suse.cz>
32376 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
32378 2018-01-04 Jakub Jelinek <jakub@redhat.com>
32380 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
32382 2018-01-03 Martin Sebor <msebor@redhat.com>
32384 PR tree-optimization/83655
32385 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
32386 checking calls with invalid arguments.
32388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32390 * tree-vect-stmts.c (vect_get_store_rhs): New function.
32391 (vectorizable_mask_load_store): Delete.
32392 (vectorizable_call): Return false for masked loads and stores.
32393 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
32394 instead of gimple_assign_rhs1.
32395 (vectorizable_load): Handle IFN_MASK_LOAD.
32396 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
32398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32400 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
32402 (vectorizable_mask_load_store): ...here.
32403 (vectorizable_load): ...and here.
32405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32407 * tree-vect-stmts.c (vect_build_all_ones_mask)
32408 (vect_build_zero_merge_argument): New functions, split out from...
32409 (vectorizable_load): ...here.
32411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32413 * tree-vect-stmts.c (vect_check_store_rhs): New function,
32415 (vectorizable_mask_load_store): ...here.
32416 (vectorizable_store): ...and here.
32418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32420 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
32422 (vectorizable_mask_load_store): ...here.
32424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32426 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
32427 (vect_model_store_cost): Take a vec_load_store_type instead of a
32429 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
32430 (vect_model_store_cost): Take a vec_load_store_type instead of a
32432 (vectorizable_mask_load_store): Update accordingly.
32433 (vectorizable_store): Likewise.
32434 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
32436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32438 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
32439 IFN_MASK_LOAD calls here rather than...
32440 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
32442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32443 Alan Hayward <alan.hayward@arm.com>
32444 David Sherwood <david.sherwood@arm.com>
32446 * expmed.c (extract_bit_field_1): For vector extracts,
32447 fall back to extract_bit_field_as_subreg if vec_extract
32450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32451 Alan Hayward <alan.hayward@arm.com>
32452 David Sherwood <david.sherwood@arm.com>
32454 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
32455 they are variable or constant sized.
32456 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
32457 slots for constant-sized data.
32459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32460 Alan Hayward <alan.hayward@arm.com>
32461 David Sherwood <david.sherwood@arm.com>
32463 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
32464 handling COND_EXPRs with boolean comparisons, try to find a better
32465 basis for the mask type than the boolean itself.
32467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32469 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
32470 is calculated and how it can be overridden.
32471 * genmodes.c (max_bitsize_mode_any_mode): New variable.
32472 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
32474 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
32477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32478 Alan Hayward <alan.hayward@arm.com>
32479 David Sherwood <david.sherwood@arm.com>
32481 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
32482 Remove the mode argument.
32483 (aarch64_simd_valid_immediate): Remove the mode and inverse
32485 * config/aarch64/iterators.md (bitsize): New iterator.
32486 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
32487 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
32488 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
32489 aarch64_simd_valid_immediate.
32490 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
32491 (aarch64_reg_or_bic_imm): Likewise.
32492 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
32493 with an insn_type enum and msl with a modifier_type enum.
32494 Replace element_width with a scalar_mode. Change the shift
32495 to unsigned int. Add constructors for scalar_float_mode and
32496 scalar_int_mode elements.
32497 (aarch64_vect_float_const_representable_p): Delete.
32498 (aarch64_can_const_movi_rtx_p)
32499 (aarch64_simd_scalar_immediate_valid_for_move)
32500 (aarch64_simd_make_constant): Update call to
32501 aarch64_simd_valid_immediate.
32502 (aarch64_advsimd_valid_immediate_hs): New function.
32503 (aarch64_advsimd_valid_immediate): Likewise.
32504 (aarch64_simd_valid_immediate): Remove mode and inverse
32505 arguments. Rewrite to use the above. Use const_vec_duplicate_p
32506 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
32507 and aarch64_float_const_representable_p on the result.
32508 (aarch64_output_simd_mov_immediate): Remove mode argument.
32509 Update call to aarch64_simd_valid_immediate and use of
32510 simd_immediate_info.
32511 (aarch64_output_scalar_simd_mov_immediate): Update call
32514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32515 Alan Hayward <alan.hayward@arm.com>
32516 David Sherwood <david.sherwood@arm.com>
32518 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
32519 (mode_nunits): Likewise CONST_MODE_NUNITS.
32520 * machmode.def (ADJUST_NUNITS): Document.
32521 * genmodes.c (mode_data::need_nunits_adj): New field.
32522 (blank_mode): Update accordingly.
32523 (adj_nunits): New variable.
32524 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
32526 (emit_mode_size_inline): Set need_bytesize_adj for all modes
32527 listed in adj_nunits.
32528 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
32529 listed in adj_nunits. Don't emit case statements for such modes.
32530 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
32531 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
32532 nothing if adj_nunits is nonnull.
32533 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
32534 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
32535 (emit_mode_fbit): Update use of print_maybe_const_decl.
32536 (emit_move_size): Likewise. Treat the array as non-const
32538 (emit_mode_adjustments): Handle adj_nunits.
32540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32542 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
32543 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
32544 (VECTOR_MODES): Use it.
32545 (make_vector_modes): Take the prefix as an argument.
32547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32548 Alan Hayward <alan.hayward@arm.com>
32549 David Sherwood <david.sherwood@arm.com>
32551 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
32552 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
32553 for MODE_VECTOR_BOOL.
32554 * machmode.def (VECTOR_BOOL_MODE): Document.
32555 * genmodes.c (VECTOR_BOOL_MODE): New macro.
32556 (make_vector_bool_mode): New function.
32557 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
32559 * lto-streamer-in.c (lto_input_mode_table): Likewise.
32560 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
32562 * stor-layout.c (int_mode_for_mode): Likewise.
32563 * tree.c (build_vector_type_for_mode): Likewise.
32564 * varasm.c (output_constant_pool_2): Likewise.
32565 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
32566 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
32567 for MODE_VECTOR_BOOL.
32568 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
32569 of mode class checks.
32570 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
32571 instead of a list of mode class checks.
32572 (expand_vector_scalar_condition): Likewise.
32573 (type_for_widest_vector_mode): Handle BImode as an inner mode.
32575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32576 Alan Hayward <alan.hayward@arm.com>
32577 David Sherwood <david.sherwood@arm.com>
32579 * machmode.h (mode_size): Change from unsigned short to
32581 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
32582 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
32583 or if measurement_type is not polynomial.
32584 (fixed_size_mode::includes_p): Check for constant-sized modes.
32585 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
32586 return a poly_uint16 rather than an unsigned short.
32587 (emit_mode_size): Change the type of mode_size from unsigned short
32588 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
32589 (emit_mode_adjustments): Cope with polynomial vector sizes.
32590 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
32592 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
32594 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
32595 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
32596 * caller-save.c (setup_save_areas): Likewise.
32597 (replace_reg_with_saved_mem): Likewise.
32598 * calls.c (emit_library_call_value_1): Likewise.
32599 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
32600 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
32601 (gen_lowpart_for_combine): Likewise.
32602 * convert.c (convert_to_integer_1): Likewise.
32603 * cse.c (equiv_constant, cse_insn): Likewise.
32604 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
32605 (cselib_subst_to_values): Likewise.
32606 * dce.c (word_dce_process_block): Likewise.
32607 * df-problems.c (df_word_lr_mark_ref): Likewise.
32608 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
32609 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
32610 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
32611 (rtl_for_decl_location): Likewise.
32612 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
32613 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
32614 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
32615 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
32616 (expand_expr_real_1): Likewise.
32617 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
32618 (pad_below): Likewise.
32619 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
32620 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
32621 * ira.c (get_subreg_tracking_sizes): Likewise.
32622 * ira-build.c (ira_create_allocno_objects): Likewise.
32623 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
32624 (ira_sort_regnos_for_alter_reg): Likewise.
32625 * ira-costs.c (record_operand_costs): Likewise.
32626 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
32627 (resolve_simple_move): Likewise.
32628 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
32629 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
32630 (lra_constraints): Likewise.
32631 (CONST_POOL_OK_P): Reject variable-sized modes.
32632 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
32633 (add_pseudo_to_slot, lra_spill): Likewise.
32634 * omp-low.c (omp_clause_aligned_alignment): Likewise.
32635 * optabs-query.c (get_best_extraction_insn): Likewise.
32636 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
32637 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
32638 (expand_mult_highpart, valid_multiword_target_p): Likewise.
32639 * recog.c (offsettable_address_addr_space_p): Likewise.
32640 * regcprop.c (maybe_mode_change): Likewise.
32641 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
32642 * regrename.c (build_def_use): Likewise.
32643 * regstat.c (dump_reg_info): Likewise.
32644 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
32645 (find_reloads, find_reloads_subreg_address): Likewise.
32646 * reload1.c (eliminate_regs_1): Likewise.
32647 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
32648 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
32649 (simplify_binary_operation_1, simplify_subreg): Likewise.
32650 * targhooks.c (default_function_arg_padding): Likewise.
32651 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
32652 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
32653 (verify_gimple_assign_ternary): Likewise.
32654 * tree-inline.c (estimate_move_cost): Likewise.
32655 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32656 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
32657 (get_address_cost_ainc): Likewise.
32658 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
32659 (vect_supportable_dr_alignment): Likewise.
32660 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
32661 (vectorizable_reduction): Likewise.
32662 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
32663 (vectorizable_operation, vectorizable_load): Likewise.
32664 * tree.c (build_same_sized_truth_vector_type): Likewise.
32665 * valtrack.c (cleanup_auto_inc_dec): Likewise.
32666 * var-tracking.c (emit_note_insn_var_location): Likewise.
32667 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
32668 (ADDR_VEC_ALIGN): Likewise.
32670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32671 Alan Hayward <alan.hayward@arm.com>
32672 David Sherwood <david.sherwood@arm.com>
32674 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
32676 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
32677 or if measurement_type is polynomial.
32678 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
32679 * combine.c (make_extraction): Likewise.
32680 * dse.c (find_shift_sequence): Likewise.
32681 * dwarf2out.c (mem_loc_descriptor): Likewise.
32682 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
32683 (extract_bit_field, extract_low_bits): Likewise.
32684 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
32685 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
32686 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
32687 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
32688 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
32689 * reload.c (find_reloads): Likewise.
32690 * reload1.c (alter_reg): Likewise.
32691 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
32692 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
32693 * tree-if-conv.c (predicate_mem_writes): Likewise.
32694 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
32695 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
32696 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
32697 * valtrack.c (dead_debug_insert_temp): Likewise.
32698 * varasm.c (mergeable_constant_section): Likewise.
32699 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
32701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32702 Alan Hayward <alan.hayward@arm.com>
32703 David Sherwood <david.sherwood@arm.com>
32705 * expr.c (expand_assignment): Cope with polynomial mode sizes
32706 when assigning to a CONCAT.
32708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32709 Alan Hayward <alan.hayward@arm.com>
32710 David Sherwood <david.sherwood@arm.com>
32712 * machmode.h (mode_precision): Change from unsigned short to
32714 (mode_to_precision): Return a poly_uint16 rather than an unsigned
32716 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
32717 or if measurement_type is not polynomial.
32718 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
32719 in which the mode is already known to be a scalar_int_mode.
32720 * genmodes.c (emit_mode_precision): Change the type of mode_precision
32721 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
32723 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
32724 for GET_MODE_PRECISION.
32725 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
32726 for GET_MODE_PRECISION.
32727 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
32729 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
32730 (expand_field_assignment, make_extraction): Likewise.
32731 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
32732 (get_last_value): Likewise.
32733 * convert.c (convert_to_integer_1): Likewise.
32734 * cse.c (cse_insn): Likewise.
32735 * expr.c (expand_expr_real_1): Likewise.
32736 * lra-constraints.c (simplify_operand_subreg): Likewise.
32737 * optabs-query.c (can_atomic_load_p): Likewise.
32738 * optabs.c (expand_atomic_load): Likewise.
32739 (expand_atomic_store): Likewise.
32740 * ree.c (combine_reaching_defs): Likewise.
32741 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
32742 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
32743 * tree.h (type_has_mode_precision_p): Likewise.
32744 * ubsan.c (instrument_si_overflow): Likewise.
32746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32747 Alan Hayward <alan.hayward@arm.com>
32748 David Sherwood <david.sherwood@arm.com>
32750 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
32751 polynomial numbers of units.
32752 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
32753 (valid_vector_subparts_p): New function.
32754 (build_vector_type): Remove temporary shim and take the number
32755 of units as a poly_uint64 rather than an int.
32756 (build_opaque_vector_type): Take the number of units as a
32757 poly_uint64 rather than an int.
32758 * tree.c (build_vector_from_ctor): Handle polynomial
32759 TYPE_VECTOR_SUBPARTS.
32760 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
32761 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
32762 (build_vector_from_val): If the number of units is variable,
32763 use build_vec_duplicate_cst for constant operands and
32764 VEC_DUPLICATE_EXPR otherwise.
32765 (make_vector_type): Remove temporary is_constant ().
32766 (build_vector_type, build_opaque_vector_type): Take the number of
32767 units as a poly_uint64 rather than an int.
32768 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
32770 * cfgexpand.c (expand_debug_expr): Likewise.
32771 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
32772 (store_constructor, expand_expr_real_1): Likewise.
32773 (const_scalar_mask_from_tree): Likewise.
32774 * fold-const-call.c (fold_const_reduction): Likewise.
32775 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
32776 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
32777 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
32778 (fold_relational_const): Likewise.
32779 (native_interpret_vector): Likewise. Change the size from an
32780 int to an unsigned int.
32781 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
32782 TYPE_VECTOR_SUBPARTS.
32783 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
32784 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
32785 duplicating a non-constant operand into a variable-length vector.
32786 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
32787 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
32788 * ipa-icf.c (sem_variable::equals): Likewise.
32789 * match.pd: Likewise.
32790 * omp-simd-clone.c (simd_clone_subparts): Likewise.
32791 * print-tree.c (print_node): Likewise.
32792 * stor-layout.c (layout_type): Likewise.
32793 * targhooks.c (default_builtin_vectorization_cost): Likewise.
32794 * tree-cfg.c (verify_gimple_comparison): Likewise.
32795 (verify_gimple_assign_binary): Likewise.
32796 (verify_gimple_assign_ternary): Likewise.
32797 (verify_gimple_assign_single): Likewise.
32798 * tree-pretty-print.c (dump_generic_node): Likewise.
32799 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32800 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
32801 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
32802 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
32803 (vect_shift_permute_load_chain): Likewise.
32804 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
32805 (expand_vector_condition, optimize_vector_constructor): Likewise.
32806 (lower_vec_perm, get_compute_type): Likewise.
32807 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
32808 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
32809 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
32810 (vect_recog_mask_conversion_pattern): Likewise.
32811 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
32812 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
32813 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32814 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
32815 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
32816 (vectorizable_shift, vectorizable_operation, vectorizable_store)
32817 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
32818 (supportable_widening_operation): Likewise.
32819 (supportable_narrowing_operation): Likewise.
32820 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
32822 * varasm.c (output_constant): Likewise.
32824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32825 Alan Hayward <alan.hayward@arm.com>
32826 David Sherwood <david.sherwood@arm.com>
32828 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
32829 so that both the length == 3 and length != 3 cases set up their
32830 own permute vectors. Add comments explaining why we know the
32831 number of elements is constant.
32832 (vect_permute_load_chain): Likewise.
32834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32835 Alan Hayward <alan.hayward@arm.com>
32836 David Sherwood <david.sherwood@arm.com>
32838 * machmode.h (mode_nunits): Change from unsigned char to
32840 (ONLY_FIXED_SIZE_MODES): New macro.
32841 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
32842 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
32843 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
32845 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
32846 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
32847 or if measurement_type is not polynomial.
32848 * genmodes.c (ZERO_COEFFS): New macro.
32849 (emit_mode_nunits_inline): Make mode_nunits_inline return a
32851 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
32852 Use ZERO_COEFFS when emitting initializers.
32853 * data-streamer.h (bp_pack_poly_value): New function.
32854 (bp_unpack_poly_value): Likewise.
32855 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
32856 for GET_MODE_NUNITS.
32857 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
32858 for GET_MODE_NUNITS.
32859 * tree.c (make_vector_type): Remove temporary shim and make
32860 the real function take the number of units as a poly_uint64
32861 rather than an int.
32862 (build_vector_type_for_mode): Handle polynomial nunits.
32863 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
32864 * emit-rtl.c (const_vec_series_p_1): Likewise.
32865 (gen_rtx_CONST_VECTOR): Likewise.
32866 * fold-const.c (test_vec_duplicate_folding): Likewise.
32867 * genrecog.c (validate_pattern): Likewise.
32868 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
32869 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
32870 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
32871 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
32872 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
32873 * rtlanal.c (subreg_get_info): Likewise.
32874 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32875 (vect_grouped_load_supported): Likewise.
32876 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
32877 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
32878 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
32879 (simplify_const_unary_operation, simplify_binary_operation_1)
32880 (simplify_const_binary_operation, simplify_ternary_operation)
32881 (test_vector_ops_duplicate, test_vector_ops): Likewise.
32882 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
32883 instead of CONST_VECTOR_NUNITS.
32884 * varasm.c (output_constant_pool_2): Likewise.
32885 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
32886 explicit-encoded elements in the XVEC for variable-length vectors.
32888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32890 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
32892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32893 Alan Hayward <alan.hayward@arm.com>
32894 David Sherwood <david.sherwood@arm.com>
32896 * coretypes.h (fixed_size_mode): Declare.
32897 (fixed_size_mode_pod): New typedef.
32898 * builtins.h (target_builtins::x_apply_args_mode)
32899 (target_builtins::x_apply_result_mode): Change type to
32900 fixed_size_mode_pod.
32901 * builtins.c (apply_args_size, apply_result_size, result_vector)
32902 (expand_builtin_apply_args_1, expand_builtin_apply)
32903 (expand_builtin_return): Update accordingly.
32905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32907 * cse.c (hash_rtx_cb): Hash only the encoded elements.
32908 * cselib.c (cselib_hash_rtx): Likewise.
32909 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
32910 CONST_VECTOR encoding.
32912 2018-01-03 Jakub Jelinek <jakub@redhat.com>
32913 Jeff Law <law@redhat.com>
32916 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
32917 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
32918 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
32919 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
32922 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
32923 explicitly probe *sp in a noreturn function if there were any callee
32924 register saves or frame pointer is needed.
32926 2018-01-03 Jakub Jelinek <jakub@redhat.com>
32929 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
32930 BLKmode for ternary, binary or unary expressions.
32933 * var-tracking.c (delete_vta_debug_insn): New inline function.
32934 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
32935 insns from get_insns () to NULL instead of each bb separately.
32936 Use delete_vta_debug_insn. No longer static.
32937 (vt_debug_insns_local, variable_tracking_main_1): Adjust
32938 delete_vta_debug_insns callers.
32939 * rtl.h (delete_vta_debug_insns): Declare.
32940 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
32941 instead of variable_tracking_main.
32943 2018-01-03 Martin Sebor <msebor@redhat.com>
32945 PR tree-optimization/83603
32946 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
32947 arguments past the endof the argument list in functions declared
32948 without a prototype.
32949 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
32950 Avoid checking when arguments are null.
32952 2018-01-03 Martin Sebor <msebor@redhat.com>
32955 * doc/extend.texi (attribute const): Fix a typo.
32956 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
32957 issuing -Wsuggest-attribute for void functions.
32959 2018-01-03 Martin Sebor <msebor@redhat.com>
32961 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
32962 offset_int::from instead of wide_int::to_shwi.
32963 (maybe_diag_overlap): Remove assertion.
32964 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
32965 * gimple-ssa-sprintf.c (format_directive): Same.
32966 (parse_directive): Same.
32967 (sprintf_dom_walker::compute_format_length): Same.
32968 (try_substitute_return_value): Same.
32970 2018-01-03 Jeff Law <law@redhat.com>
32972 PR middle-end/83654
32973 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
32974 non-constant residual for zero at runtime and avoid probing in
32975 that case. Reorganize code for trailing problem to mirror handling
32978 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
32980 PR tree-optimization/83501
32981 * tree-ssa-strlen.c (get_string_cst): New.
32982 (handle_char_store): Call get_string_cst.
32984 2018-01-03 Martin Liska <mliska@suse.cz>
32986 PR tree-optimization/83593
32987 * tree-ssa-strlen.c: Include tree-cfg.h.
32988 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
32989 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
32990 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
32992 (strlen_dom_walker::before_dom_children): Call
32993 gimple_purge_dead_eh_edges. Dump tranformation with details
32995 (strlen_dom_walker::before_dom_children): Update call by adding
32996 new argument cleanup_eh.
32997 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
32999 2018-01-03 Martin Liska <mliska@suse.cz>
33002 * cif-code.def (VARIADIC_THUNK): New enum value.
33003 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
33006 2018-01-03 Jan Beulich <jbeulich@suse.com>
33008 * sse.md (mov<mode>_internal): Tighten condition for when to use
33009 vmovdqu<ssescalarsize> for TI and OI modes.
33011 2018-01-03 Jakub Jelinek <jakub@redhat.com>
33013 Update copyright years.
33015 2018-01-03 Martin Liska <mliska@suse.cz>
33018 * ipa-visibility.c (function_and_variable_visibility): Skip
33019 functions with noipa attribure.
33021 2018-01-03 Jakub Jelinek <jakub@redhat.com>
33023 * gcc.c (process_command): Update copyright notice dates.
33024 * gcov-dump.c (print_version): Ditto.
33025 * gcov.c (print_version): Ditto.
33026 * gcov-tool.c (print_version): Ditto.
33027 * gengtype.c (create_file): Ditto.
33028 * doc/cpp.texi: Bump @copying's copyright year.
33029 * doc/cppinternals.texi: Ditto.
33030 * doc/gcc.texi: Ditto.
33031 * doc/gccint.texi: Ditto.
33032 * doc/gcov.texi: Ditto.
33033 * doc/install.texi: Ditto.
33034 * doc/invoke.texi: Ditto.
33036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33038 * vector-builder.h (vector_builder::m_full_nelts): Change from
33039 unsigned int to poly_uint64.
33040 (vector_builder::full_nelts): Update prototype accordingly.
33041 (vector_builder::new_vector): Likewise.
33042 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
33043 (vector_builder::operator ==): Likewise.
33044 (vector_builder::finalize): Likewise.
33045 * int-vector-builder.h (int_vector_builder::int_vector_builder):
33046 Take the number of elements as a poly_uint64 rather than an
33048 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
33049 from unsigned int to poly_uint64.
33050 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
33051 (vec_perm_indices::new_vector): Likewise.
33052 (vec_perm_indices::length): Likewise.
33053 (vec_perm_indices::nelts_per_input): Likewise.
33054 (vec_perm_indices::input_nelts): Likewise.
33055 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
33056 number of elements per input as a poly_uint64 rather than an
33057 unsigned int. Use the original encoding for variable-length
33058 vectors, rather than clamping each individual element.
33059 For the second and subsequent elements in each pattern,
33060 clamp the step and base before clamping their sum.
33061 (vec_perm_indices::series_p): Handle polynomial element counts.
33062 (vec_perm_indices::all_in_range_p): Likewise.
33063 (vec_perm_indices_to_tree): Likewise.
33064 (vec_perm_indices_to_rtx): Likewise.
33065 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
33066 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
33067 (tree_vector_builder::new_binary_operation): Handle polynomial
33068 element counts. Return false if we need to know the number
33069 of elements at compile time.
33070 * fold-const.c (fold_vec_perm): Punt if the number of elements
33071 isn't known at compile time.
33073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33075 * vec-perm-indices.h (vec_perm_builder): Change element type
33076 from HOST_WIDE_INT to poly_int64.
33077 (vec_perm_indices::element_type): Update accordingly.
33078 (vec_perm_indices::clamp): Handle polynomial element_types.
33079 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
33080 (vec_perm_indices::all_in_range_p): Likewise.
33081 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
33083 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
33084 polynomial vec_perm_indices element types.
33085 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
33086 * fold-const.c (fold_vec_perm): Likewise.
33087 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
33088 * tree-vect-generic.c (lower_vec_perm): Likewise.
33089 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
33090 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
33091 element type to HOST_WIDE_INT.
33093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33094 Alan Hayward <alan.hayward@arm.com>
33095 David Sherwood <david.sherwood@arm.com>
33097 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
33098 rather than an int. Use plus_constant.
33099 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
33100 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
33102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33103 Alan Hayward <alan.hayward@arm.com>
33104 David Sherwood <david.sherwood@arm.com>
33106 * calls.c (emit_call_1, expand_call): Change struct_value_size from
33107 a HOST_WIDE_INT to a poly_int64.
33109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33110 Alan Hayward <alan.hayward@arm.com>
33111 David Sherwood <david.sherwood@arm.com>
33113 * calls.c (load_register_parameters): Cope with polynomial
33114 mode sizes. Require a constant size for BLKmode parameters
33115 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
33116 forces a parameter to be padded at the lsb end in order to
33117 fill a complete number of words, require the parameter size
33118 to be ordered wrt UNITS_PER_WORD.
33120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33121 Alan Hayward <alan.hayward@arm.com>
33122 David Sherwood <david.sherwood@arm.com>
33124 * reload1.c (spill_stack_slot_width): Change element type
33125 from unsigned int to poly_uint64_pod.
33126 (alter_reg): Treat mode sizes as polynomial.
33128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33129 Alan Hayward <alan.hayward@arm.com>
33130 David Sherwood <david.sherwood@arm.com>
33132 * reload.c (complex_word_subreg_p): New function.
33133 (reload_inner_reg_of_subreg, push_reload): Use it.
33135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33136 Alan Hayward <alan.hayward@arm.com>
33137 David Sherwood <david.sherwood@arm.com>
33139 * lra-constraints.c (process_alt_operands): Reject matched
33140 operands whose sizes aren't ordered.
33141 (match_reload): Refer to this check here.
33143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33144 Alan Hayward <alan.hayward@arm.com>
33145 David Sherwood <david.sherwood@arm.com>
33147 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
33148 that the mode size is in the set {1, 2, 4, 8, 16}.
33150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33151 Alan Hayward <alan.hayward@arm.com>
33152 David Sherwood <david.sherwood@arm.com>
33154 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
33155 Use plus_constant instead of gen_rtx_PLUS.
33157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33158 Alan Hayward <alan.hayward@arm.com>
33159 David Sherwood <david.sherwood@arm.com>
33161 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
33162 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
33163 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
33164 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
33165 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
33166 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
33167 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
33168 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
33169 * config/i386/i386.c (ix86_push_rounding): ...this new function.
33170 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
33172 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
33173 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
33174 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
33175 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
33176 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
33177 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
33178 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
33179 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
33180 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
33181 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
33183 * expr.c (emit_move_resolve_push): Treat the input and result
33184 of PUSH_ROUNDING as a poly_int64.
33185 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
33186 (emit_push_insn): Likewise.
33187 * lra-eliminations.c (mark_not_eliminable): Likewise.
33188 * recog.c (push_operand): Likewise.
33189 * reload1.c (elimination_effects): Likewise.
33190 * rtlanal.c (nonzero_bits1): Likewise.
33191 * calls.c (store_one_arg): Likewise. Require the padding to be
33192 known at compile time.
33194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33195 Alan Hayward <alan.hayward@arm.com>
33196 David Sherwood <david.sherwood@arm.com>
33198 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
33199 Use plus_constant instead of gen_rtx_PLUS.
33201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33202 Alan Hayward <alan.hayward@arm.com>
33203 David Sherwood <david.sherwood@arm.com>
33205 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
33206 rather than an int.
33208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33209 Alan Hayward <alan.hayward@arm.com>
33210 David Sherwood <david.sherwood@arm.com>
33212 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
33213 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
33214 via stack temporaries. Treat the mode size as polynomial too.
33216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33217 Alan Hayward <alan.hayward@arm.com>
33218 David Sherwood <david.sherwood@arm.com>
33220 * expr.c (expand_expr_real_2): When handling conversions involving
33221 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
33222 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
33223 as a poly_uint64 too.
33225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33226 Alan Hayward <alan.hayward@arm.com>
33227 David Sherwood <david.sherwood@arm.com>
33229 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
33231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33232 Alan Hayward <alan.hayward@arm.com>
33233 David Sherwood <david.sherwood@arm.com>
33235 * combine.c (can_change_dest_mode): Handle polynomial
33236 REGMODE_NATURAL_SIZE.
33237 * expmed.c (store_bit_field_1): Likewise.
33238 * expr.c (store_constructor): Likewise.
33239 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
33240 and polynomial REGMODE_NATURAL_SIZE.
33241 (gen_lowpart_common): Likewise.
33242 * reginfo.c (record_subregs_of_mode): Likewise.
33243 * rtlanal.c (read_modify_subreg_p): Likewise.
33245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33246 Alan Hayward <alan.hayward@arm.com>
33247 David Sherwood <david.sherwood@arm.com>
33249 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
33250 numbers of elements.
33252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33253 Alan Hayward <alan.hayward@arm.com>
33254 David Sherwood <david.sherwood@arm.com>
33256 * match.pd: Cope with polynomial numbers of vector elements.
33258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33259 Alan Hayward <alan.hayward@arm.com>
33260 David Sherwood <david.sherwood@arm.com>
33262 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
33263 in a POINTER_PLUS_EXPR.
33265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33266 Alan Hayward <alan.hayward@arm.com>
33267 David Sherwood <david.sherwood@arm.com>
33269 * omp-simd-clone.c (simd_clone_subparts): New function.
33270 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
33271 (ipa_simd_modify_function_body): Likewise.
33273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33274 Alan Hayward <alan.hayward@arm.com>
33275 David Sherwood <david.sherwood@arm.com>
33277 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
33278 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
33279 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
33280 (expand_vector_condition, vector_element): Likewise.
33281 (subparts_gt): New function.
33282 (get_compute_type): Use subparts_gt.
33283 (count_type_subparts): Delete.
33284 (expand_vector_operations_1): Use subparts_gt instead of
33285 count_type_subparts.
33287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33288 Alan Hayward <alan.hayward@arm.com>
33289 David Sherwood <david.sherwood@arm.com>
33291 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
33292 (vect_compile_time_alias): ...this new function. Do the calculation
33293 on poly_ints rather than trees.
33294 (vect_prune_runtime_alias_test_list): Update call accordingly.
33296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33297 Alan Hayward <alan.hayward@arm.com>
33298 David Sherwood <david.sherwood@arm.com>
33300 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
33302 (vect_schedule_slp_instance): Likewise.
33304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33305 Alan Hayward <alan.hayward@arm.com>
33306 David Sherwood <david.sherwood@arm.com>
33308 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
33309 constant and extern definitions for variable-length vectors.
33310 (vect_get_constant_vectors): Note that the number of units
33311 is known to be constant.
33313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33314 Alan Hayward <alan.hayward@arm.com>
33315 David Sherwood <david.sherwood@arm.com>
33317 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
33318 of units as polynomial. Choose between WIDE and NARROW based
33321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33322 Alan Hayward <alan.hayward@arm.com>
33323 David Sherwood <david.sherwood@arm.com>
33325 * tree-vect-stmts.c (simd_clone_subparts): New function.
33326 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
33328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33329 Alan Hayward <alan.hayward@arm.com>
33330 David Sherwood <david.sherwood@arm.com>
33332 * tree-vect-stmts.c (vectorizable_call): Treat the number of
33333 vectors as polynomial. Use build_index_vector for
33334 IFN_GOMP_SIMD_LANE.
33336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33337 Alan Hayward <alan.hayward@arm.com>
33338 David Sherwood <david.sherwood@arm.com>
33340 * tree-vect-stmts.c (get_load_store_type): Treat the number of
33341 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
33342 for variable-length vectors.
33343 (vectorizable_mask_load_store): Treat the number of units as
33344 polynomial, asserting that it is constant if the condition has
33345 already been enforced.
33346 (vectorizable_store, vectorizable_load): Likewise.
33348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33349 Alan Hayward <alan.hayward@arm.com>
33350 David Sherwood <david.sherwood@arm.com>
33352 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
33353 of units as polynomial. Punt if we can't tell at compile time
33354 which vector contains the final result.
33356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33357 Alan Hayward <alan.hayward@arm.com>
33358 David Sherwood <david.sherwood@arm.com>
33360 * tree-vect-loop.c (vectorizable_induction): Treat the number
33361 of units as polynomial. Punt on SLP inductions. Use an integer
33362 VEC_SERIES_EXPR for variable-length integer reductions. Use a
33363 cast of such a series for variable-length floating-point
33366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33367 Alan Hayward <alan.hayward@arm.com>
33368 David Sherwood <david.sherwood@arm.com>
33370 * tree.h (build_index_vector): Declare.
33371 * tree.c (build_index_vector): New function.
33372 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
33373 of units as polynomial, forcibly converting it to a constant if
33374 vectorizable_reduction has already enforced the condition.
33375 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
33376 to create a {1,2,3,...} vector.
33377 (vectorizable_reduction): Treat the number of units as polynomial.
33378 Choose vectype_in based on the largest scalar element size rather
33379 than the smallest number of units. Enforce the restrictions
33382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33383 Alan Hayward <alan.hayward@arm.com>
33384 David Sherwood <david.sherwood@arm.com>
33386 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
33387 number of units as polynomial.
33389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33390 Alan Hayward <alan.hayward@arm.com>
33391 David Sherwood <david.sherwood@arm.com>
33393 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
33394 * target.def (autovectorize_vector_sizes): Return the vector sizes
33395 by pointer, using vector_sizes rather than a bitmask.
33396 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
33397 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
33398 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
33400 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
33401 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
33402 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
33403 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
33404 * omp-general.c (omp_max_vf): Likewise.
33405 * omp-low.c (omp_clause_aligned_alignment): Likewise.
33406 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
33407 * tree-vect-loop.c (vect_analyze_loop): Likewise.
33408 * tree-vect-slp.c (vect_slp_bb): Likewise.
33409 * doc/tm.texi: Regenerate.
33410 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
33412 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
33413 the vector size as a poly_uint64 rather than an unsigned int.
33414 (current_vector_size): Change from an unsigned int to a poly_uint64.
33415 (get_vectype_for_scalar_type): Update accordingly.
33416 * tree.h (build_truth_vector_type): Take the size and number of
33417 units as a poly_uint64 rather than an unsigned int.
33418 (build_vector_type): Add a temporary overload that takes
33419 the number of units as a poly_uint64 rather than an unsigned int.
33420 * tree.c (make_vector_type): Likewise.
33421 (build_truth_vector_type): Take the number of units as a poly_uint64
33422 rather than an unsigned int.
33424 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33425 Alan Hayward <alan.hayward@arm.com>
33426 David Sherwood <david.sherwood@arm.com>
33428 * target.def (get_mask_mode): Take the number of units and length
33429 as poly_uint64s rather than unsigned ints.
33430 * targhooks.h (default_get_mask_mode): Update accordingly.
33431 * targhooks.c (default_get_mask_mode): Likewise.
33432 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
33433 * doc/tm.texi: Regenerate.
33435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33436 Alan Hayward <alan.hayward@arm.com>
33437 David Sherwood <david.sherwood@arm.com>
33439 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
33440 * omp-general.c (omp_max_vf): Likewise.
33441 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
33442 (expand_omp_simd): Handle polynomial safelen.
33443 * omp-low.c (omplow_simd_context): Add a default constructor.
33444 (omplow_simd_context::max_vf): Change from int to poly_uint64.
33445 (lower_rec_simd_input_clauses): Update accordingly.
33446 (lower_rec_input_clauses): Likewise.
33448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33449 Alan Hayward <alan.hayward@arm.com>
33450 David Sherwood <david.sherwood@arm.com>
33452 * tree-vectorizer.h (vect_nunits_for_cost): New function.
33453 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
33454 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
33455 (vect_analyze_slp_cost): Likewise.
33456 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
33457 (vect_model_load_cost): Likewise.
33459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33460 Alan Hayward <alan.hayward@arm.com>
33461 David Sherwood <david.sherwood@arm.com>
33463 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
33464 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
33465 from an unsigned int * to a poly_uint64_pod *.
33466 (calculate_unrolling_factor): New function.
33467 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
33469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33470 Alan Hayward <alan.hayward@arm.com>
33471 David Sherwood <david.sherwood@arm.com>
33473 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
33474 from an unsigned int to a poly_uint64.
33475 (_loop_vec_info::slp_unrolling_factor): Likewise.
33476 (_loop_vec_info::vectorization_factor): Change from an int
33478 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
33479 (vect_get_num_vectors): New function.
33480 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
33481 (vect_get_num_copies): Use vect_get_num_vectors.
33482 (vect_analyze_data_ref_dependences): Change max_vf from an int *
33483 to an unsigned int *.
33484 (vect_analyze_data_refs): Change min_vf from an int * to a
33486 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
33487 than an unsigned HOST_WIDE_INT.
33488 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
33489 (vect_analyze_data_ref_dependence): Change max_vf from an int *
33490 to an unsigned int *.
33491 (vect_analyze_data_ref_dependences): Likewise.
33492 (vect_compute_data_ref_alignment): Handle polynomial vf.
33493 (vect_enhance_data_refs_alignment): Likewise.
33494 (vect_prune_runtime_alias_test_list): Likewise.
33495 (vect_shift_permute_load_chain): Likewise.
33496 (vect_supportable_dr_alignment): Likewise.
33497 (dependence_distance_ge_vf): Take the vectorization factor as a
33498 poly_uint64 rather than an unsigned HOST_WIDE_INT.
33499 (vect_analyze_data_refs): Change min_vf from an int * to a
33501 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
33502 vfm1 as a poly_uint64 rather than an int. Make the same change
33503 for the returned bound_scalar.
33504 (vect_gen_vector_loop_niters): Handle polynomial vf.
33505 (vect_do_peeling): Likewise. Update call to
33506 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
33507 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
33509 * tree-vect-loop.c (vect_determine_vectorization_factor)
33510 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
33511 (vect_get_known_peeling_cost): Likewise.
33512 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
33513 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
33514 (vect_transform_loop): Likewise. Use the lowest possible VF when
33515 updating the upper bounds of the loop.
33516 (vect_min_worthwhile_factor): Make static. Return an unsigned int
33517 rather than an int.
33518 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
33519 polynomial unroll factors.
33520 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
33521 (vect_make_slp_decision): Likewise.
33522 (vect_supported_load_permutation_p): Likewise, and polynomial
33524 (vect_analyze_slp_cost): Handle polynomial vf.
33525 (vect_slp_analyze_node_operations): Likewise.
33526 (vect_slp_analyze_bb_1): Likewise.
33527 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
33528 than an unsigned HOST_WIDE_INT.
33529 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
33530 (vectorizable_load): Handle polynomial vf.
33531 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
33533 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
33535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33536 Alan Hayward <alan.hayward@arm.com>
33537 David Sherwood <david.sherwood@arm.com>
33539 * match.pd: Handle bit operations involving three constants
33540 and try to fold one pair.
33542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
33544 * tree-vect-loop-manip.c: Include gimple-fold.h.
33545 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
33546 niters_maybe_zero parameters. Handle other cases besides a step of 1.
33547 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
33548 Add a path that uses a step of VF instead of 1, but disable it
33550 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
33551 and niters_no_overflow parameters. Update calls to
33552 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
33553 Create a new SSA name if the latter choses to use a ste other
33554 than zero, and return it via niters_vector_mult_vf_var.
33555 * tree-vect-loop.c (vect_transform_loop): Update calls to
33556 vect_do_peeling, vect_gen_vector_loop_niters and
33557 slpeel_make_loop_iterate_ntimes.
33558 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
33559 (vect_gen_vector_loop_niters): Update declarations after above changes.
33561 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
33563 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
33564 128-bit round to integer instructions.
33565 (ceil<mode>2): Likewise.
33566 (btrunc<mode>2): Likewise.
33567 (round<mode>2): Likewise.
33569 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
33571 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
33572 unaligned VSX load/store on P8/P9.
33573 (expand_block_clear): Allow the use of unaligned VSX
33574 load/store on P8/P9.
33576 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
33578 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
33580 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
33581 swap associated with both a load and a store.
33583 2018-01-02 Andrew Waterman <andrew@sifive.com>
33585 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
33586 * config/riscv/riscv.md (clear_cache): Use it.
33588 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
33590 * web.c: Remove out-of-date comment.
33592 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33594 * expr.c (fixup_args_size_notes): Check that any existing
33595 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
33596 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
33597 (emit_single_push_insn): ...here.
33599 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33601 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
33602 (const_vector_encoded_nelts): New function.
33603 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
33604 (const_vector_int_elt, const_vector_elt): Declare.
33605 * emit-rtl.c (const_vector_int_elt_1): New function.
33606 (const_vector_elt): Likewise.
33607 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
33608 of CONST_VECTOR_ELT.
33610 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33612 * expr.c: Include rtx-vector-builder.h.
33613 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
33614 directly on the tree encoding.
33615 (const_vector_from_tree): Likewise.
33616 * optabs.c: Include rtx-vector-builder.h.
33617 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
33618 sequence of "u" values.
33619 * vec-perm-indices.c: Include rtx-vector-builder.h.
33620 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
33621 directly on the vec_perm_indices encoding.
33623 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33625 * doc/rtl.texi (const_vector): Describe new encoding scheme.
33626 * Makefile.in (OBJS): Add rtx-vector-builder.o.
33627 * rtx-vector-builder.h: New file.
33628 * rtx-vector-builder.c: Likewise.
33629 * rtl.h (rtx_def::u2): Add a const_vector field.
33630 (CONST_VECTOR_NPATTERNS): New macro.
33631 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
33632 (CONST_VECTOR_DUPLICATE_P): Likewise.
33633 (CONST_VECTOR_STEPPED_P): Likewise.
33634 (CONST_VECTOR_ENCODED_ELT): Likewise.
33635 (const_vec_duplicate_p): Check for a duplicated vector encoding.
33636 (unwrap_const_vec_duplicate): Likewise.
33637 (const_vec_series_p): Check for a non-duplicated vector encoding.
33638 Say that the function only returns true for integer vectors.
33639 * emit-rtl.c: Include rtx-vector-builder.h.
33640 (gen_const_vec_duplicate_1): Delete.
33641 (gen_const_vector): Call gen_const_vec_duplicate instead of
33642 gen_const_vec_duplicate_1.
33643 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
33644 (gen_const_vec_duplicate): Use rtx_vector_builder.
33645 (gen_const_vec_series): Likewise.
33646 (gen_rtx_CONST_VECTOR): Likewise.
33647 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
33648 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
33649 Build a new vector rather than modifying a CONST_VECTOR in-place.
33650 (handle_special_swappables): Update call accordingly.
33651 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
33652 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
33653 Build a new vector rather than modifying a CONST_VECTOR in-place.
33654 (handle_special_swappables): Update call accordingly.
33656 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33658 * simplify-rtx.c (simplify_const_binary_operation): Use
33659 CONST_VECTOR_ELT instead of XVECEXP.
33661 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33663 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
33664 the selector elements to be different from the data elements
33665 if the selector is a VECTOR_CST.
33666 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
33667 ssizetype for the selector.
33669 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33671 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
33672 before testing each element individually.
33673 * tree-vect-generic.c (lower_vec_perm): Likewise.
33675 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33677 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
33678 * selftest-run-tests.c (selftest::run_tests): Call it.
33679 * vector-builder.h (vector_builder::operator ==): New function.
33680 (vector_builder::operator !=): Likewise.
33681 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
33682 (vec_perm_indices::all_from_input_p): New function.
33683 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
33684 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
33685 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
33686 instead of reading the VECTOR_CST directly. Detect whether both
33687 vector inputs are the same before constructing the vec_perm_indices,
33688 and update the number of inputs argument accordingly. Use the
33689 utility functions added above. Only construct sel2 if we need to.
33691 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33693 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
33694 the broadcast of the low byte.
33695 (expand_mult_highpart): Use an explicit encoding for the permutes.
33696 * optabs-query.c (can_mult_highpart_p): Likewise.
33697 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
33698 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
33699 (vectorizable_bswap): Likewise.
33700 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
33701 explicit encoding for the power-of-2 permutes.
33702 (vect_permute_store_chain): Likewise.
33703 (vect_grouped_load_supported): Likewise.
33704 (vect_permute_load_chain): Likewise.
33706 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33708 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
33709 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
33710 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
33711 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
33712 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
33713 (vect_gen_perm_mask_any): Likewise.
33715 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33717 * int-vector-builder.h: New file.
33718 * vec-perm-indices.h: Include int-vector-builder.h.
33719 (vec_perm_indices): Redefine as an int_vector_builder.
33720 (auto_vec_perm_indices): Delete.
33721 (vec_perm_builder): Redefine as a stand-alone class.
33722 (vec_perm_indices::vec_perm_indices): New function.
33723 (vec_perm_indices::clamp): Likewise.
33724 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
33725 (vec_perm_indices::new_vector): New function.
33726 (vec_perm_indices::new_expanded_vector): Update for new
33727 vec_perm_indices class.
33728 (vec_perm_indices::rotate_inputs): New function.
33729 (vec_perm_indices::all_in_range_p): Operate directly on the
33730 encoded form, without computing elided elements.
33731 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
33732 encoding. Update for new vec_perm_indices class.
33733 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
33734 the given vec_perm_builder.
33735 (expand_vec_perm_var): Update vec_perm_builder constructor.
33736 (expand_mult_highpart): Use vec_perm_builder instead of
33737 auto_vec_perm_indices.
33738 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
33739 vec_perm_indices instead of auto_vec_perm_indices. Use a single
33740 or double series encoding as appropriate.
33741 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
33742 vec_perm_indices instead of auto_vec_perm_indices.
33743 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
33744 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
33745 (vect_permute_store_chain): Likewise.
33746 (vect_grouped_load_supported): Likewise.
33747 (vect_permute_load_chain): Likewise.
33748 (vect_shift_permute_load_chain): Likewise.
33749 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
33750 (vect_transform_slp_perm_load): Likewise.
33751 (vect_schedule_slp_instance): Likewise.
33752 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
33753 (vectorizable_mask_load_store): Likewise.
33754 (vectorizable_bswap): Likewise.
33755 (vectorizable_store): Likewise.
33756 (vectorizable_load): Likewise.
33757 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
33758 vec_perm_indices instead of auto_vec_perm_indices. Use
33759 tree_to_vec_perm_builder to read the vector from a tree.
33760 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
33761 vec_perm_builder instead of a vec_perm_indices.
33762 (have_whole_vector_shift): Use vec_perm_builder and
33763 vec_perm_indices instead of auto_vec_perm_indices. Leave the
33764 truncation to calc_vec_perm_mask_for_shift.
33765 (vect_create_epilog_for_reduction): Likewise.
33766 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
33767 from auto_vec_perm_indices to vec_perm_indices.
33768 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
33769 instead of changing individual elements.
33770 (aarch64_vectorize_vec_perm_const): Use new_vector to install
33771 the vector in d.perm.
33772 * config/arm/arm.c (expand_vec_perm_d::perm): Change
33773 from auto_vec_perm_indices to vec_perm_indices.
33774 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
33775 instead of changing individual elements.
33776 (arm_vectorize_vec_perm_const): Use new_vector to install
33777 the vector in d.perm.
33778 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
33779 Update vec_perm_builder constructor.
33780 (rs6000_expand_interleave): Likewise.
33781 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
33782 (rs6000_expand_interleave): Likewise.
33784 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33786 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
33787 to qimode could truncate the indices.
33788 * optabs.c (expand_vec_perm_var): Likewise.
33790 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33792 * Makefile.in (OBJS): Add vec-perm-indices.o.
33793 * vec-perm-indices.h: New file.
33794 * vec-perm-indices.c: Likewise.
33795 * target.h (vec_perm_indices): Replace with a forward class
33797 (auto_vec_perm_indices): Move to vec-perm-indices.h.
33798 * optabs.h: Include vec-perm-indices.h.
33799 (expand_vec_perm): Delete.
33800 (selector_fits_mode_p, expand_vec_perm_var): Declare.
33801 (expand_vec_perm_const): Declare.
33802 * target.def (vec_perm_const_ok): Replace with...
33803 (vec_perm_const): ...this new hook.
33804 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
33805 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
33806 * doc/tm.texi: Regenerate.
33807 * optabs.def (vec_perm_const): Delete.
33808 * doc/md.texi (vec_perm_const): Likewise.
33809 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
33810 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
33811 expand_vec_perm for constant permutation vectors. Assert that
33812 the mode of variable permutation vectors is the integer equivalent
33813 of the mode that is being permuted.
33814 * optabs-query.h (selector_fits_mode_p): Declare.
33815 * optabs-query.c: Include vec-perm-indices.h.
33816 (selector_fits_mode_p): New function.
33817 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
33818 is defined, instead of checking whether the vec_perm_const_optab
33819 exists. Use targetm.vectorize.vec_perm_const instead of
33820 targetm.vectorize.vec_perm_const_ok. Check whether the indices
33821 fit in the vector mode before using a variable permute.
33822 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
33823 vec_perm_indices instead of an rtx.
33824 (expand_vec_perm): Replace with...
33825 (expand_vec_perm_const): ...this new function. Take the selector
33826 as a vec_perm_indices rather than an rtx. Also take the mode of
33827 the selector. Update call to shift_amt_for_vec_perm_mask.
33828 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
33829 Use vec_perm_indices::new_expanded_vector to expand the original
33830 selector into bytes. Check whether the indices fit in the vector
33831 mode before using a variable permute.
33832 (expand_vec_perm_var): Make global.
33833 (expand_mult_highpart): Use expand_vec_perm_const.
33834 * fold-const.c: Includes vec-perm-indices.h.
33835 * tree-ssa-forwprop.c: Likewise.
33836 * tree-vect-data-refs.c: Likewise.
33837 * tree-vect-generic.c: Likewise.
33838 * tree-vect-loop.c: Likewise.
33839 * tree-vect-slp.c: Likewise.
33840 * tree-vect-stmts.c: Likewise.
33841 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
33843 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
33844 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
33845 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
33846 (aarch64_vectorize_vec_perm_const): ...this new function.
33847 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
33848 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33849 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
33850 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
33851 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
33852 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33853 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
33855 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
33856 check for NEON modes.
33857 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
33858 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
33859 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
33860 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
33862 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
33863 the old VEC_PERM_CONST conditions.
33864 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
33865 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
33866 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
33867 (ia64_vectorize_vec_perm_const_ok): Merge into...
33868 (ia64_vectorize_vec_perm_const): ...this new function.
33869 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
33870 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
33871 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
33872 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
33873 * config/mips/mips.c (mips_expand_vec_perm_const)
33874 (mips_vectorize_vec_perm_const_ok): Merge into...
33875 (mips_vectorize_vec_perm_const): ...this new function.
33876 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
33877 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
33878 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
33879 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
33880 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
33881 (rs6000_expand_vec_perm_const): Delete.
33882 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
33884 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33885 (altivec_expand_vec_perm_const_le): Take each operand individually.
33886 Operate on constant selectors rather than rtxes.
33887 (altivec_expand_vec_perm_const): Likewise. Update call to
33888 altivec_expand_vec_perm_const_le.
33889 (rs6000_expand_vec_perm_const): Delete.
33890 (rs6000_vectorize_vec_perm_const_ok): Delete.
33891 (rs6000_vectorize_vec_perm_const): New function.
33892 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
33893 an element count and rtx array.
33894 (rs6000_expand_extract_even): Update call accordingly.
33895 (rs6000_expand_interleave): Likewise.
33896 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
33897 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
33898 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
33899 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
33900 (rs6000_expand_vec_perm_const): Delete.
33901 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
33902 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33903 (altivec_expand_vec_perm_const_le): Take each operand individually.
33904 Operate on constant selectors rather than rtxes.
33905 (altivec_expand_vec_perm_const): Likewise. Update call to
33906 altivec_expand_vec_perm_const_le.
33907 (rs6000_expand_vec_perm_const): Delete.
33908 (rs6000_vectorize_vec_perm_const_ok): Delete.
33909 (rs6000_vectorize_vec_perm_const): New function. Remove stray
33910 reference to the SPE evmerge intructions.
33911 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
33912 an element count and rtx array.
33913 (rs6000_expand_extract_even): Update call accordingly.
33914 (rs6000_expand_interleave): Likewise.
33915 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
33916 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
33918 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
33920 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33922 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
33923 vector mode and that that mode matches the mode of the data
33925 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
33926 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
33927 directly using expand_vec_perm_1 when forcing selectors into
33929 (expand_vec_perm_var): New function, split out from expand_vec_perm.
33931 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33933 * optabs-query.h (can_vec_perm_p): Delete.
33934 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
33935 * optabs-query.c (can_vec_perm_p): Split into...
33936 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
33937 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
33938 particular selector is valid.
33939 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
33940 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
33941 (vect_grouped_load_supported): Likewise.
33942 (vect_shift_permute_load_chain): Likewise.
33943 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
33944 (vect_transform_slp_perm_load): Likewise.
33945 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
33946 (vectorizable_bswap): Likewise.
33947 (vect_gen_perm_mask_checked): Likewise.
33948 * fold-const.c (fold_ternary_loc): Likewise. Don't take
33949 implementations of variable permutation vectors into account
33950 when deciding which selector to use.
33951 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
33952 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
33953 with a false third argument.
33954 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
33955 to test whether the constant selector is valid and can_vec_perm_var_p
33956 to test whether a variable selector is valid.
33958 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33960 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
33961 * optabs-query.c (can_vec_perm_p): Likewise.
33962 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
33963 instead of vec_perm_indices.
33964 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
33965 (vect_gen_perm_mask_checked): Likewise,
33966 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
33967 (vect_gen_perm_mask_checked): Likewise,
33969 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
33971 * optabs-query.h (qimode_for_vec_perm): Declare.
33972 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
33973 (qimode_for_vec_perm): ...this new function.
33974 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
33976 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
33978 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
33979 does not have a conditional at the top.
33981 2018-01-02 Richard Biener <rguenther@suse.de>
33983 * ipa-inline.c (big_speedup_p): Fix expression.
33985 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
33988 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
33991 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
33995 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
33996 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
33997 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
33998 cond_taken_branch_cost 3->4.
34000 2018-01-01 Jakub Jelinek <jakub@redhat.com>
34002 PR tree-optimization/83581
34003 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
34004 TODO_cleanup_cfg if any changes have been made.
34006 PR middle-end/83608
34007 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
34008 convert_modes if target mode has the right side, but different mode
34011 PR middle-end/83609
34012 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
34013 last argument when extracting from CONCAT. If either from_real or
34014 from_imag is NULL, use expansion through memory. If result is not
34015 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
34016 the parts directly to inner mode, if even that fails, use expansion
34019 PR middle-end/83623
34020 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
34021 check for bswap in mode rather than HImode and use that in expand_unop
34024 Copyright (C) 2018 Free Software Foundation, Inc.
34026 Copying and distribution of this file, with or without modification,
34027 are permitted in any medium without royalty provided the copyright
34028 notice and this notice are preserved.