1 2018-09-17 David Malcolm <dmalcolm@redhat.com>
3 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
4 format_string_diagnostic_t.
6 * substring-locations.c
7 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
8 (format_warning_n_va): Convert to...
9 (format_string_diagnostic_t::emit_warning_n_va): ...this.
10 (format_warning_va): Convert to...
11 (format_string_diagnostic_t::emit_warning_va): ...this.
12 (format_warning_at_substring): Convert to...
13 (format_string_diagnostic_t::emit_warning): ...this.
14 (format_warning_at_substring_n): Convert to...
15 (format_string_diagnostic_t::emit_warning_n): ...this.
16 * substring-locations.h (class format_string_diagnostic_t): New
18 (format_warning_va): Convert to
19 format_string_diagnostic_t::emit_warning_va.
20 (format_warning_n_va): Convert to
21 format_string_diagnostic_t::emit_warning_n_va.
22 (format_warning_at_substring): Convert to
23 format_string_diagnostic_t::emit_warning.
24 (format_warning_at_substring_n): Convert to
25 format_string_diagnostic_t::emit_warning_n.
27 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
28 Bernd Schmidt <bernds_cb1@t-online.de>
30 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
33 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
35 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
36 operand 0 predicate to nonimmediate operand.
37 (rint<mode>2_frndint): Remove insn pattern.
38 (rint<mode>2): Change operand 1 predicate to general_operand.
39 Extend operand 1 to XFmode and generate rintxf2 insn.
40 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
41 Do not use X87MODEF mode macro.
42 (frndintxf2_<rounding>_i387): Rename from
43 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
44 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
45 to XFmode and generate significandxf3 insn.
47 2018-09-17 Richard Biener <rguenther@suse.de>
49 PR tree-optimization/87328
50 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
51 visiting unexecutable backedges when not iterating.
52 (do_rpo_vn): Mark all edges not executable even when not
55 2018-09-17 Martin Jambor <mjambor@suse.cz>
58 * doc/invoke.texi (Warning Options): Likewise.
60 2018-09-17 Richard Biener <rguenther@suse.de>
62 PR tree-optimization/87301
63 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
64 clean EH info from leftover copy assignments.
66 2018-09-17 Martin Liska <mliska@suse.cz>
69 * gcov.c (output_intermediate_file): Fix out of bounds
72 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
74 * config/arc/arc.c: Object attributes for core4 not reflected
76 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
79 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
81 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
83 2018-09-17 Martin Liska <mliska@suse.cz>
85 * doc/gcov.texi: Document new option --use-hotness-colors.
86 * gcov.c (struct source_info): Declare new field.
87 (source_info::source_info): Set default for maximum_count.
88 (print_usage): Add new -q option.
89 (process_args): Process it.
90 (accumulate_line_info): Save src->maximum_count.
91 (output_line_beginning): Make color line number if
92 flag_use_hotness_colors is set.
93 (output_line_details): Pass default argument value.
94 (output_lines): Pass src->maximum_count.
96 2018-09-17 Martin Liska <mliska@suse.cz>
98 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
99 Use processor_names table.
100 * config/i386/i386.c (ix86_default_align): Use
101 processor_cost_table for alignment values.
102 (ix86_option_override_internal): Use processor_names.
103 (ix86_function_specific_print): Likewise.
104 * config/i386/i386.h (struct processor_costs):
105 Add alignment values.
106 (struct ptt): Remove and replace with const char *.
107 * config/i386/x86-tune-costs.h (struct processor_costs):
108 Declare default alignments for all costs.
110 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
112 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
113 symbolics or VR_VARYING ranges for ABS_EXPR.
114 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
115 when range will wrap.
117 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
120 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
121 before and after a JUMP_TABLE_DATA.
123 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
126 * dojump.c (do_compare_and_jump): Canonicalize function pointers
127 when one operand is a function pointer. Use POINTER_TYPE_P and
128 FUNC_OR_METHOD_TYPE_P.
129 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
130 * fold-const.c (build_range_check): Likewise.
131 * match.pd (simple_comparison): Likewise.
133 2018-09-14 David Malcolm <dmalcolm@redhat.com>
136 * spellcheck.c (get_edit_distance_cutoff): New function.
137 (selftest::test_edit_distance_unit_test_oneway): Rename to...
138 (selftest::test_get_edit_distance_one_way): ...this.
139 (selftest::test_get_edit_distance_unit): Rename to...
140 (selftest::test_get_edit_distance_both_ways): ...this.
141 (selftest::test_edit_distances): Move tests to this new function,
142 and test some more pairs of strings. Update for above renaming.
143 (selftest::get_old_cutoff): New function.
144 (selftest::test_get_edit_distance_cutoff): New function.
145 (selftest::assert_suggested_for): New function.
146 (ASSERT_SUGGESTED_FOR): New macro.
147 (selftest::assert_not_suggested_for): New function.
148 (ASSERT_NOT_SUGGESTED_FOR): New macro.
149 (selftest::test_suggestions): New function.
150 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
151 tests to selftest::test_edit_distances and call it. Add calls to
152 selftest::test_get_edit_distance_cutoff and
153 selftest::test_suggestions.
154 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
155 (best_match::consider): Replace hard-coded cutoff calculation with
157 (best_match::get_cutoff): New declaration.
158 (best_match::get_best_meaningful_candidate): Likewise.
160 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
162 * builtins.c (fold_builtin_strlen): Remove TODO comment.
164 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
167 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
169 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
170 terminated string literal.
172 2018-09-14 Martin Sebor <msebor@redhat.com>
174 * builtins.c (unterminated_array): Handle ARRAY_REF.
175 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
176 * builtins.h (unterminated_array): Declare extern.
177 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
179 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
182 2018-09-14 Martin Sebor <msebor@redhat.com>
183 Jeff Law <law@redhat.com>
185 * builtins.c (unterminated_array): New.
186 (expand_builtin_strcpy): Adjust.
187 (expand_builtin_strcpy_args): Detect unterminated arrays.
188 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
190 * gimple-fold.h (get_maxval_strlen): Add argument.
191 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
193 * gimple-fold.c (get_range_strlen): Add argument.
194 (get_maxval_strlen): Adjust.
195 * gimple-fold.h (get_range_strlen): Add argument.
197 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
199 * config/i386/movdirintrin.h: Fix copyright year.
201 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
203 * reg-stack.c: Include regs.h.
204 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
205 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
206 FIRST_STACK_REG, not DFmode.
207 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
208 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
209 (change stack): Default register mode to the reg_raw_mode of
210 FIRST_STACK_REG, not DFmode.
211 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
212 (*swapxf): Rename from swapxf.
214 2018-09-14 Carl Love <cel@us.ibm.com>
216 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
217 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
219 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
222 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
225 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
228 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
230 2018-09-14 Jason Merrill <jason@redhat.com>
232 Fix --enable-gather-detailed-mem-stats.
233 * hash-table.c (hash_table_usage): Change from variable to function.
234 * hash-table.h: Adjust.
235 * Makefile.in: Add missing dependencies on hash-table.h.
237 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
239 PR tree-optimization/87259
241 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
242 execute_cse_reciprocals_1 has tried transforming.
244 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
246 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
247 VR_VARYING for PLUS/MINUS_EXPR.
249 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
251 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
254 2018-09-14 Richard Biener <rguenther@suse.de>
257 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
258 bits for the merged partition.
260 2018-09-13 Martin Sebor <msebor@redhat.com>
261 Bernd Edlinger <bernd.edlinger@hotmail.de>
263 * builtins.h (c_srlen): Add argument.
264 * builtins.c (warn_string_no_nul): New function.
265 (c_strlen): Add argument and use it. Update recursive calls.
266 Pass DECL argument to string_constant to get info on non
267 terminated strings. Update *NONSTR as needed.
268 (fold_builtin_strlen): Add argument to calls to c_strlen.
269 Warn for unterminated arrays.
270 (warn_string_no_null): Add prototype.
271 * expr.c (string_constant): Update arguments. Update recursive
272 calls appropriately. Detect missing NUL terminator and outermost
273 declaration its missing in.
274 Improve checks for arrays with nonzero lower bound or elements
275 that are not a single byte. Simplify offset computation.
276 Simplify checks for non-NUL terminated strings.
277 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
278 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
280 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
282 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
284 * fold-const.c (c_getstr): Fix function comment. Remove unused third
285 argument. Fix range checks.
286 * fold-const.h (c_getstr): Adjust protoype.
287 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
288 string is constant but contains no NUL byte.
290 * expr.c (string_constant): Adjust function comment.
291 Remove bogus check for zero termination.
293 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
295 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
296 (get_constant_size): Don't make STRING_CSTs larger than they are.
297 (check_string_literal): New check function for STRING_CSTs.
298 (output_constant): Use it.
300 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
303 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
305 2018-09-13 Richard Biener <rguenther@suse.de>
307 PR tree-optimization/87263
308 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
309 (struct unwind_state): Add max_rpo field.
310 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
311 Compute max_rpo, the max RPO number a block can be backwards reached
312 from. Re-write non-iterating mode to a RPO ordered worklist approach,
313 separating it from the iterating mode.
315 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
317 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
318 (rfs_decision): New scheduling decision.
320 2018-09-13 Richard Biener <rguenther@suse.de>
323 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
324 (vn_nary_op_insert_pieces_predicated): Do not write useless
325 valid_dominated_by_p entry outside of the allocated storage.
327 2018-09-13 Omar Sandoval <osandov@osandov.com>
328 Tom de Vries <tdevries@suse.de>
331 * dwarf2out.c (is_c): New function.
332 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
334 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
337 * config/aarch64/aarch64.md (*aarch64_bfxil):
339 * config/aarch64/constraints.md (Ulc): Define.
340 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
342 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
345 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
347 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
348 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
349 aarch64_layout_frame call.
350 (aarch64_expand_epilogue): Likewise.
351 (aarch64_initial_elimination_offset): Likewise.
352 (aarch64_get_separate_components): Likewise.
353 (aarch64_use_return_insn_p): Likewise.
354 (aarch64_layout_frame): Remove unneeded check.
356 2018-09-13 Jakub Jelinek <jakub@redhat.com>
358 * configure.ac: Only append
359 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
360 gcc_config_arguments if it was never reconfigured or last reconfigure
361 was with different arguments.
362 * configure: Regenerated.
364 2018-09-13 Jakub Jelinek <jakub@redhat.com>
365 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
368 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
369 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
371 2018-09-13 Jakub Jelinek <jakub@redhat.com>
373 PR tree-optimization/87287
374 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
375 X % C == 0 to X % (unsigned) C == 0 optimization to ...
376 * match.pd (X % C == 0): ... here. New optimization.
378 2018-09-12 Jakub Jelinek <jakub@redhat.com>
381 * expr.h (maybe_optimize_mod_cmp): Declare.
382 * expr.c (mod_inv): New function.
383 (maybe_optimize_mod_cmp): New function.
384 (do_store_flag): Use it.
385 * cfgexpand.c (expand_gimple_cond): Likewise.
387 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
388 Julian Brown <julian@codesourcery.com>
391 * gimplify.c (gimplify_scan_omp_clauses): Set
392 target_firstprivatize_array_bases in OpenACC parallel and kernels
393 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
394 OpenACC data regions.
396 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
398 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
399 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
401 2018-09-12 Richard Biener <rguenther@suse.de>
403 PR tree-optimization/87280
404 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
405 edge but unreachable target.
406 (do_rpo_vn): For conservatively handling a PHI only mark
407 the backedge executable but not the block reachable.
409 2018-09-12 Richard Biener <rguenther@suse.de>
411 PR tree-optimization/87266
412 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
415 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
417 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
419 ("trunc<BFP:mode><DFP_ALL:mode>2")
420 ("trunc<DFP_ALL:mode><BFP:mode>2")
421 ("extend<BFP:mode><DFP_ALL:mode>2")
422 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
423 according to the target operand type.
425 2018-09-12 Jakub Jelinek <jakub@redhat.com>
426 Andreas Krebbel <krebbel@linux.ibm.com>
428 PR tree-optimization/86844
429 * gimple-ssa-store-merging.c
430 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
431 there are any overlapping stores in between them, make sure they are
432 also coalesced or we give up completely.
434 2018-09-12 Jakub Jelinek <jakub@redhat.com>
437 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
438 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
440 2018-09-12 Tom de Vries <tdevries@suse.de>
442 * common.opt (gdescribe-dies): Add option.
443 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
444 attribute for artifical and nameless decls.
445 (dwarf2out_register_external_die): Add description attribute to
446 external reference die.
447 (add_desc_attribute): New functions.
448 (gen_subprogram_die): Add description attribute to
449 DW_TAG_call_site_parameter.
450 * tree-pretty-print.c (print_generic_expr_to_str): New function.
451 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
452 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
454 (@item -gdescribe-dies): Add.
456 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
458 * tree-vrp.c (vrp_shift_undefined_p): Remove.
459 (extract_range_from_binary_expr_1: Call
460 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
461 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
464 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
466 * gimple-ssa-warn-alloca.c
467 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
468 field for ALLOCA_BOUND_*_LARGE.
470 2018-09-11 Nathan Sidwell <nathan@acm.org>
472 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
474 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
476 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
477 for clobbers. Remove obsolete comment.
479 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
481 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
482 mpxchk, mpxld and mpxst types.
483 (define_attr length_immediate): Remove all processing of mpx types.
484 (define_attr prefix_0f): Ditto.
485 (define_attr memory): Ditto.
487 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
489 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
490 (log<mode>2): Change operand 1 predicate to general_operand.
491 Extend operand 1 to XFmode and generate logxf3 insn.
492 (log10<mode>2): Change operand 1 predicate to general_operand.
493 Extend operand 1 to XFmode and generate log10xf3 insn.
494 (log2<mode>2): Change operand 1 predicate to general_operand.
495 Extend operand 1 to XFmode and generate log2xf3 insn.
496 (fyl2xp1_extend<mode>xf3_i387): Remove.
497 (log1p<mode>2): Change operand 1 predicate to general_operand.
498 Extend operand 1 to XFmode and generate log1pxf3 insn.
499 (fxtract_extend<mode>xf3_i387): Remove.
500 (logb<mode>2): Change operand 1 predicate to general_operand.
501 Extend operand 1 to XFmode and generate logbxf3 insn.
502 (ilogb<mode>2): Change operand 1 predicate to general_operand.
503 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
504 (significand<mode>2): Change operand 1 predicate to general_operand.
505 Extend operand 1 to XFmode and generate significandxf3 insn.
507 2018-09-11 Nathan Sidwell <nathan@acm.org>
509 * gcc.c (perror_with_name, pfatal_with_name): Delete.
510 (load_specs): Use fatal_error.
511 (DELETE_IF_ORDINARY, process_command): Use error.
512 (execute, run_attempt): Use fatal_error.
514 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
516 * diagnostic-core.h (sorry_at): New prototype.
517 * diagnostic.c (sorry_at): New function.
519 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
521 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
522 by zero as VR_UNDEFINED.
524 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
526 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
527 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
528 (<sincos>mode2): New expander.
529 (sincos_extend<mode>xf3_i387): Remove insn pattern.
530 (sincos -> sin, cos splitters): Remove splitter patterns.
531 (sincos<mode>3): Change operand 2 predicate to general_operand.
532 Extend operand 2 to XFmode and generate sincosxf3 insn.
533 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
534 Change operand 3 predicate to const1_operand.
535 (fptan_extend<mode>xf4_i387): Remove insn pattern.
536 (tanxf2): Update operands in the call to fptanxf4_i387.
537 (tan<mode>2): Change operand 1 predicate to general_operand.
538 Extend operand 1 to XFmode and generate tanxf3 insn.
539 (atan2xf3): Rename from *fpatanxf3_i387.
540 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
541 (atan2xf3): Remove expander.
542 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
543 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
544 (atan<mode>2): Change operand 1 predicate to general_operand.
545 Extend operand 1 to XFmode and generate atanxf3 insn.
547 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
549 * config/i386/i386.md (x87/SSE constant load splitter): Use
550 memory_operand instead of nonimmediate_operand for input operand
553 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
555 * config/i386/i386.md (float partial SSE register stall splitter): Move
556 splitter near its instruction pattern.
557 (float_extend partial SSE register stall splitter): Ditto.
558 (float_truncate partial SSE register stall splitter): Ditto.
560 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
563 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
564 to speculation_safe_value_not_needed.
567 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
568 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
569 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
570 leaf_function_p, instead use has_hard_reg_initial_val.
572 2018-09-09 Nathan Sidwell <nathan@acm.org>
574 * gcc.h (pfatal_with_name): Don't declare here.
575 * gcc.c (pfatal_with_name): Make static.
577 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
579 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
582 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
584 PR rtl-optimization/85458
585 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
586 priority hook to reduce the priority of EXPR.
588 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
590 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
591 DImode for x87 on 32bit targets. Conditionally disable x87 modes
592 with X87_ENABLE_FLOAT. Remove preparation code.
593 (*float<SWI48:mode><MODEF:mode>2): Rename from
594 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
595 math using "enabled" attribute.
596 (*floatdi<MODEF:mode>2_i387): Rename from
597 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
598 enable for 32bit targets only.
599 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
601 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
602 as operand 1 predicate. Rewrite as define_insn_and_split.
603 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
605 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
607 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
608 to fallthru to FLOAT case.
610 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
613 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
614 around folding of vec_sl to handle out of range shift values.
616 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
618 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
619 Update callers to gen_fix_trunc<mode>_i387_fisttp
620 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
621 nonimmediate_operand.
622 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
623 and corresponding splitters.
624 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
625 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
626 (fix_truncdi_i387_with_temp): Remove insn pattern
627 and corresponding splitters.
628 (fix_trunc<mode>_i387): Change operand 0 predicate to
629 nonimmediate_operand.
630 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
631 and corresponding splitters.
632 (*fistdi2_1): Remove.
634 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
635 (lrintxfdi2): Remove expander. Reimplement as define_insn.
636 (*fist<mode>2_1): Remove.
637 (fist<mode>2): Ditto.
638 (fist<mode>2_with_temp): Remove insn pattern and corresponding
640 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
641 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
642 (fistdi2_<rounding>): Change operand 0 predicate to
643 nonimmediate_operand.
644 (fistdi2_<rounding>_with_temp): Remove insn pattern
645 and corresponding splitters.
646 (fist<mode>2_<rounding>): Change operand 0 predicate to
647 nonimmediate_operand.
648 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
649 and corresponding splitters.
651 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
653 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
655 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
658 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
660 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
661 early gimple folding of vec_splat().
662 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
663 * gimple-fold.h: Add an extern define for tree_vec_extract().
665 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
667 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
668 wrappers around TREE_TYPE comparisons.
670 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
673 * config/s390/predicates.md: Add nonsym_memory_operand.
674 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
675 contains a SYMBOL_REF, load it into an intermediate pseudo.
676 (s390_emit_compare_and_swap): Legitimize operand.
677 * config/s390/s390.md: Use the new nonsym_memory_operand
678 with UNSPECV_CAS patterns.
680 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
683 * config/s390/s390-passes.def: New file.
684 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
686 (make_pass_s390_early_mach): Add declaration.
687 * config/s390/s390.c (make_pass_s390_early_mach):
688 (s390_option_override): Remove dynamic registration.
689 * config/s390/t-s390: Add s390-passes.def.
691 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
693 * config/s390/s390.c (s390_decompose_constant_pool_ref):
694 Remove UNSPEC_LTREL_BASE check.
695 (annotate_constant_pool_refs): Likewise.
696 (find_constant_pool_ref): Likewise.
697 (find_ltrel_base): Removed.
698 (replace_ltrel_base): Removed.
699 (s390_mainpool_finish): Remove replace_ltrel_base call.
700 (s390_chunkify_start): Remove pending LTREL_BASE logic.
701 (s390_chunkify_finish): Remove replace_ltrel_base call.
702 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
704 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
707 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
708 to speculation_safe_value_not_needed.
710 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
711 Bernd Schmidt <bernds_cb1@t-online.de>
713 * config/nvptx/nvptx-opts.h: New file.
714 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
715 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
717 (TARGET_SM35): New macro.
718 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
720 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
723 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
725 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
727 * config/i386/i386.md (truncdfsf2): Remove expander.
728 (truncdfsf2_with_temp): Ditto.
729 (truncxf<mode>2): Ditto.
730 (*truncdfsf_fast_mixed): Remove insn pattern.
731 (*truncdfsf_fast_i387): Ditto.
732 (*truncdfsf_mixed): Ditto.
733 (*truncdfsf_i387): Ditto.
734 (*truncdfsf2_i387_1): Ditto.
735 (*truncxfsf2_mixed): Ditto.
736 (*truncxfdf2_mixed): Ditto.
737 (*truncxf<mode>2_i387_noop): Ditto. Update callers
738 to call gen_truncxf<mode>2 instead.
739 (*truncxf<mode>2_i387): Remove.
740 (reg->reg splitters): Remove splitter pattern.
741 (reg->mem splitters): Ditto.
743 (truncdfsf2): New insn pattern.
744 (truncxf<mode>2): Ditto.
746 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
748 * tree-ssa-math-opts.c (is_mult_by): New function.
749 (is_square_of): Use the above.
750 (optimize_recip_sqrt): New function.
751 (pass_cse_reciprocals::execute): Use the above.
753 2018-09-05 Richard Biener <rguenther@suse.de>
756 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
757 to zero-init the emplaced vec.
759 2018-09-05 Martin Liska <mliska@suse.cz>
761 PR tree-optimization/87205
762 * tree-switch-conversion.c (pass_lower_switch::execute):
763 Group cases for switch statements.
765 2018-09-05 Richard Biener <rguenther@suse.de>
767 PR tree-optimization/87217
768 * tree-ssa-sccvn.c (vuse_valueize): New.
769 (vn_reference_lookup_pieces): Use it.
770 (vn_reference_lookup): Likewise.
772 2018-09-05 Nathan Sidwell <nathan@acm.org>
775 * stor-layout.c (place_field): Scan forwards to check last
776 bitfield when ms_bitfield_placement is in effect.
778 2018-09-05 Richard Biener <rguenther@suse.de>
781 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
784 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
785 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
787 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
788 * config.gcc (extra_objs): Build it.
789 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
791 * config/aarch64/aarch64-passes.def
792 (pass_tag_collision_avoidance): New pass.
793 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
794 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
795 (aarch64_classify_address): Remove static qualifier.
796 (aarch64_address_info, aarch64_address_type): Move to...
797 * config/aarch64/aarch64-protos.h: ... here.
798 (make_pass_tag_collision_avoidance): New function.
799 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
802 2018-09-05 Martin Liska <mliska@suse.cz>
804 * doc/gcov.texi: Update documentation of humar
806 * gcov.c (format_count): Print one decimal place, it provides
807 more fine number of situations like '1G' vs. '1.4G'.
809 2018-09-05 Martin Liska <mliska@suse.cz>
812 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
813 * optc-gen.awk: Allow 'Var' for Deprecated options in order
814 to generate a MASK value.
816 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
819 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
820 if frame pointer isn't used.
821 (compute_frame_pointer_to_fb_displacement): Likewise.
823 2018-09-04 Jakub Jelinek <jakub@redhat.com>
826 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
827 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
828 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
829 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
830 and OPTION_MASK_ISA_XSAVEC_UNSET.
832 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
834 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
835 XOR operations in NAND case.
837 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
839 * wide-int-range.cc (wide_int_range_convert): New.
840 * wide-int-range.h (wide_int_range_convert): New.
841 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
842 code into wide_int_range_convert.
843 (extract_range_into_wide_ints): Do not munge anti range constants
844 into the entire domain. Just return the range back.
846 2018-09-04 Martin Liska <mliska@suse.cz>
848 * genmatch.c (output_line_directive): Add new argument
850 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
852 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
854 * doc/invoke.texi (Option Summary): Add whitespace.
856 * doc/invoke.texi (Option Summary): Add -Waligned-new.
858 2018-09-04 Richard Biener <rguenther@suse.de>
860 PR tree-optimization/87211
861 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
862 backedge value we're supposed to treat as VARYING also number
863 the PHI to VARYING in case it got a different value-number already.
865 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
867 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
868 (extract_range_from_binary_expr_1): Do not call
869 vrp_can_optimize_bit_op.
870 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
872 (wide_int_range_get_mask_and_bounds): New.
873 (wide_int_range_optimize_bit_op): New.
874 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
875 (wide_int_range_bit_and): Same.
876 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
877 (wide_int_range_optimize_bit_op): New.
878 (wide_int_range_get_mask_and_bounds): New.
880 2018-09-04 Richard Biener <rguenther@suse.de>
882 PR tree-optimization/87176
883 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
884 variable. When value-numbering a virtual PHI node make sure
885 to not value-number to the backedge value.
887 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
889 * doc/extend.texi (Long Long, Hex Floats): Document support for
890 long long and hex floats in more recent versions of ISO C++.
892 2018-09-03 Richard Biener <rguenther@suse.de>
894 PR tree-optimization/87177
895 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
898 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
900 * bb-reorder.c (edge_order): Convert to C-qsort-style
901 tri-state comparator.
902 (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
904 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
906 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
907 tri-state comparator.
908 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
910 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
912 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
913 (mergesort): ... here as maximum count for using netsort.
914 (gcc_qsort): Set nlim to 3 if stable sort is requested.
915 (gcc_stablesort): New.
916 * system.h (gcc_stablesort): Declare.
918 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
920 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
921 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
922 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
924 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
926 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
927 lxsdx and stxsdx alternatives.
928 (*mov<mode>_hardfloat64): Ditto.
929 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
931 2018-09-03 Richard Biener <rguenther@suse.de>
933 PR tree-optimization/87200
934 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
937 2018-09-03 Martin Liska <mliska@suse.cz>
939 PR tree-optimization/87201
940 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
941 Fix parenthesis in an expression.
943 2018-09-03 Richard Biener <rguenther@suse.de>
945 PR tree-optimization/87197
946 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
947 visited. CSE the VN_INFO hashtable lookup.
949 PR tree-optimization/87169
950 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
951 iterating make sure there's no extra backedges from irreducible
952 regions feeding the header. Mark the destination block
955 2018-09-03 Martin Liska <mliska@suse.cz>
958 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
959 * common/common-targhooks.c (default_get_valid_option_values):
961 * common/common-targhooks.h (default_get_valid_option_values):
963 * common/config/i386/i386-common.c: Move processor_target_table
965 (ix86_get_valid_option_values): New function.
966 (TARGET_GET_VALID_OPTION_VALUES): New macro.
967 * config/i386/i386.c (struct ptt): Move to i386-common.c.
968 (PTA_*): Move all defined masks into i386-common.c.
969 (ix86_function_specific_restore): Use new processor_cost_table.
970 * config/i386/i386.h (struct ptt): Moved from i386.c.
971 (struct pta): Likewise.
972 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
973 * doc/tm.texi.in: Likewise.
974 * opt-suggestions.c (option_proposer::suggest_option):
975 Pass prefix to build_option_suggestions.
976 (option_proposer::get_completions): Likewise.
977 (option_proposer::build_option_suggestions): Use the new target
979 * opts.c (struct option_help_tuple): New struct.
980 (print_filtered_help): Use the new target hook.
982 2018-09-03 Martin Liska <mliska@suse.cz>
985 * predict.c (set_even_probabilities): Add likely_edges
986 argument and handle cases where we have precisely one
988 (combine_predictions_for_bb): Catch also likely_edges.
989 (tree_predict_by_opcode): Handle gswitch statements.
990 * tree-cfg.h (find_case_label_for_value): New declaration.
991 (find_taken_edge_switch_expr): Likewise.
992 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
993 Find pivot in decision tree based on probabily, not by number of
996 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
998 * doc/standards.texi (Standards): Update Objective-C reference.
1000 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1002 * doc/install.texi (Prerequisites): Update link for MPC.
1004 2018-09-01 Michael Matz <matz@suse.de>
1006 PR tree-optimization/87074
1007 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
1008 PHIs for outer-loop uses.
1010 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1012 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
1013 * doc/invoke.texi (C Dialect Options): Ditto.
1015 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
1017 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
1019 2018-08-31 Richard Biener <rguenther@suse.de>
1021 PR tree-optimization/87168
1022 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
1023 (rpo_elim::eliminate_avail): When OP was not visited it must
1026 2018-08-31 David Malcolm <dmalcolm@redhat.com>
1028 * tree-vrp.c (copy_value_range): Convert param "from" from
1029 "value_range *" to "const value_range *".
1030 (range_is_null): Likewise for param "vr".
1031 (range_int_cst_p): Likewise.
1032 (range_int_cst_singleton_p): Likewise.
1033 (symbolic_range_p): Likewise.
1034 (value_ranges_intersect_p): Likewise for both params.
1035 (value_range_nonnegative_p): Likewise for param "vr".
1036 (value_range_constant_singleton): Likewise.
1037 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
1038 (extract_range_into_wide_ints): Likewise for param "vr".
1039 (extract_range_from_multiplicative_op): Likewise for params "vr0"
1041 (vrp_can_optimize_bit_op): Likewise.
1042 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
1044 (extract_range_from_unary_expr): Likewise.
1045 (debug_value_range): Likewise for param "vr".
1046 (value_range::dump): Add "const" qualifier.
1047 (vrp_prop::check_array_ref): Convert local "vr" from
1048 "value_range *" to "const value_range *".
1049 (vrp_prop::check_mem_ref): Likewise.
1050 (vrp_prop::visit_stmt): Likewise for local "old_vr".
1051 (vrp_intersect_ranges_1): Likewise for param "vr_1".
1052 (vrp_intersect_ranges): Likewise.
1053 (simplify_stmt_for_jump_threading): Likewise for local "vr".
1054 (vrp_prop::vrp_finalize): Likewise.
1055 * tree-vrp.h (value_range::dump): Add "const" qualifier.
1056 (vrp_intersect_ranges): Add "const" qualifier to params as above.
1057 (extract_range_from_unary_expr): Likewise.
1058 (value_range_constant_singleton): Likewise.
1059 (symbolic_range_p): Likewise.
1060 (copy_value_range): Likewise.
1061 (extract_range_from_binary_expr_1): Likewise.
1062 (range_int_cst_p): Likewise.
1063 (vrp_set_zero_nonzero_bits): Likewise.
1064 (range_int_cst_singleton_p): Likewise.
1066 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
1068 * config/aarch64/arm_neon.h (vabsd_s64): New.
1069 (vnegd_s64): Likewise.
1071 2018-08-31 Martin Jambor <mjambor@suse.cz>
1073 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
1075 2018-08-31 Martin Liska <mliska@suse.cz>
1077 * ipa-icf.c (sem_item::add_type): Use
1078 sem_item::m_type_hash_cache.
1079 * ipa-icf.h: Move the cache from sem_item_optimizer
1082 2018-08-31 Nathan Sidwell <nathan@acm.org>
1084 * doc/extend.texi (Backwards Compatibility): Remove implicit
1085 extern C leeway of () being (...).
1087 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1089 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
1091 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
1095 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
1097 2018-08-31 Jakub Jelinek <jakub@redhat.com>
1100 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
1101 gen_int_mode. Formatting fixes.
1103 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
1105 * target.def (custom_function_descriptors): Improve documentation.
1106 * doc/tm.texi.in (Trampolines): Expand discussion of function
1107 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
1108 beginning of the section.
1109 * doc/tm.texi: Regenerated.
1111 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
1113 * cfg.h (class auto_edge_flag): Spell out the template-id of the
1114 base class in the initializer list. This is a workaround for
1115 building with older GCC.
1116 (class auto_bb_flag): Likewise.
1118 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
1120 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
1121 (altivec_vcmpequ<VI_char>_p): Remove star.
1122 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
1124 (expand_strncmp_vec_sequence): New function.
1125 (emit_final_str_compare_vec): New function.
1126 (expand_strn_compare): Add support for vector strncmp.
1127 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
1128 length specification to bytes.
1129 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
1130 (vcmpnezb_p): New pattern.
1131 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
1132 for option -mstring-compare-inline-limit.
1134 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
1136 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
1137 (PTA_SKYLAKE): Add PTA_AES.
1138 (PTA_GOLDMONT): Likewise.
1140 2018-08-29 Jan Hubicka <jh@suse.cz>
1143 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
1144 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
1146 2018-08-29 Jan Hubicka <jh@suse.cz>
1148 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
1150 (hash_tree): Do not visit TYPE_STUB_DECL.
1151 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
1152 stream TYPE_STUB_DECL.
1153 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
1154 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
1155 after free_lang_data.
1156 (type_in_anonymous_namespace_p): Likewise.
1158 2018-08-29 Jan Hubicka <jh@suse.cz>
1160 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1161 comment that it has to be even number.
1162 (class sreal): Change m_sig type to int32_t.
1163 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1164 int64_t for temporary calculations.
1165 (sreal_verify_basics): Drop one bit from minimum and maximum.
1167 2018-08-30 Richard Biener <rguenther@suse.de>
1169 PR tree-optimization/87147
1170 * tree-ssa-sccvn.c (SSA_VISITED): New function.
1171 (visit_phi): When the degenerate result is from the backedge and
1172 we didn't visit its definition yet drop to VARYING.
1173 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
1175 2018-08-29 Jan Hubicka <jh@suse.cz>
1177 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
1179 (hash_tree): Likewise.
1181 2018-08-29 Jan Hubicka <jh@suse.cz>
1183 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
1184 and TYPE_NEXT_REF_TO.
1186 2018-08-29 Jan Hubicka <jh@suse.cz>
1188 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1189 comment that it has to be even number.
1190 (class sreal): Change m_sig type to int32_t.
1191 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1192 int64_t for temporary calculations.
1193 (sreal_verify_basics): Drop one bit from minimum and maximum.
1195 2018-08-30 Tamar Christina <tamar.christina@arm.com>
1197 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
1199 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
1202 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
1203 if to_add is negative.
1205 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
1208 * builtins.c (c_strlen): Improve range checks.
1210 2018-08-29 Martin Sebor <msebor@redhat.com>
1211 Jeff Law <law@redhat.com>
1213 PR tree-optimization/86714
1214 PR tree-optimization/86711
1215 * builtins.c (c_strlen): Add arguments to call to string_constant.
1216 * expr.c (string_constant): Add argument. Detect missing nul
1217 terminator and outermost declaration it's missing in.
1218 * expr.h (string_constant): Add argument.
1219 * fold-const.c (read_from_constant_string): Add arguments to call to
1221 (c_getstr): Likewise.
1222 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1224 * tree-ssa-strlen.c (get_stridx): Likewise.
1226 2018-08-29 Jan Hubicka <jh@suse.cz>
1228 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
1229 Do not stream DECL_VINDEX.
1230 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
1231 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
1232 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
1234 2018-08-29 Richard Biener <rguenther@suse.de>
1236 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
1237 virtual operands that are not default defs to honor region
1239 (rpo_vn_valueize): Remove ineffective code here.
1241 2018-08-29 Richard Biener <rguenther@suse.de>
1243 PR tree-optimization/87132
1244 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
1245 when skipping defs reachable over backedges.
1247 2018-08-29 Richard Biener <rguenther@suse.de>
1249 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
1250 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
1251 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
1252 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
1253 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
1254 (vn_reference_lookup_3): Remove use of const_parms.
1255 (free_rpo_vn): Do not free const_parms.
1256 (do_rpo_vn): Do not call init_const_parms.
1257 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
1258 SSA_NAME_POINTS_TO_READONLY_MEMORY.
1259 (call_may_clobber_ref_p_1): Likewise.
1261 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
1264 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
1266 (-ftree-scev-cprop): Document.
1268 2018-08-29 Jan Hubicka <jh@suse.cz>
1270 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
1272 (sreal constructor): Update.
1273 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
1274 sreal:operator/): Update.
1276 2018-08-29 Martin Liska <mliska@suse.cz>
1278 * tree-switch-conversion.c (switch_conversion::expand):
1279 Strenghten assumption about gswitch statements.
1281 2018-08-29 Richard Biener <rguenther@suse.de>
1283 PR tree-optimization/87117
1284 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1285 re-value-number released SSA VDEFs.
1287 2018-08-29 Richard Biener <rguenther@suse.de>
1289 PR tree-optimization/87126
1290 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
1292 2018-08-28 Jim Wilson <jimw@sifive.com>
1294 * config/riscv/pic.md: Rewrite.
1295 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
1297 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
1298 (SOFTF, default_load, softload, softstore): New.
1300 2018-08-28 Jeff Law <law@redhat.com>
1302 * fold-const.c (fold_binary_loc): Remove recently added assert.
1304 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
1306 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
1307 to OP parmeter of generated function.
1309 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
1311 PR tree-optimization/87009
1312 * match.pd: Add boolean optimizations.
1314 2018-08-28 Martin Sebor <msebor@redhat.com>
1317 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
1318 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
1319 (pass_walloca::gate): Use it.
1320 (alloca_call_type): Same.
1321 (pass_walloca::execute): Same.
1322 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
1324 2018-08-28 David Malcolm <dmalcolm@redhat.com>
1326 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
1327 GCC_VERSION for usage of "__gcc_dump_printf__" format from
1330 2018-08-28 Richard Biener <rguenther@suse.de>
1332 PR tree-optimization/87124
1333 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
1334 constants before looking up avail.
1336 2018-08-28 Jakub Jelinek <jakub@redhat.com>
1339 * calls.c (maybe_warn_nonstring_arg): Punt early if
1340 warn_stringop_overflow is zero. Don't call get_range_strlen
1341 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
1342 Swap comparison operands to have constants on rhs. Only use
1343 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
1344 increment lenrng[0].
1346 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
1348 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
1349 use of tree_to_shwi. Remove duplicated test for the size being
1350 a whole number of bytes.
1352 2018-08-28 Richard Biener <rguenther@suse.de>
1354 PR tree-optimization/87117
1355 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
1356 Handle removed stmt without LHS (GIMPLE_NOP).
1358 2018-08-28 Richard Biener <rguenther@suse.de>
1360 PR tree-optimization/87117
1361 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
1362 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
1364 2018-08-28 Richard Biener <rguenther@suse.de>
1366 PR tree-optimization/87117
1367 * tree-ssa-pre.c (compute_avail): Do not make expressions
1368 with predicated values available.
1369 (get_expr_value_id): Assert we do not run into predicated value
1372 2018-08-28 Richard Biener <rguenther@suse.de>
1374 PR tree-optimization/87117
1375 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
1376 get virtual operands.
1377 (get_expr_operands): Handle STRING_CST like other decls.
1379 2018-08-28 Martin Liska <mliska@suse.cz>
1381 * tree.h: Update documentation of fndecl_built_in_p
1385 2018-08-27 Jeff Law <law@redhat.com>
1386 PR tree-optimization/87110
1387 * tree-ssa-dse.c (compute_trims): Handle non-constant
1390 2018-08-27 Martin Sebor <msebor@redhat.com>
1392 PR tree-optimization/86914
1393 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
1395 2018-08-27 Martin Sebor <msebor@redhat.com>
1397 PR tree-optimization/87112
1398 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
1399 the type of the bound argument.
1401 2018-08-27 Jeff Law <law@redhat.com>
1403 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
1404 type does not have a TYPE_SIZE_UNIT.
1406 2018-08-27 Steve Ellcey <sellcey@cavium.com>
1408 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
1409 with include of backend.h.
1411 2018-08-27 Richard Biener <rguenther@suse.de>
1413 PR tree-optimization/86927
1414 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
1415 use const cond reduction code.
1417 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
1419 PR tree-optimization/85758
1420 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
1422 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1425 * diagnostic-show-locus.c (class layout_range): Update for
1426 conversion of show_caret_p to a tri-state.
1427 (layout_range::layout_range): Likewise.
1428 (make_range): Likewise.
1429 (layout::maybe_add_location_range): Likewise.
1430 (layout::should_print_annotation_line_p): Don't show annotation
1431 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
1432 (layout::get_state_at_point): Update for conversion of
1433 show_caret_p to a tri-state. Bail out early for
1434 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
1435 underlining or source colorization.
1436 (gcc_rich_location::add_location_if_nearby): Update for conversion
1437 of show_caret_p to a tri-state.
1438 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
1439 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
1441 (selftest::test_one_liner_labels): Likewise.
1442 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
1443 conversion of show_caret_p to a tri-state.
1444 * pretty-print.c (text_info::set_location): Likewise.
1445 * pretty-print.h (text_info::set_location): Likewise.
1446 * substring-locations.c (format_warning_n_va): Likewise.
1447 * tree-diagnostic.c (default_tree_printer): Likewise.
1448 * tree-pretty-print.c (newline_and_indent): Likewise.
1450 2018-08-27 David Malcolm <dmalcolm@redhat.com>
1453 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
1454 line above for line-insertion fix-it hints.
1455 (selftest::test_fixit_insert_containing_newline): Update the
1456 expected results, and add a test with line-numbering enabled.
1458 2018-08-27 Martin Liska <mliska@suse.cz>
1461 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
1462 params with DECL_HAS_VALUE_EXPR_P.
1464 2018-08-27 Martin Liska <mliska@suse.cz>
1466 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
1467 selected expansion strategy.
1469 2018-08-27 Martin Liska <mliska@suse.cz>
1471 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
1472 * builtins.c (is_builtin_fn): Likewise.
1473 * attribs.c (diag_attr_exclusions): Use new function
1474 fndecl_built_in_p and remove check for FUNCTION_DECL if
1476 (builtin_mathfn_code): Likewise.
1477 (fold_builtin_expect): Likewise.
1478 (fold_call_expr): Likewise.
1479 (fold_builtin_call_array): Likewise.
1480 (fold_call_stmt): Likewise.
1481 (set_builtin_user_assembler_name): Likewise.
1482 (is_simple_builtin): Likewise.
1483 * calls.c (gimple_alloca_call_p): Likewise.
1484 (maybe_warn_nonstring_arg): Likewise.
1485 * cfgexpand.c (expand_call_stmt): Likewise.
1486 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
1487 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
1488 (cgraph_node::verify_node): Likewise.
1489 * cgraphclones.c (build_function_decl_skip_args): Likewise.
1490 (cgraph_node::create_clone): Likewise.
1491 * config/arm/arm.c (arm_insert_attributes): Likewise.
1492 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
1493 * dse.c (scan_insn): Likewise.
1494 * expr.c (expand_expr_real_1): Likewise.
1495 * fold-const.c (operand_equal_p): Likewise.
1496 (fold_binary_loc): Likewise.
1497 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
1498 * gimple-low.c (lower_stmt): Likewise.
1499 * gimple-pretty-print.c (dump_gimple_call): Likewise.
1500 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
1501 * gimple.c (gimple_build_call_from_tree): Likewise.
1502 (gimple_call_builtin_p): Likewise.
1503 (gimple_call_combined_fn): Likewise.
1504 * gimplify.c (gimplify_call_expr): Likewise.
1505 (gimple_boolify): Likewise.
1506 (gimplify_modify_expr): Likewise.
1507 (gimplify_addr_expr): Likewise.
1508 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
1509 * ipa-cp.c (determine_versionability): Likewise.
1510 * ipa-fnsummary.c (compute_fn_summary): Likewise.
1511 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
1512 * ipa-split.c (visit_bb): Likewise.
1513 (split_function): Likewise.
1514 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1515 * lto-cgraph.c (input_node): Likewise.
1516 * lto-streamer-out.c (write_symbol): Likewise.
1517 * omp-low.c (setjmp_or_longjmp_p): Likewise.
1518 (lower_omp_1): Likewise.
1519 * predict.c (strip_predict_hints): Likewise.
1520 * print-tree.c (print_node): Likewise.
1521 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
1522 * trans-mem.c (is_tm_irrevocable): Likewise.
1523 (is_tm_load): Likewise.
1524 (is_tm_simple_load): Likewise.
1525 (is_tm_store): Likewise.
1526 (is_tm_simple_store): Likewise.
1527 (is_tm_abort): Likewise.
1528 (tm_region_init_1): Likewise.
1529 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
1530 * tree-cfg.c (verify_gimple_call): Likewise.
1531 (move_stmt_r): Likewise.
1532 (stmt_can_terminate_bb_p): Likewise.
1533 * tree-eh.c (lower_eh_constructs_2): Likewise.
1534 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
1535 * tree-inline.c (remap_gimple_stmt): Likewise.
1536 (copy_bb): Likewise.
1537 (estimate_num_insns): Likewise.
1538 (fold_marked_statements): Likewise.
1539 * tree-sra.c (scan_function): Likewise.
1540 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
1541 (optimize_stack_restore): Likewise.
1542 (pass_fold_builtins::execute): Likewise.
1543 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
1544 (mark_all_reaching_defs_necessary_1): Likewise.
1545 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
1546 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1547 (pass_forwprop::execute): Likewise.
1548 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1549 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1550 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
1551 * tree-ssa-strlen.c (get_string_length): Likewise.
1552 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
1553 (find_func_aliases_for_call): Likewise.
1554 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
1555 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
1556 * tree-tailcall.c (find_tail_calls): Likewise.
1557 * tree.c (need_assembler_name_p): Likewise.
1558 (free_lang_data_in_decl): Likewise.
1559 (get_call_combined_fn): Likewise.
1560 * ubsan.c (is_ubsan_builtin_p): Likewise.
1561 * varasm.c (incorporeal_function_p): Likewise.
1562 * tree.h (DECL_BUILT_IN): Remove and replace with
1564 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
1565 (fndecl_built_in_p): New.
1567 2018-08-27 Martin Liska <mliska@suse.cz>
1569 PR tree-optimization/86847
1570 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
1571 Dump also subtree probability.
1572 (switch_decision_tree::do_jump_if_equal): New function.
1573 (switch_decision_tree::emit_case_nodes): Handle special
1574 situations in balanced tree that can be emitted much simpler.
1575 Fix calculation of probabilities that happen in tree expansion.
1576 * tree-switch-conversion.h (struct cluster): Add
1578 (struct simple_cluster): Likewise.
1579 (struct case_tree_node): Add new function has_child.
1580 (do_jump_if_equal): New.
1582 2018-08-27 Martin Liska <mliska@suse.cz>
1584 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
1585 Add new argument to bit_test_cluster constructor.
1586 (bit_test_cluster::emit): Set bits really number of values
1588 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
1589 probability argument.
1590 * tree-switch-conversion.h (struct bit_test_cluster):
1591 Add m_handles_entire_switch.
1593 2018-08-27 Martin Liska <mliska@suse.cz>
1595 PR tree-optimization/86702
1596 * tree-switch-conversion.c (jump_table_cluster::emit):
1597 Make probabilities even for values in jump table
1598 according to number of cases handled.
1599 (switch_decision_tree::compute_cases_per_edge): Pass
1600 argument to reset_out_edges_aux function.
1601 (switch_decision_tree::analyze_switch_statement): Likewise.
1602 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
1605 2018-08-27 Martin Liska <mliska@suse.cz>
1607 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
1608 cfun argument explicitly.
1609 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
1610 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
1611 function gimple_switch_default_bb.
1612 (convert_switch_statements):
1614 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
1615 * stmt.c (label_to_block_fn): Use label_to_block and pass
1616 cfun argument explicitly and use gimple_switch_label_bb.
1617 (expand_case): Likewise.
1618 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
1619 cfun argument explicitly. Likewise.
1620 (make_edges_bb): Likewise.
1621 (make_cond_expr_edges): Likewise.
1622 (get_cases_for_edge): Likewise.
1623 (make_gimple_switch_edges): Likewise.
1624 (label_to_block_fn): Likewise.
1625 (label_to_block): Likewise.
1626 (make_goto_expr_edges): Likewise.
1627 (make_gimple_asm_edges): Likewise.
1628 (main_block_label): Likewise.
1629 (group_case_labels_stmt): Likewise.
1630 (find_taken_edge_computed_goto): Likewise.
1631 (find_taken_edge_switch_expr): Likewise.
1632 (gimple_verify_flow_info): Likewise.
1633 (gimple_redirect_edge_and_branch): Likewise.
1634 (gimple_switch_label_bb): New function.
1635 (gimple_switch_default_bb): Likewise.
1636 (gimple_switch_edge): Likewise.
1637 (gimple_switch_default_edge): Likewise.
1638 * tree-cfg.h (label_to_block_fn): Remove and replace ...
1639 (label_to_block): ... with this.
1640 (gimple_switch_label_bb): New.
1641 (gimple_switch_default_bb): Likewise.
1642 (gimple_switch_edge): Likewise.
1643 (gimple_switch_default_edge): Likewise.
1644 * tree-cfgcleanup.c (convert_single_case_switch): Use
1645 new gimple functions and pass new argument to label_to_block.
1646 (cleanup_control_flow_bb):
1647 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
1648 cfun argument explicitly.
1649 (make_eh_edges): Likewise.
1650 (redirect_eh_dispatch_edge): Likewise.
1651 (lower_resx): Likewise.
1652 (lower_eh_dispatch): Likewise.
1653 (maybe_remove_unreachable_handlers): Likewise.
1654 (unsplit_eh): Likewise.
1655 (cleanup_empty_eh): Likewise.
1656 (verify_eh_edges): Likewise.
1657 (verify_eh_dispatch_edge): Likewise.
1658 * tree-ssa-dom.c (record_edge_info): Likewise.
1659 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
1660 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
1661 (thread_through_normal_block): Likewise.
1662 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
1663 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
1664 * tree-switch-conversion.c (switch_conversion::collect): Use new
1666 (switch_conversion::check_final_bb): Likewise.
1667 (switch_conversion::gather_default_values): Pass new argument
1669 (switch_conversion::build_constructors): Likewise.
1670 (switch_decision_tree::compute_cases_per_edge): Use new
1671 gimple_switch_edge function.
1672 (switch_decision_tree::analyze_switch_statement): Pass new argument
1674 (switch_decision_tree::try_switch_expansion): Use
1675 gimple_switch_default_edge.
1676 * tree-vrp.c (find_switch_asserts): Pass new argument
1678 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
1679 (vr_values::simplify_switch_using_ranges): Likewise.
1681 2018-08-27 Richard Biener <rguenther@suse.de>
1683 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
1684 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
1686 * tree-ssa-sccvn.h (struct vn_pval): New structure.
1687 (struct vn_nary_op_s): Add unwind_to member. Add
1688 predicated_values flag and put result into a union together
1689 with a linked list of vn_pval.
1690 (struct vn_ssa_aux): Add name member to make maintaining
1691 a map of SSA name to vn_ssa_aux possible. Remove no longer
1692 needed info, dfsnum, low, visited, on_sccstack, use_processed
1693 and range_info_anti_range_p members.
1694 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
1695 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
1697 (vn_valueize): New global.
1698 (vn_context_bb): Likewise.
1699 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
1700 VN_INFO_PTR_INFO): Remove.
1701 * tree-ssa-sccvn.c: ... (rewrite)
1702 (pass_fre::execute): For -O2+ initialize loops and run
1703 RPO VN in optimistic mode (iterating). For -O1 and -Og
1704 run RPO VN in non-optimistic mode.
1705 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
1706 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
1707 * doc/invoke.texi (sccvn-max-scc-size): Remove.
1708 (rpo-vn-max-loop-depth): Document.
1709 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
1710 when valuezing the VUSE signals we walked out of the region.
1711 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
1712 (phi_translate): Set VN context block to use for availability
1714 (compute_avail): Likewise.
1715 (pre_valueize): New function.
1716 (pass_pre::execute): Adjust to the RPO VN API.
1718 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
1719 (propagate_constants_for_unrolling): Remove.
1720 (tree_unroll_loops_completely): Perform value-numbering
1721 on the unrolled bodies loop parent.
1723 2018-08-27 Richard Biener <rguenther@suse.de>
1725 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
1726 for partial antic compute.
1728 2018-08-27 Jakub Jelinek <jakub@redhat.com>
1730 PR rtl-optimization/87065
1731 * combine.c (simplify_if_then_else): Formatting fix.
1732 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
1734 (known_cond): Don't return const_true_rtx for vector modes. Use
1735 CONST0_RTX instead of const0_rtx. Formatting fixes.
1737 2018-08-27 Martin Liska <mliska@suse.cz>
1739 PR gcov-profile/87069
1740 * gcov.c (process_file): Record files already processed
1741 and warn about a file being processed multiple times.
1743 2018-08-27 Martin Liska <mliska@suse.cz>
1746 * config/aarch64/aarch64.c (aarch64_override_options_internal):
1747 Set default values for x_aarch64_*_string strings.
1748 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
1749 prefix. For -mabi do not print '=ABI' in help and use
1750 <option_value> format for -msve-vector-bits and -moverride
1753 2018-08-26 Jeff Law <law@redhat.com>
1755 * config/mips/frame-header-opt.c: Include "backend.h" rather than
1758 2018-08-26 Marek Polacek <polacek@redhat.com>
1760 PR c++/87029, Implement -Wredundant-move.
1761 * doc/invoke.texi: Document -Wredundant-move.
1763 2018-08-25 Martin Sebor <msebor@redhat.com>
1765 PR tree-optimization/87059
1766 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
1767 to the same type as the other.
1768 * fold-const.c (fold_binary_loc): Assert expectation.
1770 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
1772 * config/darwin.c (machopic_legitimize_pic_address): Clean up
1773 extraneous parentheses, dead code section and formatting.
1775 2018-08-24 David Malcolm <dmalcolm@redhat.com>
1778 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
1779 wide enough for jumps in the line-numbering to be visible.
1780 (layout::print_gap_in_line_numbering): New member function.
1781 (layout::calculate_line_spans): When using line numbering, merge
1782 line spans that are only 1 line apart.
1783 (diagnostic_show_locus): When printing line numbers, show gaps in
1784 line numbering directly, rather than printing headers.
1785 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
1786 line-numbering with multiple line spans.
1787 (selftest::test_fixit_insert_containing_newline_2): Add test of
1788 line-numbering, in which the spans are close enough to be merged.
1790 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1792 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
1793 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
1794 * tree-vrp.c (range_is_nonnull): Remove.
1795 (range_includes_zero_p): Accept value_range instead of min/max.
1796 (extract_range_from_binary_expr_1): Do not early bail on
1798 Use range_includes_zero_p instead of range_is_nonnull.
1799 (extract_range_from_unary_expr): Use range_includes_zero_p instead
1800 of range_is_nonnull.
1801 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
1802 special case VR_ANTI_RANGE.
1803 (vrp_finalize): Same.
1804 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
1806 (range_is_nonnull): Remove.
1807 * vr-values.c (vrp_stmt_computes_nonzero): Use
1808 range_includes_zero_p instead of range_is_nonnull.
1809 (extract_range_basic): Pass value_range to range_includes_zero_p
1810 instead of range_is_nonnull.
1812 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
1814 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
1815 * emit-rtl.h (rtl_data): Remove return_bnd.
1816 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
1817 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
1818 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
1819 (POINTER_BOUNDS_MODE): Remove definition.
1820 (make_pointer_bounds_mode): Remove.
1821 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
1822 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
1823 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
1824 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
1825 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
1826 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
1827 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
1829 * config/i386/i386-modes.def (BND32, BND64): Remove.
1830 * config/i386/i386.c (dbx_register_map): Remove bound registers.
1831 (dbx64_register_map): Ditto.
1832 (svr4_dbx_register_map): Ditto.
1833 (indirect_thunk_bnd_needed): Remove.
1834 (indirect_thunks_bnd_used): Ditto.
1835 (indirect_return_bnd_needed): Ditto.
1836 (indirect_return_via_cx_bnd): Ditto.
1837 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
1838 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
1839 (output_indirect_thunk): Ditto. Remove need_prefix argument.
1840 (output_indirect_thunk_function): Remove handling of
1841 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
1842 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
1843 (ix86_save_reg): Remove handling of crtl->return_bnd.
1844 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
1845 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
1846 and UNSPEC_BNDLX_ADDR.
1847 (ix86_output_indirect_branch_via_reg): Remove handling of
1848 indirect_thunk_prefix_bnd.
1849 (ix86_output_indirect_branch_via_push): Ditto.
1850 (ix86_output_function_return): Ditto.
1851 (ix86_output_indirect_function_return): Ditto.
1852 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
1853 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
1854 (CALL_USED_REGISTERS): Ditto.
1855 (REG_ALLOC_ORDER): Update for removal of bound registers.
1856 (HI_REGISTER_NAMES): Ditto.
1857 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
1858 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
1859 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
1860 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
1861 (FIRST_PSEUDO_REG): Update.
1862 (BND): Remove mode iterator.
1863 * config/i386/predicates.md (bnd_mem_operator): Remove.
1865 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1867 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
1870 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
1872 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
1873 the case in which the permute needs only a single element and
1874 repeats for every vector of the result. Extend that case to
1875 handle variable-length vectors.
1876 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1878 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
1881 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
1882 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
1884 2018-08-24 Richard Biener <rguenther@suse.de>
1886 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
1887 bb_flags_allocated members.
1888 (auto_flag): New RAII class for allocating flags.
1889 (auto_edge_flag): New RAII class for allocating edge flags.
1890 (auto_bb_flag): New RAII class for allocating bb flags.
1891 * cfgloop.c (verify_loop_structure): Allocate temporary edge
1893 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
1894 in favor of temporarily allocated BB flag.
1895 * hsa-brig.c: Re-order includes.
1896 * hsa-dump.c: Likewise.
1897 * hsa-regalloc.c: Likewise.
1898 * print-rtl.c: Likewise.
1899 * profile-count.c: Likewise.
1901 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
1904 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
1907 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
1910 * wide-int-range.cc (wide_int_range_div): Do not ignore result
1911 from wide_int_range_multiplicative_op.
1913 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1915 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
1918 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
1920 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
1923 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
1925 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
1926 full GENERIC expression used for replacement.
1928 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
1930 * tree-vrp.c (abs_extent_range): Remove.
1931 (extract_range_into_wide_ints): Pass wide ints by reference.
1932 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
1933 Pass wide ints by reference in all calls to
1934 extract_range_into_wide_ints.
1935 * wide-int-range.cc (wide_int_range_div): New.
1936 * wide-int-range.h (wide_int_range_div): New.
1937 (wide_int_range_includes_zero_p): New.
1938 (wide_int_range_zero_p): New.
1940 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
1942 * config/aarch64/aarch64.md (arches): New enum.
1943 (arch): New enum attr.
1944 (arch_enabled): New attr.
1945 (enabled): Now uses arch_enabled only.
1946 (simd, sve, fp16): Removed attribute.
1947 (fp): Attr now defined in terms of 'arch'.
1948 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
1949 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
1950 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
1951 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
1952 attributes into 'arch'.
1953 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
1954 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
1955 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
1956 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
1957 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
1958 'simd' attribute into 'arch'.
1959 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
1960 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
1961 Convert use of 'fp' attribute to 'arch'.
1962 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
1963 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
1965 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
1966 (different modes) Merge 'fp' and 'simd' into 'arch'.
1967 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
1970 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
1972 PR rtl-optimization/87026
1973 * expmed.c (canonicalize_comparison): If we can no longer create
1974 pseudoregisters, don't.
1976 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
1979 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
1981 * config/arm/arm.c (speculation_barrier_libfunc): New static
1983 (arm_init_libfuncs): Initialize it.
1984 (arm_emit_speculation_barrier): New function.
1985 * config/arm/arm.md (speculation_barrier): Call
1986 arm_emit_speculation_barrier for architectures that do not have
1988 (speculation_barrier_insn): Only match on Armv7 or later.
1990 2018-08-23 Richard Biener <rguenther@suse.de>
1993 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
1996 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
1998 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
1999 of single-vector TBLs.
2000 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
2003 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
2006 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
2007 aarch64_evpc_tbl guard.
2009 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
2011 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
2014 2018-08-22 Martin Sebor <msebor@redhat.com>
2017 * tree-pretty-print.c (pretty_print_string): Add argument.
2018 (dump_generic_node): Call to pretty_print_string with string size.
2020 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
2022 PR rtl-optimization/86771
2023 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
2024 of two SETs into those two SETs, one to be placed at i2, if that SETs
2025 destination is modified between i2 and i3.
2027 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2029 PR tree-optimization/86725
2030 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
2032 (vect_analyze_scalar_cycles_1): Check it.
2034 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2036 PR tree-optimization/86725
2037 * tree-vect-loop.c (vect_is_simple_reduction): When treating
2038 an outer loop phi as a double reduction, make sure that the
2039 single user of the phi result is an inner loop phi.
2041 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2043 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
2044 grouped stores with gaps to a strided group.
2046 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
2048 * tree-vect-stmts.c (get_group_load_store_type)
2049 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
2050 first statement in a group.
2052 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2054 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
2055 the sequence used in gcc/gcc.c.
2057 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2060 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
2063 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2065 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
2066 Darwin10-specific unwinder-shim.
2067 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
2068 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
2069 New to cater for Darwin10 Rosetta.
2071 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2073 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
2074 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
2076 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
2081 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2082 (dwarf2out_switch_text_section): Generate a local label for the
2083 second function sub-section and apply it as the second FDE start
2085 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2086 second sub-section start.
2088 2018-08-22 Richard Biener <rguenther@suse.de>
2090 PR tree-optimization/86988
2091 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
2093 2018-08-22 Richard Biener <rguenther@suse.de>
2095 PR tree-optimization/86945
2096 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
2098 2018-08-22 Alexandre Oliva <oliva@adacore.com>
2100 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
2101 a comment about how uses of r2 for .sdata2 come about.
2103 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
2105 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
2107 2018-08-21 Marek Polacek <polacek@redhat.com>
2109 PR c++/86981, Implement -Wpessimizing-move.
2110 * doc/invoke.texi: Document -Wpessimizing-move.
2112 2018-08-21 Jan Hubicka <jh@suse.cz>
2114 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
2115 * tree.h (is_redundant_typedef): Remove.
2116 * dwarf2out.c (is_redundant_typedef): Turn into static function.
2118 2018-08-21 Jan Hubicka <jh@suse.cz>
2120 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
2123 2018-08-21 Tamar Christina <tamar.christina@arm.com>
2125 * expmed.c (extract_low_bits): Reject invalid subregs early.
2127 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2130 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
2133 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2135 * config/vxworks.h: Guard vxworks_asm_out_constructor and
2136 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
2137 * config/vxworks.c: Likewise.
2139 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2141 * config/vxworks.c: Set targetm.have_ctors_dtors
2142 if HAVE_INITFINI_ARRAY_SUPPORT.
2143 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
2144 if HAVE_INITFINI_ARRAY_SUPPORT.
2146 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2148 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
2149 default search path for VxWorks < 7.
2151 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
2153 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
2154 (get_format_string): Refer to c_getstr.
2156 2018-08-21 Tom de Vries <tdevries@suse.de>
2158 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2159 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2160 (debuginfo_early_stop): Declare.
2161 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2162 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2163 (debuginfo_early_stop): New function.
2164 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
2165 and debuginfo_early_stop.
2166 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
2167 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
2168 (general_init): Call debuginfo_early_init.
2169 (finalize): Call debuginfo_fini.
2170 (do_compile): Call debuginfo_init.
2171 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
2173 (@item -fdump-debug, @item -fdump-earlydebug): Add.
2175 2018-08-21 Tom de Vries <tdevries@suse.de>
2177 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
2178 flag_dump_noaddr and flag_dump_unnumbered.
2180 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
2182 * wide-int-range.cc (wide_int_range_abs): New.
2183 (wide_int_range_order_set): Rename from wide_int_range_min_max.
2184 * wide-int-range.h (wide_int_range_abs): New.
2185 (wide_int_range_min_max): New.
2186 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
2187 case to call wide_int_range_abs.
2188 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
2189 (extract_range_from_abs_expr): Delete.
2191 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
2194 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2195 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
2198 2018-08-20 Nathan Sidwell <nathan@acm.org>
2199 Jeff Law <law@redhat.com>
2201 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
2202 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
2204 2018-08-20 David Malcolm <dmalcolm@redhat.com>
2207 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
2208 (decl_attributes): Likewise.
2209 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
2211 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2212 * diagnostic-core.h (class auto_diagnostic_group): New class.
2213 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
2214 (diagnostic_report_diagnostic): Handle the first diagnostics within
2216 (emit_diagnostic): Add auto_diagnostic_group instance.
2218 (inform_n): Likewise.
2219 (warning): Likewise.
2220 (warning_at): Likewise.
2221 (warning_n): Likewise.
2222 (pedwarn): Likewise.
2223 (permerror): Likewise.
2225 (error_n): Likewise.
2226 (error_at): Likewise.
2228 (fatal_error): Likewise.
2229 (internal_error): Likewise.
2230 (internal_error_no_backtrace): Likewise.
2231 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
2232 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
2233 * diagnostic.h (struct diagnostic_context): Add fields
2234 "diagnostic_group_nesting_depth",
2235 "diagnostic_group_emission_count", "begin_group_cb",
2237 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
2238 Add auto_diagnostic_group instance(s).
2239 (find_explicit_erroneous_behavior): Likewise.
2240 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
2241 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
2242 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
2243 (gimplify_va_arg_expr): Likewise.
2244 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
2245 (HSA_SORRY_AT): Likewise.
2246 * ipa-devirt.c (compare_virtual_tables): Likewise.
2247 (warn_odr): Likewise.
2248 * multiple_target.c (expand_target_clones): Likewise.
2249 * opts-common.c (cmdline_handle_error): Likewise.
2250 * reginfo.c (globalize_reg): Likewise.
2251 * substring-locations.c (format_warning_n_va): Likewise.
2252 * tree-inline.c (expand_call_inline): Likewise.
2253 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
2254 * tree-ssa-loop-niter.c
2255 (do_warn_aggressive_loop_optimizations): Likewise.
2256 * tree-ssa-uninit.c (warn_uninit): Likewise.
2257 * tree.c (warn_deprecated_use): Likewise.
2259 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
2262 * config/i386/i386.md (eh_return): Always update EH return
2263 address in word_mode.
2265 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
2267 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
2268 TARGET_SPLIT_COMPLEX_ARG is defined.
2270 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2272 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
2274 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
2277 * expr.c (expand_assignment): Assert that bitpos is positive.
2278 (store_field): Likewise
2279 (expand_expr_real_1): Make sure that bitpos is positive.
2280 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
2283 2018-08-20 Nathan Sidwell <nathan@acm.org>
2285 * Makefile.in (CPP_ID_DATA_H): Delete.
2286 (CPP_INTERNAL_H): Don't add it.
2287 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
2288 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
2290 2018-08-20 Richard Biener <rguenther@suse.de>
2292 PR tree-optimization/78655
2293 * tree-vrp.c (extract_range_from_binary_expr_1): Make
2294 pointer + offset nonnull if either operand is nonnull work.
2296 2018-08-20 Tom de Vries <tdevries@suse.de>
2298 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
2299 unless the referenced die describes the added property using
2300 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
2301 Otherwise, add a DW_AT_location to the referenced die.
2303 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
2306 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
2307 register_operand when calling ix86_set_reg_reg_cost.
2308 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
2309 Set *total to 0 for operands that satisfy x86_64_immediate_operand
2310 predicate and to 1 otherwise.
2312 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
2314 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2315 emit a diagnostic that it is not supported and reset the option.
2316 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2317 supported and consume the option. (ASM_FINAL_SPEC): New.
2319 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2321 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
2324 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
2326 C-SKY port: Documentation
2328 * doc/extend.texi (C-SKY Function Attributes): New section.
2329 * doc/invoke.texi (Option Summary): Add C-SKY options.
2330 (C-SKY Options): New section.
2331 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
2333 2018-08-17 Jojo <jijie_rong@c-sky.com>
2334 Huibin Wang <huibin_wang@c-sky.com>
2335 Sandra Loosemore <sandra@codesourcery.com>
2336 Chung-Lin Tang <cltang@codesourcery.com>
2338 C-SKY port: Backend implementation
2340 * config/csky/*: New.
2341 * common/config/csky/*: New.
2343 2018-08-17 Jojo <jijie_rong@c-sky.com>
2344 Huibin Wang <huibin_wang@c-sky.com>
2345 Sandra Loosemore <sandra@codesourcery.com>
2346 Chung-Lin Tang <cltang@codesourcery.com>
2347 Andrew Jenner <andrew@codesourcery.com>
2349 C-SKY port: Configury
2351 * config.gcc (csky-*-*): New.
2352 * configure.ac: Add csky to targets for dwarf2 debug_line support.
2353 * configure: Regenerated.
2355 2018-08-17 David Malcolm <dmalcolm@redhat.com>
2357 * dump-context.h: Include "dumpfile.h".
2358 (dump_context::dump_printf_va): Convert final param from va_list
2359 to va_list *. Convert from ATTRIBUTE_PRINTF to
2360 ATTRIBUTE_GCC_DUMP_PRINTF.
2361 (dump_context::dump_printf_loc_va): Likewise.
2362 * dumpfile.c: Include "stringpool.h".
2363 (make_item_for_dump_printf_va): Delete.
2364 (make_item_for_dump_printf): Delete.
2365 (class dump_pretty_printer): New class.
2366 (dump_pretty_printer::dump_pretty_printer): New ctor.
2367 (dump_pretty_printer::emit_items): New member function.
2368 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
2370 (dump_pretty_printer::emit_item): New member function.
2371 (dump_pretty_printer::stash_item): New member function.
2372 (dump_pretty_printer::format_decoder_cb): New member function.
2373 (dump_pretty_printer::decode_format): New member function.
2374 (dump_context::dump_printf_va): Reimplement in terms of
2375 dump_pretty_printer.
2376 (dump_context::dump_printf_loc_va): Convert final param from va_list
2378 (dump_context::begin_scope): Reimplement call to
2379 make_item_for_dump_printf.
2380 (dump_printf): Update for change to dump_printf_va.
2381 (dump_printf_loc): Likewise.
2382 (selftest::test_capture_of_dump_calls): Convert "stmt" from
2383 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
2384 with %T, %E, and %G.
2385 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
2386 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
2387 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
2388 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
2389 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
2390 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
2391 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
2392 within a dump_printf_loc call to "%wu".
2393 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
2394 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
2395 missing space after "=".
2396 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
2397 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
2398 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
2399 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
2400 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
2401 duplicate "vectorized" from message.
2403 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
2405 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
2406 polyNxK_t element's TYPE_STRING_FLAG.
2408 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
2410 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
2411 (they were unnamed before). Fix comments.
2413 2018-08-17 Nathan Sidwell <nathan@acm.org>
2415 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
2417 2018-08-17 Richard Biener <rguenther@suse.de>
2419 PR tree-optimization/86841
2420 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
2422 2018-08-17 Martin Liska <mliska@suse.cz>
2424 * common.opt: Remove Warn, Init and Report for options with
2425 Ignore/Deprecated flag. Warning is done automatically for
2427 * config/i386/i386.opt: Likewise.
2428 * config/ia64/ia64.opt: Likewise.
2429 * config/rs6000/rs6000.opt: Likewise.
2430 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2431 Remove usage of flag_check_pointer_bounds.
2432 * lto-wrapper.c (merge_and_complain): Do not handle
2433 OPT_fcheck_pointer_bounds.
2434 (append_compiler_options): Likewise.
2435 * opt-functions.awk: Do not handle Deprecated.
2436 * optc-gen.awk: Check that Var, Report and Init are not
2437 used for an option with Ignore/Deprecated flag.
2438 * opts-common.c (decode_cmdline_option): Do not report
2440 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
2442 * opts.h (struct cl_option): Remove cl_deprecated flag.
2443 (CL_ERR_DEPRECATED): Remove error enum value.
2445 2018-08-17 Richard Biener <rguenther@suse.de>
2448 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2449 across a va-arg-pack using call adjust its return value accordingly.
2451 2018-08-16 Martin Sebor <msebor@redhat.com>
2453 PR tree-optimization/86853
2454 * gimple-ssa-sprintf.c (struct format_result): Rename member.
2455 (struct fmtresult): Add member and initialize it in ctors.
2456 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
2457 (format_string): Handle %S the same as %ls. Set MAYFAIL.
2458 (format_directive): Set POSUNDER4K when MAYFAIL is set.
2459 (parse_directive): Handle %C same as %c.
2460 (sprintf_dom_walker::compute_format_length): Adjust.
2461 (is_call_safe): Adjust.
2463 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
2465 * builtins.c (c_strlen): Add new parameter eltsize. Use it
2466 for determining how to count the elements.
2467 * builtins.h (c_strlen): Adjust prototype.
2468 * expr.c (string_constant): Add new parameter mem_size.
2469 Set *mem_size appropriately.
2470 * expr.h (string_constant): Adjust protoype.
2471 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
2472 * gimple-fold.h (get_range_strlen): Adjust prototype.
2473 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
2474 (format_string): Call get_string_length with eltsize.
2476 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2478 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
2479 to emit the span, rather than setting it as the prefix.
2481 2018-08-16 David Malcolm <dmalcolm@redhat.com>
2483 * diagnostic-show-locus.c (layout::start_annotation_line): Add
2484 "margin_char" parameter, defaulting to space. Use it in place
2485 of pp_space for the initial part of the margin.
2486 (layout::print_leading_fixits): Use '+' when filling the margin
2487 of line-insertion fix-it hints.
2489 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2491 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
2494 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2496 * config/rs6000/altivec.md: Don't set length attribute to the default
2498 * config/rs6000/darwin.md: Ditto.
2499 * config/rs6000/dfp.md: Ditto.
2500 * config/rs6000/htm.md: Ditto.
2501 * config/rs6000/rs6000.md: Ditto.
2502 * config/rs6000/sync.md: Ditto.
2503 * config/rs6000/vsx.md: Ditto.
2505 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2507 * config/rs6000/altivec.md: Don't set length attribute to the default
2508 value, for branch instructions.
2509 * config/rs6000/darwin.md: Ditto.
2510 * config/rs6000/rs6000.md: Ditto.
2512 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2514 * config/rs6000/rs6000.md (length): Always define as const_int 4.
2515 (unnamed conditional branch define_insn): Set length to 4 or 8
2516 depending on offset.
2517 (<bd>_<mode>): Similar, for alternative 0.
2518 (<bd>tf_<mode>): Ditto.
2520 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2522 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
2524 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
2526 * doc/rtl.texi: Replace old RTX class names with new names.
2529 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
2531 * expmed.h (canonicalize_comparison): New declaration.
2532 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
2533 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
2534 * optabs.c (prepare_cmp_insn): Likewise.
2535 * rtl.h (unsigned_condition_p): New function which checks if a
2536 comparison operator is unsigned.
2538 2018-08-16 Nathan Sidwell <nathan@acm.org>
2540 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
2541 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
2543 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2546 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
2547 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
2548 (mov<mov>): ..this and enable unconditionally.
2550 2018-08-16 Tamar Christina <tamar.christina@arm.com>
2552 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
2554 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
2556 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
2557 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
2558 "Common" with "Target".
2560 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2562 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
2563 * doc/invoke.texi (mmitigate-rop): Remove.
2564 * config/i386/i386.c: Do not include "regrename.h".
2565 (ix86_rop_should_change_byte_p, reg_encoded_number)
2566 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
2568 (ix86_reorg): Remove call to ix86_mitigate_rop.
2569 * config/i386/i386.md (attr "modrm_class"): Remove.
2570 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
2571 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
2572 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
2574 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
2576 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
2577 allow folding of mergeh() and mergel() for the float and double types.
2578 (fold_mergehl_helper): Rework to handle building a permute tree
2581 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2583 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
2586 2018-08-15 David Malcolm <dmalcolm@redhat.com>
2588 * common.opt (fdiagnostics-show-labels): New option.
2589 * diagnostic-show-locus.c (class layout_range): Add field
2591 (class layout): Add field "m_show_labels_p".
2592 (layout_range::layout_range): Add param "label" and use it to
2594 (make_range): Pass in NULL for new "label" param of layout_range's
2596 (layout::layout): Initialize m_show_labels_p.
2597 (layout::maybe_add_location_range): Pass in loc_range->m_label
2598 when constructing layout_range instances.
2599 (struct line_label): New struct.
2600 (layout::print_any_labels): New member function.
2601 (layout::print_line): Call it if label-printing is enabled.
2602 (selftest::test_one_liner_labels): New test.
2603 (selftest::test_diagnostic_show_locus_one_liner): Call it.
2604 * diagnostic.c (diagnostic_initialize): Initialize
2605 context->show_labels_p.
2606 * diagnostic.h (struct diagnostic_context): Add field
2608 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2609 -fno-diagnostics-show-labels.
2610 * dwarf2out.c (gen_producer_string): Add
2611 OPT_fdiagnostics_show_labels to the ignored options.
2612 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
2614 (gcc_rich_location::maybe_add_expr): Likewise.
2615 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
2616 label" param, defaulting to NULL.
2617 (gcc_rich_location::add_expr): Add "label" param.
2618 (gcc_rich_location::maybe_add_expr): Likewise.
2619 (class text_range_label): New class.
2620 (class range_label_for_type_mismatch): New class.
2621 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
2622 of format_warning_va.
2623 (fmtwarn_n): Likewise for new params of format_warning_n_va.
2624 * lto-wrapper.c (merge_and_complain): Add
2625 OPT_fdiagnostics_show_labels to the "pick one setting" options.
2626 (append_compiler_options): Likewise to the dropped options.
2627 (append_diag_options): Likewise to the passed-on options.
2628 * opts.c (common_handle_option): Handle the new option.
2629 * selftest-diagnostic.c
2630 (test_diagnostic_context::test_diagnostic_context): Enable
2632 * substring-locations.c: Include "gcc-rich-location.h".
2633 (format_warning_n_va): Add "fmt_label" and "param_label" params
2634 and use them as appropriate.
2635 (format_warning_va): Add "fmt_label" and "param_label" params,
2636 passing them on to format_warning_n_va.
2637 (format_warning_at_substring): Likewise.
2638 (format_warning_at_substring_n): Likewise.
2639 * substring-locations.h (format_warning_va): Add "fmt_label" and
2640 "param_label" params.
2641 (format_warning_n_va): Likewise.
2642 (format_warning_at_substring): Likewise.
2643 (format_warning_at_substring_n): Likewise.
2644 * toplev.c (general_init): Initialize global_dc->show_labels_p.
2646 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
2649 * builtins.c (expand_builtin_memcmp): Do not expand the call
2650 when overflow is detected.
2652 2018-08-15 Martin Sebor <msebor@redhat.com>
2654 PR tree-optimization/71625
2655 * config/aarch64/aarch64-builtins.c
2656 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
2658 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
2660 * config/s390/s390.c (s390_reorg): Remove loop.
2662 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2665 (darwin_function_switched_text_sections): Delete.
2666 * gcc/config/darwin.h
2667 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
2669 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
2672 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
2673 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
2674 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
2676 2018-08-15 Martin Liska <mliska@suse.cz>
2678 PR tree-optimization/86925
2679 * predict.c (expr_expected_value_1): When taking
2680 later predictor, assign also probability.
2681 Use fold_build2_initializer_loc in order to fold
2682 the expression in -frounding-math.
2684 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
2686 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
2688 (expand_vec_perm_1): Try the new method.
2690 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
2693 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
2694 Check whether lra_live_max_point is 0 before dividing.
2696 2018-08-14 Martin Sebor <msebor@redhat.com>
2698 PR tree-optimization/86650
2699 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
2700 (vrp_prop::check_mem_ref): Same.
2702 2018-08-13 Liu Hao <lh_mouse@126.com>
2704 * pretty-print.c (eat_esc_sequence): Swap the foreground and
2705 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
2706 and clear it thereafter, as it only works for DBCS.
2708 2018-08-13 Liu Hao <lh_mouse@126.com>
2710 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
2711 handle returned by _get_osf_handle().
2713 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2715 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
2716 for folding vec_perm.
2718 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2720 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
2721 Add support for gimple-folding of vec_pack() and vec_unpack()
2724 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
2726 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
2727 vec_xst variants to the list.
2728 (rs6000_gimple_fold_builtin): Add support for folding unaligned
2729 vector loads and stores.
2731 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
2733 * config.gcc (rs6000-ibm-aix4.x): Delete.
2734 (rs6000-ibm-aix5.1): Delete.
2735 (rs6000-ibm-aix5.2): Delete.
2736 (rs6000-ibm-aix5.3): Delete.
2737 * config/rs6000/aix43.h: Delete.
2738 * config/rs6000/aix51.h: Delete.
2739 * config/rs6000/aix52.h: Delete.
2740 * config/rs6000/t-aix43: Delete.
2742 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
2744 * config/s390/s390.c (s390_decompose_constant_pool_ref):
2746 (s390_decompose_address): Factor out constant pool ref
2749 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2751 * config/nds32/nds32-predicates.c
2752 (nds32_can_use_bclr_p): Change return type as bool.
2753 (nds32_can_use_bset_p): Ditto.
2754 (nds32_can_use_btgl_p): Ditto.
2755 (nds32_can_use_bitci_p): Ditto.
2756 * config/nds32/nds32-protos.h
2757 (nds32_can_use_bclr_p): Change declaration.
2758 (nds32_can_use_bset_p): Ditto.
2759 (nds32_can_use_btgl_p): Ditto.
2760 (nds32_can_use_bitci_p): Ditto.
2762 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2764 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
2765 Support -msched-prolog-epilog option.
2766 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
2768 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2770 * common/config/nds32/nds32-common.c
2771 (nds32_option_optimization_table): Enalbe -malways-align.
2773 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
2775 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
2777 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
2778 OPT_misr_secure_ case.
2779 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
2780 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
2781 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
2783 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
2784 (nds32_isr_info): New field security_level.
2785 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
2786 * config/nds32/nds32.md (return_internal): Consider critical attribute.
2787 * config/nds32/nds32.opt (misr-secure): New option.
2788 * config/nds32/nds32_init.inc: New file.
2789 * config/nds32/nds32_isr.h: New file.
2791 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
2793 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
2794 Update comment for atomic instructions.
2795 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
2796 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
2798 (atomic_loaddi): Revise fence expansion to only emit fence prior to
2799 load for __ATOMIC_SEQ_CST model.
2800 (atomic_loaddi_1): Remove float register target.
2801 (atomic_storedi): Handle CONST_INT values.
2802 (atomic_storedi_1): Remove float register source. Add special case
2804 (memory_barrier): New expander and insn.
2806 2018-08-11 Jakub Jelinek <jakub@redhat.com>
2808 PR tree-optimization/86835
2809 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
2810 new_stmt after def_gsi, make sure to insert new_square_stmt after
2811 that stmt, not 2 stmts before it.
2813 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
2816 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
2819 2018-08-10 Martin Liska <mliska@suse.cz>
2822 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
2824 * builtins.c (expand_builtin_expect_with_probability):
2826 (expand_builtin_expect_with_probability): New function.
2827 (build_builtin_expect_predicate): Add new argumnet probability
2828 for BUILT_IN_EXPECT_WITH_PROBABILITY.
2829 (fold_builtin_expect):
2832 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
2833 * builtins.h (fold_builtin_expect): Set new argument.
2834 * doc/extend.texi: Document __builtin_expect_with_probability.
2835 * doc/invoke.texi: Likewise.
2836 * gimple-fold.c (gimple_fold_call): Pass new argument.
2837 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
2838 also BUILT_IN_EXPECT_WITH_PROBABILITY.
2839 * predict.c (get_predictor_value): New function.
2840 (expr_expected_value): Add new argument probability. Assume
2841 that predictor and probability are always non-null.
2842 (expr_expected_value_1): Likewise. For __builtin_expect and
2843 __builtin_expect_with_probability set probability. Handle
2844 combination in binary expressions.
2845 (tree_predict_by_opcode): Simplify code by simply calling
2846 get_predictor_value.
2847 (pass_strip_predict_hints::execute): Add handling of
2848 BUILT_IN_EXPECT_WITH_PROBABILITY.
2849 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
2851 * tree.h (DECL_BUILT_IN_P): New function.
2853 2018-08-10 Martin Liska <mliska@suse.cz>
2855 PR tree-optimization/85799
2856 * passes.def: Add argument for pass_strip_predict_hints.
2857 * predict.c (class pass_strip_predict_hints): Add new argument
2859 (strip_predictor_early): New function.
2860 (pass_strip_predict_hints::execute): Call the function to
2862 (strip_predict_hints): New function.
2863 * predict.def: Fix comment.
2865 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
2867 * Makefile.in: Clarify which tm.texi to copy over to assert the
2868 right to grant a GFDL license for all.
2870 2018-08-09 Jeff Law <law@redhat.com>
2872 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
2875 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
2877 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
2880 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2882 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
2883 reductions for variable-length vectors.
2885 2018-08-09 David Malcolm <dmalcolm@redhat.com>
2888 * common.opt (fdiagnostics-show-line-numbers): New option.
2889 * diagnostic-show-locus.c (class layout): Add fields
2890 "m_show_line_numbers_p" and "m_linenum_width";
2891 (num_digits): New function.
2892 (test_num_digits): New function.
2893 (layout::layout): Initialize new fields. Update m_x_offset
2894 logic to handle any left margin.
2895 (layout::print_source_line): Print line number when requested.
2896 (layout::start_annotation_line): New member function.
2897 (layout::print_annotation_line): Call it.
2898 (layout::print_leading_fixits): Likewise.
2899 (layout::print_trailing_fixits): Likewise. Update calls to
2900 move_to_column for new parameter.
2901 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
2902 it to potentially call start_annotation_line.
2903 (layout::show_ruler): Call start_annotation_line.
2904 (selftest::test_line_numbers_multiline_range): New selftest.
2905 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
2906 and selftest::test_line_numbers_multiline_range.
2907 * diagnostic.c (diagnostic_initialize): Initialize
2908 show_line_numbers_p.
2909 * diagnostic.h (struct diagnostic_context): Add field
2910 "show_line_numbers_p".
2911 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2912 -fno-diagnostics-show-line-numbers.
2913 * dwarf2out.c (gen_producer_string): Add
2914 OPT_fdiagnostics_show_line_numbers to the ignored options.
2915 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
2916 one setting" options.
2917 (append_compiler_options): Likewise to the dropped options.
2918 (append_diag_options): Likewise to the passed-on options.
2919 * opts.c (common_handle_option): Handle the new option.
2920 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
2922 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
2924 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
2925 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
2926 third argument of type "const signed char" to descriptions of
2927 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
2928 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
2929 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
2930 __builtin_bcdsub_ov functions.
2932 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2934 PR tree-optimization/86858
2935 * tree-vect-loop.c (vect_is_simple_reduction): Restore
2936 flow_bb_inside_loop_p calls.
2938 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
2940 PR tree-optimization/86871
2941 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
2942 instead of gimple_assign_lhs.
2944 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
2947 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
2948 register constraint to operand 0.
2949 (add<mode>3_carryinC): Likewise.
2950 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
2952 2018-08-09 Martin Liska <mliska@suse.cz>
2955 * common.opt: Remove extra line.
2957 2018-08-09 Martin Liska <mliska@suse.cz>
2959 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
2960 at the end of a line, make first letter capital and end up
2961 a sentence with a dot.
2962 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
2963 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
2964 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
2965 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
2966 (PARAM_MAX_ISL_OPERATIONS): Likewise.
2967 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
2968 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
2969 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
2970 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
2971 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
2972 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
2973 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
2974 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
2975 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
2976 (PARAM_TREE_REASSOC_WIDTH): Likewise.
2977 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
2978 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
2979 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
2981 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
2984 * config/s390/s390.c (s390_option_override_internal): Reduce the
2985 stack-clash-protection-probe-interval param if it would be too big
2988 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
2991 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
2992 * config/m68k/m68k.c (handle_move_double): Don't call
2993 m68k_final_prescan_insn.
2994 (m68k_adjust_decorated_operand): Renamed from
2995 m68k_final_prescan_insn, remove first and third operand and
2997 (print_operand): Call it.
2998 (print_operand_address): Call it.
3000 2018-08-08 Nathan Sidwell <nathan@acm.org>
3002 * diagnostic.c (diagnostic_report_current_module): Use
3003 linemap_included_from & linemap_included_from_linemap.
3005 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
3007 * config/aarch64/aarch64-cores.def: Add phecda core.
3008 * config/aarch64/aarch64-tune.md: Regenerate.
3009 * doc/invoke.texi: Add phecda core.
3011 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3014 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
3016 * config/s390/s390.md ("movti"): Add more alternatives for
3017 constant to GPR copies.
3019 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
3021 * config/s390/s390.c: Fix whitespace damage throughout the file.
3022 * config/s390/s390.h: Likewise.
3023 * config/s390/tpf.h: Likewise.
3025 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3027 * config/s390/s390.c (s390_loadrelative_operand_p):
3028 Remove TARGET_CPU_ZARCH usages.
3029 (s390_rtx_costs): Likewise.
3030 (s390_legitimate_constant_p): Likewise.
3031 (s390_cannot_force_const_mem): Likewise.
3032 (legitimate_reload_constant_p): Likewise.
3033 (s390_preferred_reload_class): Likewise.
3034 (legitimize_pic_address): Likewise.
3035 (legitimize_tls_address): Likewise.
3036 (s390_split_branches): Removed.
3037 (s390_add_execute): Removed.
3038 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
3039 (s390_mainpool_start): Likewise.
3040 (s390_mainpool_finish): Likewise.
3041 (s390_mainpool_cancel): Removed.
3042 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
3043 (s390_chunkify_cancel): Likewise.
3044 (s390_return_addr_rtx): Likewise.
3045 (s390_register_info): Remove split_branches_pending_p uages.
3046 (s390_optimize_register_info): Likewise.
3047 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
3048 split_branches_pending_p usages.
3049 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
3050 (s390_load_got): Likewise.
3051 (s390_expand_split_stack_prologue): Likewise.
3052 (output_asm_nops): Likewise.
3053 (s390_function_profiler): Likewise.
3054 (s390_emit_call): Likewise.
3055 (s390_conditional_register_usage): Likewise.
3056 (s390_optimize_prologue): Likewise.
3057 (s390_reorg): Remove TARGET_CPU_ZARCH and
3058 split_branches_pending_p usages.
3059 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
3061 (s390_output_indirect_thunk_function): Likewise.
3062 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
3063 (TARGET_CPU_ZARCH_P): Removed.
3064 (struct machine_function): Remove split_branches_pending_p.
3065 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
3067 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
3069 * common/config/s390/s390-common.c (processor_flags_table):
3071 * config.gcc: Remove with_arch/with_tune support.
3072 * config/s390/2064.md: Remove cpu attribute comparisons.
3073 * config/s390/driver-native.c (s390_host_detect_local_cpu):
3075 * config/s390/linux.h (ASM_SPEC):
3076 Remove -march support.
3077 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
3078 Use a table to get an arch level.
3079 * config/s390/s390-opts.h (enum processor_type):
3081 * config/s390/s390.c
3082 (processor_table): Remove entries, add arch_level values.
3083 (s390_issue_rate): Remove cases.
3084 (s390_option_override): Adjust
3085 s390_option_override_internal() call.
3086 (s390_option_override_internal): Remove deprecation warning.
3087 (s390_valid_target_attribute_tree): Adjust
3088 s390_option_override_internal() call.
3089 * config/s390/s390.h (struct s390_processor):
3090 Share with s390-c.c, add arch_level field.
3091 * config/s390/s390.md:
3092 Remove occurrences in cpu attribute.
3093 * config/s390/s390.opt: Remove -march/-mtune support.
3094 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
3095 * doc/invoke.texi: Remove deprecation warning.
3097 2018-08-08 Luis Machado <luis.machado@linaro.org>
3099 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
3101 (qdf24xx_tunings): Set vector cost structure to
3102 qdf24xx_vector_cost.
3104 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
3105 <register_sextend>: Set to 3.
3107 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
3110 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
3111 * config/aarch64/aarch64-simd.md
3112 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
3113 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
3114 (aarch64_frecpx<mode>): ...this new pattern.
3115 * config/aarch64/aarch64-simd-builtins.def: Remove comment
3116 about aarch64_frecp<FRECP:frecp_suffix><mode>.
3118 2018-08-07 Martin Liska <mliska@suse.cz>
3121 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
3122 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
3123 * predict.def (PRED_MALLOC_NONNULL): New predictor.
3124 * doc/extend.texi: Document that malloc attribute adds
3127 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
3130 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3131 Define to speculation_safe_value_not_needed.
3133 2018-08-06 Jeff Law <law@redhat.com>
3135 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
3136 the vr_values instance to cprop_into_stmt.
3137 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
3138 (cprop_operand): Also query EVRP to determine if OP is a constant.
3140 2018-08-06 Nathan Sidwell <nathan@acm.org>
3142 * diagnostic.c (diagnostic_report_current_module): Reroll
3143 included-at loop. Translate text.
3145 2018-08-06 David Malcolm <dmalcolm@redhat.com>
3147 * function-tests.c (selftest::test_expansion_to_rtl): Call
3148 free_after_compilation.
3150 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3152 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
3154 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
3156 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
3157 loops with memory block operations from getting unrolled.
3159 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
3162 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3163 Define to speculation_safe_value_not_needed.
3165 2018-08-06 Jeff Law <law@redhat.com>
3167 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
3170 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3173 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
3174 with all enabled __intN types.
3176 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
3178 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3180 * alias.c (record_set): Check for clobber high.
3181 * cfgexpand.c (expand_gimple_stmt): Likewise.
3182 * combine-stack-adj.c (single_set_for_csa): Likewise.
3183 * combine.c (find_single_use_1): Likewise.
3184 (set_nonzero_bits_and_sign_copies): Likewise.
3185 (get_combine_src_dest): Likewise.
3186 (is_parallel_of_n_reg_sets): Likewise.
3187 (try_combine): Likewise.
3188 (record_dead_and_set_regs_1): Likewise.
3189 (reg_dead_at_p_1): Likewise.
3190 (reg_dead_at_p): Likewise.
3191 * dce.c (deletable_insn_p): Likewise.
3192 (mark_nonreg_stores_1): Likewise.
3193 (mark_nonreg_stores_2): Likewise.
3194 * df-scan.c (df_find_hard_reg_defs): Likewise.
3195 (df_uses_record): Likewise.
3196 (df_get_call_refs): Likewise.
3197 * dwarf2out.c (mem_loc_descriptor): Likewise.
3198 * haifa-sched.c (haifa_classify_rtx): Likewise.
3199 * ira-build.c (create_insn_allocnos): Likewise.
3200 * ira-costs.c (scan_one_insn): Likewise.
3201 * ira.c (equiv_init_movable_p): Likewise.
3202 (rtx_moveable_p): Likewise.
3203 (interesting_dest_for_shprep): Likewise.
3204 * jump.c (mark_jump_label_1): Likewise.
3205 * postreload-gcse.c (record_opr_changes): Likewise.
3206 * postreload.c (reload_cse_simplify): Likewise.
3207 (struct reg_use): Add source expr.
3208 (reload_combine): Check for clobber high.
3209 (reload_combine_note_use): Likewise.
3210 (reload_cse_move2add): Likewise.
3211 (move2add_note_store): Likewise.
3212 * print-rtl.c (print_pattern): Likewise.
3213 * recog.c (decode_asm_operands): Likewise.
3214 (store_data_bypass_p): Likewise.
3215 (if_test_bypass_p): Likewise.
3216 * regcprop.c (kill_clobbered_value): Likewise.
3217 (kill_set_value): Likewise.
3218 * reginfo.c (reg_scan_mark_refs): Likewise.
3219 * reload1.c (maybe_fix_stack_asms): Likewise.
3220 (eliminate_regs_1): Likewise.
3221 (elimination_effects): Likewise.
3222 (mark_not_eliminable): Likewise.
3223 (scan_paradoxical_subregs): Likewise.
3224 (forget_old_reloads_1): Likewise.
3225 * reorg.c (find_end_label): Likewise.
3226 (try_merge_delay_insns): Likewise.
3227 (redundant_insn): Likewise.
3228 (own_thread_p): Likewise.
3229 (fill_simple_delay_slots): Likewise.
3230 (fill_slots_from_thread): Likewise.
3231 (dbr_schedule): Likewise.
3232 * resource.c (update_live_status): Likewise.
3233 (mark_referenced_resources): Likewise.
3234 (mark_set_resources): Likewise.
3235 * rtl.c (copy_rtx): Likewise.
3236 * rtlanal.c (reg_referenced_p): Likewise.
3237 (single_set_2): Likewise.
3238 (noop_move_p): Likewise.
3239 (note_stores): Likewise.
3240 * sched-deps.c (sched_analyze_reg): Likewise.
3241 (sched_analyze_insn): Likewise.
3243 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3245 * cse.c (invalidate_reg): New function extracted from...
3246 (invalidate): ...here.
3247 (canonicalize_insn): Check for clobber high.
3248 (invalidate_from_clobbers): invalidate clobber highs.
3249 (invalidate_from_sets_and_clobbers): Likewise.
3250 (count_reg_usage): Check for clobber high.
3251 (insn_live_p): Likewise.
3252 * cselib.c (cselib_expand_value_rtx_1):Likewise.
3253 (cselib_invalidate_regno): Check for clobber in setter.
3254 (cselib_invalidate_rtx): Pass through setter.
3255 (cselib_invalidate_rtx_note_stores):
3256 (cselib_process_insn): Check for clobber high.
3257 * cselib.h (cselib_invalidate_rtx): Add operand.
3259 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3261 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
3262 (mark_not_eliminable): Likewise.
3263 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
3264 * lra-lives.c (process_bb_lives): Check for clobber high.
3265 * lra.c (new_insn_reg): Remember clobber highs.
3266 (collect_non_operand_hard_regs): Check for clobber high.
3267 (lra_set_insn_recog_data): Likewise.
3268 (add_regs_to_insn_regno_info): Likewise.
3269 (lra_update_insn_regno_info): Likewise.
3271 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3273 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
3274 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
3276 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3278 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
3279 (copy_insn_1): Likewise.
3280 (gen_hard_reg_clobber_high): New gen function.
3281 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
3282 * genemit.c (gen_exp): Likewise.
3283 (gen_emit_seq): Pass through info.
3284 (gen_insn): Check for CLOBBER_HIGH.
3285 (gen_expand): Pass through info.
3286 (gen_split): Likewise.
3287 (output_add_clobbers): Likewise.
3288 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
3289 (remove_clobbers): Likewise.
3290 * rtl.h (gen_hard_reg_clobber_high): New declaration.
3292 2018-08-06 Alan Hayward <alan.hayward@arm.com>
3294 * doc/rtl.texi (clobber_high): Add.
3295 (parallel): Add in clobber high
3296 * rtl.c (rtl_check_failed_code3): Add function.
3297 * rtl.def (CLOBBER_HIGH): Add expression.
3298 * rtl.h (RTL_CHECKC3): Add macro.
3299 (rtl_check_failed_code3): Add declaration.
3300 (XC3EXP): Add macro.
3302 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
3305 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3306 cfun->machine->max_used_stack_alignment if needed.
3308 2018-08-04 Martin Sebor <msebor@redhat.com>
3310 PR tree-optimization/86571
3311 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
3314 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
3317 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3320 2018-08-03 Jeff Law <law@redhat.com>
3323 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3324 Define to speculation_safe_value_not_needed.
3326 2018-08-03 David Malcolm <dmalcolm@redhat.com>
3328 * doc/gcov.texi (-x): Remove duplicate "to".
3329 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
3330 (-Wif-not-aligned): Remove duplicate "is".
3331 (-flto): Remove duplicate "the".
3332 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
3333 duplicate "v5.00.b".
3334 (MSP430 Options): Remove duplicate "and" from the description
3335 of "-mgprel-sec=regexp".
3336 (x86 Options): Remove duplicate copies of "vmldLog102" and
3337 vmlsLog104 from description of "-mveclibabi=type".
3339 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
3341 * internal-fn.h (first_commutative_argument): Declare.
3342 * internal-fn.c (first_commutative_argument): New function.
3343 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
3344 restrictions for pattern statements. Use first_commutative_argument
3345 to look for commutative operands in calls to internal functions.
3347 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
3349 * Makefile.in (wide-int-range.o): New.
3350 * tree-vrp.c: Move all the wide_int_* functions to...
3351 * wide-int-range.cc: ...here.
3352 * tree-vrp.h: Move all the wide_int_* prototypes to...
3353 * wide-int-range.h: ...here.
3355 2018-08-03 Tom de Vries <tdevries@suse.de>
3357 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
3359 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
3360 * except.c (output_function_exception_table): Do early exit if
3361 targetm_common.except_unwind_info (&global_options) == UI_NONE.
3363 2018-08-03 Martin Liska <mliska@suse.cz>
3365 * predict.c (dump_prediction): Change to 2 digits
3368 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
3370 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
3372 (falkor_am_1_gtov_gtov): ... a new insn reservation.
3374 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
3376 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
3377 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
3378 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
3379 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
3380 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
3382 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3384 * diagnostic-show-locus.c (diagnostic_show_locus): Use
3385 pp_take_prefix when saving the existing prefix.
3386 * diagnostic.c (diagnostic_append_note): Likewise.
3387 * langhooks.c (lhd_print_error_function): Likewise.
3388 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
3389 param's type. Free the existing prefix.
3390 (pp_take_prefix): New function.
3391 (pretty_printer::pretty_printer): Drop the prefix parameter.
3392 Rename the length parameter to match the comment.
3393 (pretty_printer::~pretty_printer): Free the prefix.
3394 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
3396 (struct pretty_printer): Drop the "const" from "prefix" field's
3397 type and clarify memory management.
3398 (pp_set_prefix): Drop the "const" from the 2nd param.
3399 (pp_take_prefix): New decl.
3401 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
3403 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
3404 for word_mode_ok here instead of passing as argument.
3405 (expand_block_compare): Change select_block_compare_mode() call.
3406 (expand_strncmp_gpr_sequence): New function.
3407 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
3409 2018-08-02 Jeff Law <law@redhat.com>
3412 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3413 Define to speculation_safe_value_not_needed.
3416 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3417 Define to speculation_safe_value_not_needed.
3419 2018-08-02 Tom de Vries <tdevries@suse.de>
3422 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
3423 function. Return UI_TARGET unconditionally.
3424 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
3425 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
3427 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3429 * genemit.c (print_overload_test): Fix typo.
3431 2018-08-02 Richard Biener <rguenther@suse.de>
3433 PR tree-optimization/86816
3434 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
3435 which checks for value availability before querying it.
3436 (gvn_uses_equal): Use it.
3437 (same_succ_hash): Likewise.
3438 (gimple_equal_p): Likewise.
3440 2018-08-02 Nick Clifton <nickc@redhat.com>
3443 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3444 Define to speculation_safe_value_not_needed.
3447 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3448 Define to speculation_safe_value_not_needed.
3451 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3452 Define to speculation_safe_value_not_needed.
3455 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3456 Define to speculation_safe_value_not_needed.
3459 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3460 Define to speculation_safe_value_not_needed.
3463 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3464 Define to speculation_safe_value_not_needed.
3467 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3468 Define to speculation_safe_value_not_needed.
3471 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3472 Define to speculation_safe_value_not_needed.
3475 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3476 speculation_safe_value_not_needed.
3479 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
3480 to speculation_safe_value_not_needed.
3482 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3484 * doc/md.texi: Expand the documentation of instruction names
3485 to mention port-local uses. Document '@' in pattern names.
3486 * read-md.h (overloaded_instance, overloaded_name): New structs.
3488 (md_reader::handle_overloaded_name): New member function.
3489 (md_reader::get_overloads): Likewise.
3490 (md_reader::m_first_overload): New member variable.
3491 (md_reader::m_next_overload_ptr): Likewise.
3492 (md_reader::m_overloads_htab): Likewise.
3493 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
3494 m_next_overload_ptr and m_overloads_htab.
3495 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
3496 (get_mode_token, get_code_token, get_int_token): New functions.
3497 (map_attr_string): Add an optional argument that passes back
3498 the associated iterator.
3499 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
3500 (md_reader::handle_overloaded_name, add_overload_instance): New
3502 (apply_iterators): Handle '@' names. Report an error if '@'
3503 is used without iterators.
3504 (initialize_iterators): Initialize the new iterator_group fields.
3505 * genopinit.c (handle_overloaded_code_for)
3506 (handle_overloaded_gen): New functions.
3507 (main): Use them to print declarations of maybe_code_for_* and
3508 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
3509 * genemit.c (print_overload_arguments, print_overload_test)
3510 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
3511 (main): Use it to print definitions of maybe_code_for_* and
3512 maybe_gen_* functions.
3513 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
3514 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
3515 instead of explicit mode checks.
3516 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
3517 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
3518 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
3519 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
3520 (aarch64_expand_compare_and_swap): Likewise
3521 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
3522 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
3523 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
3524 (aarch64_constant_pool_reload_icode): Delete.
3525 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
3526 instead of aarch64_constant_pool_reload_icode. Use
3527 code_for_aarch64_reload_mov instead of explicit mode checks.
3528 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
3529 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
3530 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
3531 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
3532 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
3533 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
3534 (aarch64_atomic_load_op_code): Delete.
3535 (aarch64_emit_atomic_load_op): Likewise.
3536 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
3537 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
3538 instead of aarch64_emit_atomic_load_op.
3539 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
3540 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
3541 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
3542 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
3543 character before the pattern name.
3544 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
3545 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
3546 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
3547 (aarch64_frecps<mode>): Likewise.
3548 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
3549 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
3550 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
3551 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
3552 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
3554 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
3556 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
3557 Allow HFmode constants if TARGET_FP_F16INST.
3559 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
3562 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
3563 No longer check last store for clobber of address register.
3565 2018-08-02 Martin Liska <mliska@suse.cz>
3567 PR gcov-profile/86817
3568 * gcov.c (process_all_functions): New function.
3570 (process_file): Move functions processing to
3571 process_all_functions.
3573 2018-08-02 David Malcolm <dmalcolm@redhat.com>
3575 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
3576 "const" to the "gimple *" and "rtx_insn *" parameters.
3577 * dumpfile.h (dump_user_location_t::dump_user_location_t):
3579 (dump_location_t::dump_location_t): Likewise.
3581 2018-08-01 Martin Sebor <msebor@redhat.com>
3583 PR tree-optimization/86650
3584 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
3585 rather than a "gcall *". Directly pass the data of interest
3586 to percent_K_format, rather than building a temporary CALL_EXPR
3588 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
3589 (gimple_fold_builtin_strncat): Adjust.
3590 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
3591 gcall* argument with gimple*.
3592 * gimple-ssa-warn-restrict.c (check_call): Same.
3593 (wrestrict_dom_walker::before_dom_children): Same.
3594 (builtin_access::builtin_access): Same.
3595 (check_bounds_or_overlap): Same
3596 (maybe_diag_overlap): Same.
3597 (maybe_diag_offset_bounds): Same.
3598 * tree-diagnostic.c (default_tree_printer): Move usage of
3599 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
3601 * tree-pretty-print.c (percent_K_format): Add argument.
3602 * tree-pretty-print.h: Add argument.
3603 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
3604 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
3605 (maybe_diag_stxncpy_trunc): Same.
3606 (handle_builtin_stxncpy): Same.
3607 (handle_builtin_strcat): Same.
3609 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3611 * match.pd: Optimise pointer range checks.
3613 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3615 PR tree-optimization/86758
3616 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
3617 to remove pattern statements.
3619 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3621 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
3622 result of dfs_enumerate_from when constructing stmt_vec_infos,
3623 instead of additionally calling get_loop_body.
3625 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3627 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
3629 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
3630 When creating an iv, assert that the step is not known to be zero.
3631 (vect_setup_realignment): Update call accordingly.
3632 * tree-vect-stmts.c (vectorizable_store): Likewise.
3633 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
3635 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3637 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
3638 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
3639 (vectorizable_reduction): Likewise.
3640 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3641 (vect_detect_hybrid_slp_stmts): Likewise.
3642 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3644 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
3646 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
3647 (wide_int_set_zero_nonzero_bits): ...this.
3648 (zero_nonzero_bits_from_vr): Rename to...
3649 (vrp_set_zero_nonzero_bits): ...this.
3650 (extract_range_from_multiplicative_op_1): Abstract wide int
3652 (wide_int_range_multiplicative_op): ...here.
3653 (extract_range_from_binary_expr_1): Extract wide int binary
3654 operations into their own functions.
3655 (wide_int_range_lshift): New.
3656 (wide_int_range_can_optimize_bit_op): New.
3657 (wide_int_range_shift_undefined_p): New.
3658 (wide_int_range_bit_xor): New.
3659 (wide_int_range_bit_ior): New.
3660 (wide_int_range_bit_and): New.
3661 (wide_int_range_trunc_mod): New.
3662 (extract_range_into_wide_ints): New.
3663 (vrp_shift_undefined_p): New.
3664 (extract_range_from_multiplicative_op): New.
3665 (vrp_can_optimize_bit_op): New.
3666 * tree-vrp.h (value_range::dump): New.
3667 (wide_int_range_multiplicative_op): New.
3668 (wide_int_range_lshift):New.
3669 (wide_int_range_shift_undefined_p): New.
3670 (wide_int_range_bit_xor): New.
3671 (wide_int_range_bit_ior): New.
3672 (wide_int_range_bit_and): New.
3673 (wide_int_range_trunc_mod): New.
3674 (zero_nonzero_bits_from_bounds): Rename to...
3675 (wide_int_set_zero_nonzero_bits): ...this.
3676 (zero_nonzero_bits_from_vr): Rename to...
3677 (vrp_set_zero_nonzero_bits): ...this.
3678 (range_easy_mask_min_max): Rename to...
3679 (wide_int_range_can_optimize_bit_op): this.
3680 * vr-values.c (simplify_bit_ops_using_ranges): Rename
3681 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
3683 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3685 * tree-vectorizer.h (vect_orig_stmt): New function.
3686 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
3687 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
3688 (vect_create_epilog_for_reduction): Likewise.
3689 (vectorizable_live_operation): Likewise.
3690 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
3691 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
3692 * tree-vect-stmts.c (vectorizable_call): Likewise.
3693 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
3695 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3697 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
3699 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
3700 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
3701 (vect_transform_loop): Likewise.
3702 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
3704 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3706 * tree-vectorizer.h (vect_schedule_slp): Return void.
3707 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
3708 (vect_schedule_slp): Likewise.
3710 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3712 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
3714 (vect_transform_loop): Update calls accordingly. Schedule SLP
3715 instances before the main loop, if any exist.
3717 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
3719 PR tree-optimization/86749
3720 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
3721 If the lhs is used in a COND_EXPR, check that it is being used
3722 as the "then" or "else" value.
3724 2018-08-01 Tom de Vries <tdevries@suse.de>
3727 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3728 speculation_safe_value_not_needed.
3730 2018-08-01 Richard Biener <rguenther@suse.de>
3732 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
3735 2018-08-01 Martin Liska <mliska@suse.cz>
3737 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
3738 format how successful transformation is dumped.
3739 (gimple_mod_pow2_value_transform): Likewise.
3740 (gimple_mod_subtract_transform): Likewise.
3741 (gimple_stringops_transform): Likewise.
3743 2018-08-01 Martin Liska <mliska@suse.cz>
3746 * value-prof.c (interesting_stringop_to_profile_p):
3747 Simplify the code and add BUILT_IN_MEMMOVE.
3748 (gimple_stringops_transform): Likewise.
3750 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
3752 * config/aarch64/aarch64-simd.md
3753 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
3754 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
3755 use GPI iterator instead of SI mode.
3757 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
3759 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
3760 rs6000_speculation_barrier.
3761 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
3762 new barrier pattern name.
3764 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
3766 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
3767 (speculation_barrier): New insn.
3769 2018-08-01 Richard Biener <rguenther@suse.de>
3772 * graphite.h: Include isl/id.h and isl/space.h to allow build
3775 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
3778 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3779 mode for COFF targets.
3780 * defaults.h (TARGET_COFF): Define.
3781 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
3782 TARGET_COFF): Define.
3783 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
3784 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
3785 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
3787 2018-07-31 Alexandre Oliva <oliva@adacore.com>
3789 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
3790 * gimple-streamer-out.c (output_bb): Save it.
3791 * lto-streamer-in.c (input_struct_function_base): Restore
3792 instance discriminator if available. Create map on demand.
3793 * lto-streamer-out.c (output_struct_function_base): Save it if
3795 * final.c (decl_to_instance_map): Document LTO strategy.
3797 2018-07-31 Alexandre Oliva <oliva@adacore.com>
3798 Olivier Hainque <hainque@adacore.com>
3800 * debug.h (decl_to_instance_map_t): New type.
3801 (decl_to_instance_map): Declare.
3802 (maybe_create_decl_to_instance_map): New inline function.
3803 * final.c (bb_discriminator, last_bb_discriminator): New statics,
3804 to track basic block discriminators.
3805 (final_start_function_1): Initialize them.
3806 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
3808 (decl_to_instance_map): New variable.
3809 (map_decl_to_instance, maybe_set_discriminator): New functions.
3810 (notice_source_line): Set discriminator.
3812 2018-07-31 Ian Lance Taylor <iant@golang.org>
3814 * targhooks.c (default_have_speculation_safe_value): Add
3817 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3819 * dump-context.h: Include "pretty-print.h".
3820 (dump_context::refresh_dumps_are_enabled): New decl.
3821 (dump_context::emit_item): New decl.
3822 (class dump_context): Add fields "m_test_pp" and
3824 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3825 (temp_dump_context::get_dumped_text): New decl.
3826 (class temp_dump_context): Add field "m_pp".
3827 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
3828 (dump_context::refresh_dumps_are_enabled): ...and add a test for
3830 (set_dump_file): Update for above change.
3831 (set_alt_dump_file): Likewise.
3832 (dump_loc): New overload, taking a pretty_printer *.
3833 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
3834 to any test pretty-printer.
3835 (make_item_for_dump_gimple_stmt): New function, adapted from
3836 optinfo::add_gimple_stmt.
3837 (dump_context::dump_gimple_stmt): Call it, and use the result,
3838 eliminating the direct usage of dump_file and alt_dump_file in
3839 favor of indirectly using them via emit_item.
3840 (make_item_for_dump_gimple_expr): New function, adapted from
3841 optinfo::add_gimple_expr.
3842 (dump_context::dump_gimple_expr): Call it, and use the result,
3843 eliminating the direct usage of dump_file and alt_dump_file in
3844 favor of indirectly using them via emit_item.
3845 (make_item_for_dump_generic_expr): New function, adapted from
3847 (dump_context::dump_generic_expr): Call it, and use the result,
3848 eliminating the direct usage of dump_file and alt_dump_file in
3849 favor of indirectly using them via emit_item.
3850 (make_item_for_dump_printf_va): New function, adapted from
3851 optinfo::add_printf_va.
3852 (make_item_for_dump_printf): New function.
3853 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
3854 and use the result, eliminating the direct usage of dump_file and
3855 alt_dump_file in favor of indirectly using them via emit_item.
3856 (make_item_for_dump_dec): New function.
3857 (dump_context::dump_dec): Call it, and use the result,
3858 eliminating the direct usage of dump_file and alt_dump_file in
3859 favor of indirectly using them via emit_item.
3860 (make_item_for_dump_symtab_node): New function, adapted from
3861 optinfo::add_symtab_node.
3862 (dump_context::dump_symtab_node): Call it, and use the result,
3863 eliminating the direct usage of dump_file and alt_dump_file in
3864 favor of indirectly using them via emit_item.
3865 (dump_context::begin_scope): Reimplement, avoiding direct usage
3866 of dump_file and alt_dump_file in favor of indirectly using them
3868 (dump_context::emit_item): New member function.
3869 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3870 Set up test pretty-printer on the underlying context. Call
3871 refresh_dumps_are_enabled.
3872 (temp_dump_context::~temp_dump_context): Call
3873 refresh_dumps_are_enabled.
3874 (temp_dump_context::get_dumped_text): New member function.
3875 (selftest::verify_dumped_text): New function.
3876 (ASSERT_DUMPED_TEXT_EQ): New macro.
3877 (selftest::test_capture_of_dump_calls): Run all tests twice, with
3878 and then without optinfo enabled. Add uses of
3879 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
3880 * dumpfile.h: Update comment for the dump_* API.
3881 * optinfo-emit-json.cc
3882 (selftest::test_building_json_from_dump_calls): Update for new
3883 param for temp_dump_context ctor.
3884 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
3885 and "m_owned" field.
3886 (optinfo_item::~optinfo_item): Likewise.
3887 (optinfo::add_item): New member function.
3888 (optinfo::emit): Update comment.
3889 (optinfo::add_string): Delete.
3890 (optinfo::add_printf): Delete.
3891 (optinfo::add_printf_va): Delete.
3892 (optinfo::add_gimple_stmt): Delete.
3893 (optinfo::add_gimple_expr): Delete.
3894 (optinfo::add_tree): Delete.
3895 (optinfo::add_symtab_node): Delete.
3896 (optinfo::add_dec): Delete.
3897 * optinfo.h (class dump_context): New forward decl.
3898 (optinfo::add_item): New decl.
3899 (optinfo::add_string): Delete.
3900 (optinfo::add_printf): Delete.
3901 (optinfo::add_printf_va): Delete.
3902 (optinfo::add_gimple_stmt): Delete.
3903 (optinfo::add_gimple_expr): Delete.
3904 (optinfo::add_tree): Delete.
3905 (optinfo::add_symtab_node): Delete.
3906 (optinfo::add_dec): Delete.
3907 (optinfo::add_poly_int): Delete.
3908 (optinfo_item::optinfo_item): Remove "owned" param.
3909 (class optinfo_item): Remove field "m_owned".
3911 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3914 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
3915 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
3916 requested variable alignment.
3917 (expand_one_ssa_partition): Likewise.
3918 (expand_one_var): Likewise.
3920 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3922 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3923 to speculation_safe_value_not_needed.
3925 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3927 * targhooks.h (speculation_safe_value_not_needed): New prototype.
3928 * targhooks.c (speculation_safe_value_not_needed): New function.
3929 * target.def (have_speculation_safe_value): Update documentation.
3930 * doc/tm.texi: Regenerated.
3932 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3934 * config/aarch64/iterators.md (ALLI_TI): New iterator.
3935 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
3937 (despeculate_copy<ALLI:mode>_insn): New insn.
3938 (despeculate_copyti_insn): New insn.
3939 (despeculate_simple<ALLI:mode>): New insn
3940 (despeculate_simpleti): New insn.
3941 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
3943 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
3944 aarch64_speculation_safe_value.
3945 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
3947 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3949 * config/aarch64/aarch64-speculation.cc: New file.
3950 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
3951 before pass_reorder_blocks.
3952 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
3954 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
3955 X14 and X15 when tracking speculation.
3956 * config/aarch64/aarch64.md (register name constants): Add
3957 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
3958 (unspec): Add UNSPEC_SPECULATION_TRACKER.
3959 (speculation_barrier): New insn attribute.
3960 (cmp<mode>): Allow SP in comparisons.
3961 (speculation_tracker): New insn.
3962 (speculation_barrier): Add speculation_barrier attribute.
3963 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
3964 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
3965 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
3967 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3969 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
3970 aarch64_track_speculation is true.
3971 (tb<optab><mode>1): Likewise.
3972 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
3973 generate CB[N]Z when tracking speculation.
3974 (aarch64_split_compare_and_swap): Likewise.
3975 (aarch64_split_atomic_op): Likewise.
3977 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3979 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
3981 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3983 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
3984 (speculation_barrier): New insn.
3986 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3988 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
3989 * config/arm/arm.md (speculation_barrier): New expand.
3990 (speculation_barrier_insn): New pattern.
3992 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
3994 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
3995 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
3996 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
3997 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
3999 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
4000 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
4001 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
4002 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
4003 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
4004 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
4005 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
4006 * builtins.c (expand_speculation_safe_value): New function.
4007 (expand_builtin): Call it.
4008 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
4009 * doc/extend.texi: Document __builtin_speculation_safe_value.
4010 * doc/md.texi: Document "speculation_barrier" pattern.
4011 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
4012 TARGET_HAVE_SPECULATION_SAFE_VALUE.
4013 * doc/tm.texi: Regenerated.
4014 * target.def (have_speculation_safe_value, speculation_safe_value): New
4016 * targhooks.c (default_have_speculation_safe_value): New function.
4017 (default_speculation_safe_value): New function.
4018 * targhooks.h (default_have_speculation_safe_value): Add prototype.
4019 (default_speculation_safe_value): Add prototype.
4021 2018-07-31 David Malcolm <dmalcolm@redhat.com>
4023 * dump-context.h (dump_context::dump_loc): New decl.
4024 * dumpfile.c (dump_context::dump_loc): New member function.
4025 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
4026 and dump_gimple_stmt.
4027 (dump_context::dump_gimple_expr_loc): Likewise, using
4029 (dump_context::dump_generic_expr_loc): Likewise, using
4031 (dump_context::dump_printf_loc_va): Likewise, using
4033 (dump_context::begin_scope): Explicitly using the global function
4034 "dump_loc", rather than the member function.
4036 2018-07-31 Martin Sebor <msebor@redhat.com>
4038 PR tree-optimization/86741
4039 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
4041 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
4043 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
4044 depend on whether prefetch instructions will be emitted or not.
4045 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
4046 will be emitted or not.
4047 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
4048 (TARGET_SETMEM_PFD): New macros.
4050 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4052 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
4053 (NULL_STMT_VEC_INFO): Delete.
4054 (stmt_vec_info::operator*): Likewise.
4055 (stmt_vec_info::operator gimple *): Likewise.
4056 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
4057 of NULL_STMT_VEC_INFO.
4058 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4059 (vect_reassociating_reduction_p): Likewise.
4060 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4061 (vectorizable_store): Likewise.
4062 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
4063 (vec_info::free_stmt_vec_infos): Likewise.
4065 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4067 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
4068 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
4070 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4072 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
4073 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4074 (vec_info::free_stmt_vec_info): New private member functions.
4075 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
4076 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
4077 * tree-parloops.c (gather_scalar_reductions): Remove calls to
4078 set_stmt_vec_info_vec and free_stmt_vec_infos.
4079 * tree-vect-loop.c (_loop_vec_info): Remove call to
4080 set_stmt_vec_info_vec.
4081 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
4082 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
4083 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
4084 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4085 (vec_info::free_stmt_vec_info): ...these new functions. Remove
4086 assignments in {vec_info::,}new_stmt_vec_info that are redundant
4087 with the clearing in the xcalloc.
4088 (stmt_vec_info_vec): Delete.
4089 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
4090 (vectorize_loops): Likewise.
4091 (vec_info::~vec_info): Remove argument from call to
4092 free_stmt_vec_infos.
4093 (vec_info::add_stmt): Remove vinfo argument from call to
4096 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4098 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
4099 rather than a gimple stmt.
4100 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
4101 information for pattern statements when passed the original
4102 statement; instead wait to be passed the pattern statement itself.
4103 Don't call set_vinfo_for_stmt here.
4104 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
4105 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
4106 stmt_vec_infos here.
4107 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
4108 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
4109 stmt_vec_infos entry.
4111 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4113 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
4114 * tree-vectorizer.c (vec_info::replace_stmt): New function.
4115 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
4116 * tree-vect-stmts.c (vectorizable_call): Likewise.
4117 (vectorizable_simd_clone_call): Likewise.
4119 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4121 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
4122 * tree-vectorizer.c (vec_info::remove_stmt): New function.
4123 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
4124 * tree-vect-loop.c (vect_transform_loop): Likewise.
4125 * tree-vect-slp.c (vect_schedule_slp): Likewise.
4126 * tree-vect-stmts.c (vect_remove_stores): Likewise.
4128 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4130 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
4131 (vect_dr_stmt): Delete.
4132 * tree-vectorizer.c (vec_info::lookup_dr): New function.
4133 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
4135 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
4136 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
4137 (vect_verify_datarefs_alignment, vect_peeling_supportable)
4138 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4139 (vect_analyze_data_refs): Likewise.
4140 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
4142 (vect_find_same_alignment_drs): Likewise.
4143 (vect_slp_analyze_node_dependences): Update calls accordingly.
4144 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4145 instead of DR_VECT_AUX.
4146 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
4147 of a vector data references. Use vec_info::lookup_dr instead of
4149 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
4150 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
4151 instead of DR_VECT_AUX.
4153 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4155 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
4157 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4159 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4160 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4161 (vect_gen_prolog_loop_niters): Likewise.
4163 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4165 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
4166 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
4167 (vect_known_alignment_in_bytes, vect_dr_behavior)
4168 (vect_get_scalar_dr_size): Take references as dr_vec_infos
4169 instead of data_references. Update calls to other routines for
4170 which the same change has been made.
4171 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
4172 dr_vec_infos instead of stmt_vec_infos.
4173 (vect_analyze_data_ref_dependence): Update call accordingly.
4174 (vect_slp_analyze_data_ref_dependence)
4175 (vect_record_base_alignments): Use DR_VECT_AUX.
4176 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
4177 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
4178 (vector_alignment_reachable_p, vect_get_data_access_cost)
4179 (vect_peeling_supportable, vect_analyze_group_access_1)
4180 (vect_analyze_group_access, vect_analyze_data_ref_access)
4181 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
4182 (vect_compile_time_alias, vect_small_gap_p)
4183 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
4184 (vect_supportable_dr_alignment): Take references as dr_vec_infos
4185 instead of data_references. Update calls to other routines for
4186 which the same change has been made.
4187 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
4188 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4189 (vect_slp_analyze_and_verify_node_alignment)
4190 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4191 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4192 (vect_setup_realignment): Use dr_vec_infos. Update calls after
4194 (_vect_peel_info::dr): Replace with...
4195 (_vect_peel_info::dr_info): ...this new field.
4196 (vect_peeling_hash_get_most_frequent)
4197 (vect_peeling_hash_choose_best_peeling): Update accordingly.
4198 (vect_peeling_hash_get_lowest_cost):
4199 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
4200 routines for which the same change has been made.
4201 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
4203 * tree-vect-loop-manip.c (get_misalign_in_elems)
4204 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
4206 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4207 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
4208 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
4209 (get_group_load_store_type, get_negative_load_store_type)
4210 (vect_get_data_ptr_increment, vectorizable_store)
4211 (vectorizable_load): Likewise.
4212 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
4213 Update calls to other routines for which the same change has been made.
4215 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4217 * tree-vectorizer.h (vec_info::move_dr): New member function.
4218 (dataref_aux): Rename to...
4219 (dr_vec_info): ...this and add "dr" and "stmt" fields.
4220 (_stmt_vec_info::dr_aux): Update accordingly.
4221 (_stmt_vec_info::data_ref_info): Delete.
4222 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
4223 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
4224 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
4225 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
4227 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
4228 (STMT_VINFO_DR_INFO): New macro.
4229 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
4230 (set_dr_misalignment): Update after rename of dataref_aux.
4231 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
4232 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
4233 initialization of STMT_VINFO_DATA_REF.
4234 * tree-vectorizer.c (vec_info::move_dr): New function.
4235 * tree-vect-patterns.c (vect_recog_bool_pattern)
4236 (vect_recog_mask_conversion_pattern)
4237 (vect_recog_gather_scatter_pattern): Use it.
4238 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
4239 the "dr" and "stmt" fields of dr_vec_info instead of
4240 STMT_VINFO_DATA_REF.
4242 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4244 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
4245 (is_pattern_stmt_p): Use it.
4246 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
4247 on pattern statements.
4249 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4251 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
4252 original stmt as a stmt_vec_info rather than a gimple stmt.
4253 (vect_pattern_recog_1): Take the statement directly as a
4254 stmt_vec_info, rather than via a gimple_stmt_iterator.
4255 Update call to vect_mark_pattern_stmts.
4256 (vect_pattern_recog): Update calls accordingly.
4258 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4260 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
4261 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
4262 a vect_def_type for the first argument.
4263 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
4264 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
4265 operand if it isn't defined by a vectorized statement.
4266 (vect_build_gather_load_calls): Remove the mask_dt argument and
4267 update calls to vect_get_vec_def_for_stmt_copy.
4268 (vectorizable_bswap): Likewise the dt argument.
4269 (vectorizable_call): Update calls to vectorizable_bswap and
4270 vect_get_vec_def_for_stmt_copy.
4271 (vectorizable_simd_clone_call, vectorizable_assignment)
4272 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
4273 (vectorizable_comparison): Update calls to
4274 vect_get_vec_def_for_stmt_copy.
4275 (vectorizable_store): Likewise. Remove now-unnecessary calls to
4277 (vect_get_loop_based_defs): Remove dt argument and update call
4278 to vect_get_vec_def_for_stmt_copy.
4279 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
4280 and vect_get_vec_def_for_stmt_copy.
4281 (vectorizable_load): Update calls to vect_build_gather_load_calls
4282 and vect_get_vec_def_for_stmt_copy.
4283 * tree-vect-loop.c (vect_create_epilog_for_reduction)
4284 (vectorizable_reduction, vectorizable_live_operation): Update calls
4285 to vect_get_vec_def_for_stmt_copy.
4287 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4289 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
4290 and gimple stmt arguments with a stmt_vec_info.
4291 (vect_record_base_alignments): Update calls accordingly.
4292 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
4293 and gimple stmt arguments with a stmt_vec_info.
4294 (vect_build_slp_tree_1): Remove vinfo argument and update call
4295 to vect_record_max_nunits.
4296 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
4297 and vect_record_max_nunits.
4299 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4301 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
4302 file and take a stmt_vec_info instead of a gimple stmt.
4303 (supportable_widening_operation, vect_finish_replace_stmt)
4304 (vect_finish_stmt_generation, vect_get_store_rhs)
4305 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
4306 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
4307 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
4308 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
4309 (vect_create_data_ref_ptr, bump_vector_ptr)
4310 (vect_permute_store_chain, vect_setup_realignment)
4311 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
4312 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
4313 (vectorizable_reduction, vectorizable_induction)
4314 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
4315 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
4316 than gimple stmts as arguments.
4317 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
4318 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
4319 (can_group_stmts_p, vect_check_gather_scatter)
4320 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4321 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
4322 (vect_permute_load_chain, vect_shift_permute_load_chain)
4323 (vect_transform_grouped_load)
4324 (vect_record_grouped_load_vectors): Likewise.
4325 * tree-vect-loop.c (vect_fixup_reduc_chain)
4326 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
4327 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
4328 (vectorizable_reduction, vectorizable_induction)
4329 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
4330 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
4331 (vect_get_load_store_mask): Likewise.
4332 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
4333 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
4334 * tree-vect-stmts.c (vect_mark_relevant)
4335 (is_simple_and_all_uses_invariant)
4336 (exist_non_indexing_operands_for_use_p, process_use)
4337 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
4338 (vect_get_vec_def_for_operand, vect_get_vec_defs)
4339 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
4340 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
4341 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
4342 (get_negative_load_store_type, get_load_store_type)
4343 (vect_check_load_store_mask, vect_check_store_rhs)
4344 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
4345 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4346 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4347 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4348 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
4349 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
4350 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
4351 (supportable_widening_operation): Likewise.
4353 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4355 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
4356 a stmt_vec_info instead of a gcall.
4357 (vect_check_gather_scatter): Update call accordingly.
4358 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
4360 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
4361 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
4362 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
4363 instead of a gimple stmt.
4364 (vect_transform_loop): Update calls accordingly.
4365 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
4366 stmt_vec_infos instead of gimple stmts.
4367 (vect_analyze_slp_instance): Update use accordingly.
4368 * tree-vect-stmts.c (read_vector_array, write_vector_array)
4369 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
4370 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
4371 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
4372 (vect_gen_widened_results_half, vect_get_loop_based_defs)
4373 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
4374 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
4375 down to subroutines.
4377 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4379 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
4380 of the worklist from a vector of gimple stmts to a vector of
4382 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4383 (vect_mark_stmts_to_be_vectorized): Likewise
4385 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4387 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
4388 statement before passing it to vect_analyze_stmt.
4389 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
4390 the chain of phi vector definitions. Track the exit phi via its
4392 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
4393 STMT_VINFO_REDUC_DEF.
4394 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
4395 stmt_vec_infos to handle the statement chains.
4396 (vect_get_slp_defs): Record the first statement in the node
4397 using a stmt_vec_info.
4398 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
4399 statements here and pass their stmt_vec_info down to subroutines.
4400 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
4401 down to vect_finish_stmt_generation.
4402 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
4403 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
4404 stmt_vec_infos to subroutines.
4405 (vect_remove_stores): Use stmt_vec_infos to handle the statement
4408 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4410 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
4411 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
4412 (vect_permute_store_chain, vect_setup_realignment)
4413 (vect_permute_load_chain, vect_shift_permute_load_chain)
4414 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
4415 stmts internally, and when passing values to other vectorizer routines.
4416 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4417 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
4418 (vect_analyze_loop_operations, get_initial_def_for_reduction)
4419 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
4420 (vectorizable_reduction, vectorizable_induction)
4421 (vectorizable_live_operation, vect_transform_loop_stmt)
4422 (vect_transform_loop): Likewise.
4423 * tree-vect-patterns.c (vect_reassociating_reduction_p)
4424 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
4425 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
4426 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4427 (vect_slp_analyze_node_operations_1): Likewise.
4428 * tree-vect-stmts.c (vect_mark_relevant, process_use)
4429 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
4430 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
4431 (vect_finish_stmt_generation_1, get_group_load_store_type)
4432 (get_load_store_type, vect_build_gather_load_calls)
4433 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4434 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4435 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4436 (vectorizable_store, vectorizable_load, vectorizable_condition)
4437 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
4438 (supportable_widening_operation): Likewise.
4439 (vect_get_vector_types_for_stmt): Likewise.
4440 * tree-vectorizer.h (vect_dr_behavior): Likewise.
4442 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4444 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4445 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
4446 (vect_permute_store_chain, vect_permute_load_chain)
4447 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
4448 repeated stmt_vec_info lookups.
4449 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4450 (vect_update_ivs_after_vectorizer): Likewise.
4451 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
4452 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4453 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
4454 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4455 (vect_bb_slp_scalar_cost): Likewise.
4456 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
4458 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4460 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
4461 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
4462 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
4463 to get gassigns and gcalls, rather than operating on generc gimple
4465 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
4466 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
4467 (vectorizable_load, vect_analyze_stmt): Likewise.
4468 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
4470 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4472 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
4473 return stmt_vec_infos rather than gimple stmts. Do not accept
4475 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
4477 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4478 Update use of get_later_stmt.
4479 (vect_get_constant_vectors): Update call accordingly.
4480 (vect_schedule_slp_instance): Likewise
4481 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
4482 (vect_slp_analyze_instance_dependence): Likewise.
4483 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
4485 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4487 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
4488 (stmt_info_for_cost::stmt_info): ...this new field.
4489 (add_stmt_costs): Update accordingly.
4490 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4491 (vect_get_known_peeling_cost): Likewise.
4492 (vect_estimate_min_profitable_iters): Likewise.
4493 * tree-vect-stmts.c (record_stmt_cost): Likewise.
4495 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4497 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
4498 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4499 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4501 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
4503 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4505 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
4506 a gimple stmt to a stmt_vec_info.
4507 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4509 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4511 * tree-vectorizer.h (vec_info::grouped_stores): Change from
4512 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4513 (_loop_vec_info::reduction_chains): Likewise.
4514 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
4516 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4518 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4520 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
4521 a gimple stmt to a stmt_vec_info.
4522 (_stmt_vec_info::next_element): Likewise.
4523 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
4524 (vect_slp_analyze_and_verify_node_alignment)
4525 (vect_analyze_group_access_1, vect_analyze_group_access)
4526 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
4527 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
4528 (vect_supportable_dr_alignment): Update accordingly.
4529 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
4530 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
4531 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
4532 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4533 * tree-vect-slp.c (vect_build_slp_tree_1)
4534 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
4535 (vect_split_slp_store_group, vect_analyze_slp_instance)
4536 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
4537 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
4538 (get_group_load_store_type, get_load_store_type)
4539 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
4540 (vect_transform_stmt, vect_remove_stores): Likewise.
4542 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4544 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
4546 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4547 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
4548 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
4549 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
4550 (vector_alignment_reachable_p, vect_get_data_access_cost)
4551 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
4552 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
4553 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4554 (vect_analyze_group_access_1, vect_analyze_group_access)
4555 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
4556 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
4557 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
4558 result of vect_dr_stmt and use the stmt_vec_info instead of
4559 the associated gimple stmt.
4560 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4561 (vect_gen_prolog_loop_niters): Likewise.
4562 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4564 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4566 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
4567 to a vec<stmt_vec_info>.
4568 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
4569 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
4571 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
4572 to a vec<stmt_vec_info>.
4573 (bst_traits::value_type, bst_traits::value_type): Likewise.
4574 (bst_traits::hash): Update accordingly.
4575 (vect_get_and_check_slp_defs): Change the stmts parameter from
4576 a vec<gimple *> to a vec<stmt_vec_info>.
4577 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
4578 (vect_build_slp_tree): Likewise.
4579 (vect_build_slp_tree_2): Likewise. Update uses of
4580 SLP_TREE_SCALAR_STMTS.
4581 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
4582 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
4583 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
4584 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
4585 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
4586 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
4587 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
4588 (vect_get_constant_vectors, vect_get_slp_defs)
4589 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
4590 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
4591 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
4592 instead of gimple stmts.
4593 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
4594 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
4595 (vect_slp_analyze_instance_dependence): Update uses of
4596 SLP_TREE_SCALAR_STMTS.
4597 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4598 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
4599 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
4600 (get_initial_defs_for_reduction): Likewise.
4601 (vect_create_epilog_for_reduction): Likewise.
4602 (vectorize_fold_left_reduction): Likewise.
4603 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
4604 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
4605 (can_vectorize_live_stmts): Likewise.
4607 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4609 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
4610 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4611 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
4613 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
4615 (gather_scalar_reductions): Update after above interface changes.
4616 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
4617 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
4619 (vect_force_simple_reduction): Likewise.
4620 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
4621 LOOP_VINFO_REDUCTIONS.
4622 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4624 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4626 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
4627 a gimple stmt to a stmt_vec_info.
4628 * tree-vect-loop.c (vect_active_double_reduction_p)
4629 (vect_force_simple_reduction, vectorizable_reduction): Update
4632 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4634 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
4635 vec<gimple *> to a vec<stmt_vec_info>.
4636 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
4637 the reduction_phis argument from a vec<gimple *> to a
4639 (vectorizable_reduction): Likewise the phis local variable that
4640 is passed to vect_create_epilog_for_reduction. Update for new type
4641 of SLP_TREE_VEC_STMTS.
4642 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
4643 (vectorizable_live_operation): Likewise.
4644 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
4645 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
4647 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4649 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
4650 a gimple stmt to a stmt_vec_info.
4651 (vectorizable_condition, vectorizable_live_operation)
4652 (vectorizable_reduction, vectorizable_induction): Pass back the
4653 vectorized statement as a stmt_vec_info.
4654 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
4655 use of STMT_VINFO_VEC_STMT.
4656 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
4657 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
4658 as stmt_vec_infos rather than gimple stmts.
4659 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
4661 (vectorizable_live_operation): Likewise.
4662 (vectorizable_reduction, vectorizable_induction): Likewise,
4663 updating use of STMT_VINFO_VEC_STMT.
4664 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
4665 of STMT_VINFO_VEC_STMT.
4666 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
4667 (vectorizable_simd_clone_call, vectorizable_conversion)
4668 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4669 (vectorizable_store, vectorizable_load, vectorizable_condition)
4670 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
4671 from a gimple stmt to a stmt_vec_info.
4672 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
4673 pointer to a stmt_vec_info to the vectorizable_* routines.
4675 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4677 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
4678 a gimple stmt to a stmt_vec_info.
4679 (is_pattern_stmt_p): Update accordingly.
4680 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
4681 (vect_record_grouped_load_vectors): Likewise.
4682 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
4683 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
4684 (vect_model_reduction_cost): Likewise.
4685 (vect_create_epilog_for_reduction): Likewise.
4686 (vectorizable_reduction, vectorizable_induction): Likewise.
4687 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4688 Return the stmt_vec_info for the pattern statement.
4689 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
4690 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
4691 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4692 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
4693 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
4694 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
4695 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
4696 (free_stmt_vec_info, vect_is_simple_use): Likewise.
4698 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4700 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
4701 (vect_finish_stmt_generation): Likewise.
4702 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4703 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
4704 (vect_build_gather_load_calls): Use the return value of the above
4705 functions instead of a separate call to vinfo_for_stmt. Use narrow
4706 scopes for the input gimple stmt and wider scopes for the associated
4707 stmt_vec_info. Use vec_info::lookup_def when setting these
4708 stmt_vec_infos from an SSA_NAME definition.
4709 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4710 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4711 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4712 (vectorizable_store, vectorizable_load, vectorizable_condition)
4713 (vectorizable_comparison): Likewise.
4714 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
4715 (vectorizable_reduction): Likewise.
4717 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4719 * tree-vectorizer.h (vect_is_simple_use): Add an optional
4720 stmt_vec_info * parameter before the optional gimple **.
4721 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4722 (process_use, vect_get_vec_def_for_operand_1): Update callers.
4723 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
4724 * tree-vect-loop.c (vectorizable_reduction): Likewise.
4725 (vectorizable_live_operation): Likewise.
4726 * tree-vect-patterns.c (type_conversion_p): Likewise.
4727 (vect_look_through_possible_promotion): Likewise.
4728 (vect_recog_rotate_pattern): Likewise.
4729 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4731 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4733 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
4734 a typedef to a wrapper class.
4735 (NULL_STMT_VEC_INFO): New macro.
4736 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
4737 (stmt_vec_info::operator*): New function.
4738 (stmt_vec_info::operator gimple *): Likewise.
4739 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
4740 (add_stmt_costs): Likewise.
4741 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
4742 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4743 (vect_get_known_peeling_cost): Likewise.
4744 (vect_estimate_min_profitable_iters): Likewise.
4745 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4746 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
4747 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4748 (vectorizable_store, free_stmt_vec_infos): Likewise.
4749 (new_stmt_vec_info): Change return type of xcalloc to
4752 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4754 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
4755 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
4756 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
4757 a single_imm_use-based sequence.
4758 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
4760 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4762 * tree-vectorizer.h (vec_info::lookup_def): Declare.
4763 * tree-vectorizer.c (vec_info::lookup_def): New function.
4764 * tree-vect-patterns.c (vect_get_internal_def): Use it.
4765 (vect_widened_op_tree): Likewise.
4766 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4767 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
4768 (vectorizable_reduction): Likewise.
4769 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
4771 (vect_is_slp_reduction): Update calls accordingly. Use
4772 vec_info::lookup_def.
4773 (vect_is_simple_reduction): Likewise
4774 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
4776 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4778 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
4779 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
4780 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
4782 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
4783 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
4784 (vect_update_vf_for_slp, vect_analyze_loop_operations)
4785 (vect_is_slp_reduction, vectorizable_induction)
4786 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
4787 * tree-vect-patterns.c (vect_init_pattern_stmt):
4788 (vect_determine_min_output_precision_1, vect_determine_precisions)
4789 (vect_pattern_recog): Likewise.
4790 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
4791 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
4792 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
4793 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4794 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
4795 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
4796 info field from a loop to a loop_vec_info.
4798 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4800 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
4801 (vec_info::add_stmt): Declare.
4802 * tree-vectorizer.c (vec_info::add_stmt): New function.
4803 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
4804 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
4805 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4806 (vectorizable_induction): Likewise.
4807 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
4808 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4809 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
4810 (vectorizable_load): Likewise.
4811 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4812 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
4813 (vect_recog_gather_scatter_pattern): Likewise.
4814 (append_pattern_def_seq): Likewise. Remove a check that is
4815 performed by add_stmt itself.
4817 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4819 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
4820 which make_ssa_name was called with new_stmt before new_stmt
4823 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4825 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
4827 (vect_is_slp_reduction): ...here...
4828 (vect_is_simple_reduction): ...and here. Remove repetition of tests
4829 that are already known to be false.
4831 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4833 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
4834 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
4835 STMT_VINFO_NUM_SLP_USES when it's true.
4836 (vect_free_slp_instance): Add a final_p parameter and pass it to
4838 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
4839 (vect_analyze_slp_instance): Likewise.
4840 (vect_slp_analyze_operations): Likewise.
4841 (vect_slp_analyze_bb_1): Likewise.
4842 * tree-vectorizer.c (vec_info): Likewise.
4843 * tree-vect-loop.c (vect_transform_loop): Likewise.
4845 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4847 * tree-vect-loop.c (vectorizable_reduction): Assert that the
4848 function is not called for second and subsequent members of
4851 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4853 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
4854 cases for nested loops from here to ...
4855 (vect_create_epilog_for_reduction): ...here. Only call
4856 vect_is_simple_use for inner-loop reductions.
4858 2018-07-31 Martin Liska <mliska@suse.cz>
4860 PR gcov-profile/85338
4861 PR gcov-profile/85350
4862 PR gcov-profile/85372
4863 * profile.c (struct location_triplet): New.
4864 (struct location_triplet_hash): Likewise.
4865 (output_location): Do not output a BB that
4866 is already recorded for a line.
4867 (branch_prob): Use streamed_locations.
4869 2018-07-31 Martin Liska <mliska@suse.cz>
4871 PR gcov-profile/85370
4872 * coverage.c (coverage_begin_function): Do not mark target
4873 clones as artificial functions.
4875 2018-07-31 Martin Liska <mliska@suse.cz>
4877 PR gcov-profile/83813
4878 PR gcov-profile/84758
4879 PR gcov-profile/85217
4880 PR gcov-profile/85332
4881 * profile.c (branch_prob): Do not record GOTO expressions
4882 for GIMPLE statements which locations are already streamed.
4884 2018-07-31 Olivier Hainque <hainque@adacore.com>
4886 * gcc.c (handle_spec_function): Accept a soft_matched_part
4887 argument, as do_spec_1. Pass it down to ...
4888 (eval_spec_function): Accept a soft_matched_part argument,
4889 and pass it down to ...
4890 (do_spec_2): Accept a soft_matched_part argument, and pass
4891 it down to do_spec_1.
4892 (do_spec_1): Pass soft_matched_part to handle_spec_function.
4893 (handle_braces): Update call to handle_spec_function.
4894 (driver::set_up_specs): Update calls to do_spec_2.
4895 (compare_debug_dump_opt_spec_function): Likewise.
4896 (compare_debug_self_opt_spec_function): Likewise.
4898 2018-07-31 Olivier Hainque <hainque@adacore.com>
4900 * common.opt (nolibc): New option.
4901 * doc/invoke.texi (Link Options): Document it.
4902 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
4903 * config/alpha/linux.h: Likewise.
4904 * config/arc/elf.h: Likewise.
4905 * config/arm/uclinux-elf.h: Likewise.
4906 * config/arm/unknown-elf.h: Likewise.
4907 * config/avr/avrlibc.h: Likewise.
4908 * config/bfin/bfin.h: Likewise.
4909 * config/bfin/linux.h: Likewise.
4910 * config/bfin/uclinux.h: Likewise.
4911 * config/darwin.h: Likewise.
4912 * config/darwin10.h: Likewise.
4913 * config/darwin12.h: Likewise.
4914 * config/gnu-user.h: Likewise.
4915 * config/lm32/uclinux-elf.h: Likewise.
4916 * config/pa/pa-hpux11.h: Likewise.
4917 * config/pa/pa64-hpux.h: Likewise.
4918 * config/sparc/sparc.h: Likewise.
4920 2018-07-31 Olivier Hainque <hainque@adacore.com>
4922 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
4923 undefined variables.
4925 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4928 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
4931 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
4933 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
4934 terminated string literal.
4936 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
4938 PR rtl-optimization/85160
4939 * combine.c (is_just_move): New function.
4940 (try_combine): Allow combining two instructions into two if neither of
4941 the original instructions was a move.
4943 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
4946 * doc/extend.texi (Global Register Variables): Discourage use of type
4948 (Local Register Variables): Likewise.
4950 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
4952 PR tree-optimization/86506
4953 * hwint.h (ceil_log2): Resync with hwint.c implementation.
4955 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
4958 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
4959 hard_regno, make sure no insn between `from` and `to` clobbers it.
4961 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
4962 Tom de Vries <tdevries@suse.de>
4964 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
4965 (PTX_DEFAULT_RUNTIME_DIM): ... this.
4966 (nvptx_goacc_validate_dims): Set default worker and gang dims to
4967 PTX_DEFAULT_RUNTIME_DIM.
4968 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
4970 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
4972 * config/pa/pa.c (pa_output_addr_vec): Align address table.
4973 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
4974 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
4976 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
4978 * config/rs6000/constraints.md (wG constraint): Delete, no longer
4980 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
4981 predicate to reflect toc fusion has been deleted.
4982 (toc_fusion_mem_raw): Delete, no longer used.
4983 (toc_fusion_mem_wrapped): Likewise.
4984 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
4986 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
4987 Delete, no longer used.
4988 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
4989 meant to be used for toc fusion.
4990 (rs6000_debug_print_mode): Delete toc fusion debugging.
4991 (rs6000_debug_reg_global): Likewise.
4992 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
4993 fusion and secondary reload support that were never used.
4994 (rs6000_option_override_internal): Delete TOC fusion, that was only
4995 partially defined, and it did not work unless you also used the
4997 (rs6000_legitimate_address_p): Delete TOC fusion support.
4998 (rs6000_opt_masks): Likewise.
4999 (fusion_wrap_memory_address): Delete function, no longer used.
5000 (fusion_split_address); Delete TOC fusion support.
5001 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
5002 longer used with toc fusion being deleted.
5003 (TARGET_TOC_FUSION_FP): Likewise.
5004 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
5006 (toc fusion spliter): Delete TOC fusion support.
5007 (toc_fusionload_<mode>): Likewise.
5008 (toc_fusionload_di): Likewise.
5009 (fusion_gpr_load_<mode>): Delete generator function, this insn no
5010 longer needs to be named. Rename predicate to delete TOC fusion.
5011 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5012 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5013 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5014 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5015 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
5017 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
5019 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
5020 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
5021 __int128 in built-in function prototypes.
5022 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
5023 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
5025 2018-07-27 Martin Sebor <msebor@redhat.com>
5027 PR tree-optimization/86696
5028 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
5029 types, including enums.
5030 (handle_char_store): Be prepared for the above function to fail.
5032 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
5034 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
5035 when optimization level is lower than 2 or optimize for size.
5037 2018-07-26 Martin Sebor <msebor@redhat.com>
5039 PR tree-optimization/86043
5040 PR tree-optimization/86042
5041 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
5042 (get_string_cst_length): Rename...
5043 (get_min_string_length): ...to this. Add argument.
5044 (handle_char_store): Extend to handle multi-character stores by
5046 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
5047 * tree.h (initializer_zerop): Add argument.
5049 2018-07-26 Jakub Jelinek <jakub@redhat.com>
5052 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
5053 declare target to variables if they have always,{to,from,tofrom} map
5056 2018-07-26 Martin Liska <mliska@suse.cz>
5059 * lto-wrapper.c: Add linker_output as prefix
5060 for ltrans_output_file.
5062 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
5064 PR rtl-optimization/85805
5065 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
5066 value for hard registers if that was written in the same mode.
5068 2018-07-26 Martin Liska <mliska@suse.cz>
5070 PR gcov-profile/86536
5071 * gcov.c (format_gcov): Use printf format %.*f directly
5072 and do not handle special values.
5074 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5076 * common/config/arc/arc-common.c (arc_option_optimization_table):
5077 Update default optimizations for size.
5079 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5081 * config/arc/arc.md (movsf_insn): Add short instruction selection.
5082 * config/arc/constraints.md (CfZ): New constraint.
5083 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
5084 (subsf3_fpu): Likewise.
5085 (cmpsf_fpu): Likewise.
5086 (cmpsf_fpu_uneq): Likewise.
5088 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5090 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
5092 (arc_is_uncached_mem_p): uncached applies to both the variable and
5095 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5097 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
5100 2018-07-25 David Malcolm <dmalcolm@redhat.com>
5102 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
5103 field "m_scopes" from vec to auto_vec.
5105 2018-07-25 Martin Liska <mliska@suse.cz>
5107 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
5110 2018-07-25 Richard Biener <rguenther@suse.de>
5113 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
5114 special wrt context_die late.
5115 (gen_subprogram_die): Re-use DIEs in local scope.
5117 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
5119 PR tree-optimization/86644
5120 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
5122 2018-07-25 Martin Liska <mliska@suse.cz>
5125 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
5126 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
5128 2018-07-25 Martin Liska <mliska@suse.cz>
5131 * params.def: Explain ASan abbreviation and provide
5132 a documentation link.
5134 2018-07-24 Martin Sebor <msebor@redhat.com>
5136 PR tree-optimization/86622
5137 PR tree-optimization/86532
5138 * builtins.h (string_length): Declare.
5139 * builtins.c (c_strlen): Correct handling of non-constant offsets.
5140 (check_access): Be prepared for non-constant length ranges.
5141 (string_length): Make extern.
5142 * expr.c (string_constant): Only handle the minor non-constant
5143 array index. Use string_constant to compute the length of
5144 a generic string constant.
5146 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
5148 PR tree-optimization/86618
5149 * tree-vect-stmts.c (vectorizable_call): Don't take the address
5150 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
5152 2018-07-24 David Malcolm <dmalcolm@redhat.com>
5154 PR tree-optimization/86636
5155 * json.cc (json::object::set): Fix comment. Add assertions.
5156 (json::array::append): Move here from json.h. Add comment and an
5158 (json::string::string): Likewise.
5159 * json.h (json::array::append): Move to json.cc.
5160 (json::string::string): Likewise.
5161 * optinfo-emit-json.cc
5162 (optrecord_json_writer::impl_location_to_json): Assert that we
5163 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
5164 wrapper around it. Expand the location once, rather than three
5166 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
5167 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
5169 (optrecord_json_writer::optinfo_to_json): Likewise, in four
5170 places. Fix some overlong lines.
5172 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
5174 * config/aarch64/aarch64-simd.md
5175 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
5176 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
5177 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
5178 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
5179 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
5180 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
5181 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
5182 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
5183 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
5185 2018-07-24 Jakub Jelinek <jakub@redhat.com>
5188 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
5189 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
5190 and abs_d == d, do the power of two handling if profitable.
5192 2018-07-24 Richard Biener <rguenther@suse.de>
5194 * match.pd: Add BIT_FIELD_REF canonicalizations.
5196 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5199 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
5201 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5203 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
5204 terminated STRING_CST object.
5206 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
5208 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
5210 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5212 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
5213 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
5214 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
5215 the elements into a register.
5216 (rs6000_split_v4si_init_di_reg): Delete.
5217 (rs6000_split_v4si_init): Delete.
5218 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
5219 (vsx_init_v4si): Rewrite as a define_expand.
5221 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
5223 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
5224 zero_extend argument from memory): New.
5226 2018-07-22 Martin Sebor <msebor@redhat.com>
5229 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
5230 diagnosing calls with unknown arguments unless -Walloca-larger-than
5231 is restricted to less than PTRDIFF_MAX bytes.
5233 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
5235 * doc/gcov.texi (Invoking Gcov): Editorial changes.
5237 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5239 * pretty-print.c (text_info::set_location): Remove redundant
5240 "line_table" parameter from call to rich_location::set_range.
5242 2018-07-20 Martin Sebor <msebor@redhat.com>
5245 * builtins.c (expand_builtin_alloca): Adjust.
5246 * calls.c (alloc_max_size): Simplify.
5247 * cgraphunit.c (cgraph_node::expand): Adjust.
5248 * common.opt (larger_than_size, warn_frame_larger_than): Remove
5250 (frame_larger_than_size): Same.
5251 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
5252 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
5254 * doc/invoke.texi (GCC Command Options): Document option arguments.
5255 Explain byte-size arguments and suffixes.
5256 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
5257 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
5258 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
5259 * doc/options.texi (UInteger): Expand.
5260 (Host_Wide_Int, ByteSize): Document new properties.
5261 * final.c (final_start_function_1): Include sizes in an error message.
5262 * function.c (frame_offset_overflow): Same.
5263 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
5264 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
5265 Diagnose unbounded alloca calls only for limits of less than
5267 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
5268 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
5270 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
5271 only for limits of less than PTRDIFF_MAX.
5272 * langhooks-def.h (lhd_handle_option): Change function argument
5274 * langhooks.c (lhd_handle_option): Same.
5275 * langhooks.h (handle_option): Same.
5276 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
5278 (var_type, var_type_struct): Same.
5279 (var_set): Handle ByteSize flag.
5280 * optc-gen.awk: Add comments to output to ease debugging. Make
5281 use of HOST_WIDE_INT where appropriate.
5282 * opts-gen-save.awk: Use %lx to format unsigned long.
5283 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
5284 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
5285 arguments. Parse bytes-size suffixes.
5286 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
5287 (enum_value_to_arg): Same.
5288 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
5289 (handle_option): Adjust.
5290 (generate_option): Change function argument to HOST_WIDE_INT.
5291 (cmdline_handle_error): Adjust.
5292 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
5293 (set_option): Change function argument to HOST_WIDE_INT.
5294 (option_enabled): Handle cl_host_wide_int.
5295 (get_option_state): Handle CLVC_SIZE.
5296 (control_warning_option): Same.
5297 * opts.c (common_handle_option): Change function argument to
5298 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
5299 OPT_Wvla_larger_than_.
5300 * opts.h (enum cl_var_type): Add an enumerator.
5301 * stor-layout.c (layout_decl): Print a more meaningful warning.
5302 * toplev.c (output_stack_usage): Adjust.
5304 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
5306 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
5307 call to inline_expand_builtin_string_cmp.
5308 (expand_builtin_strcmp): Likewise.
5309 (expand_builtin_strncmp): Likewise.
5310 (inline_string_cmp): Delete the last parameter, change char_type_node
5311 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
5313 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
5314 the inlining expansion on target where the type of the call has same or
5315 narrower precision than unsigned char.
5317 2018-07-20 David Malcolm <dmalcolm@redhat.com>
5319 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
5320 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
5321 * common.opt (fsave-optimization-record): New option.
5322 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
5323 * doc/invoke.texi (-fsave-optimization-record): New option.
5324 * dumpfile.c: Include "optinfo-emit-json.h".
5325 (struct kv_pair): Move to coretypes.h.
5326 (optgroup_options): Make non-static.
5327 (dump_context::end_scope): Call
5328 optimization_records_maybe_pop_dump_scope.
5329 * dumpfile.h (optgroup_options): New decl.
5330 * json.cc: New file.
5332 * optinfo-emit-json.cc: New file.
5333 * optinfo-emit-json.h: New file.
5334 * optinfo.cc: Include "optinfo-emit-json.h".
5335 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
5336 (optinfo_enabled_p): Check optimization_records_enabled_p.
5337 (optinfo_wants_inlining_info_p): Likewise.
5338 * optinfo.h: Update comment.
5339 * profile-count.c (profile_quality_as_string): New function.
5340 * profile-count.h (profile_quality_as_string): New decl.
5341 (profile_count::quality): New accessor.
5342 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
5343 and optinfo_emit_json_cc_tests.
5344 * selftest.h (selftest::json_cc_tests): New decl.
5345 (selftest::optinfo_emit_json_cc_tests): New decl.
5346 * toplev.c: Include "optinfo-emit-json.h".
5347 (compile_file): Call optimization_records_finish.
5348 (do_compile): Call optimization_records_start.
5349 * tree-ssa-live.c: Include optinfo.h.
5350 (remove_unused_scope_block_p): Retain inlining information if
5351 optinfo_wants_inlining_info_p returns true.
5353 2018-07-20 Richard Biener <rguenther@suse.de>
5356 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
5357 to cover -flto-partition=none.
5359 2018-07-20 Martin Liska <mliska@suse.cz>
5361 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
5362 (get_decl_source_range): Remove unused function.
5364 2018-07-20 Richard Biener <rguenther@suse.de>
5366 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
5367 (struct vn_phi_s): Likewise.
5368 (struct vn_reference_s): Likewise.
5369 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
5370 for searching the slot of an entry known to be in the hash itself.
5371 (vn_phi_hasher::equal): Likewise.
5372 (vn_reference_hasher::equal): Likewise.
5373 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
5375 (optimistic_info, current_info): Remove, keeping only valid_info.
5376 (vn_reference_lookup_1): Remove fallback lookup.
5377 (vn_reference_lookup_2): Likewise.
5378 (vn_nary_op_lookup_1): Likewise.
5379 (vn_phi_lookup): Likewise.
5380 (vn_nary_build_or_lookup_1): Make sure to not chain the built
5382 (vn_reference_insert): Adjust, chain the inserted hash element
5383 at last_inserted_ref.
5384 (vn_reference_insert_pieces): Likewise.
5385 (visit_reference_op_call): Likewise.
5386 (vn_nary_op_insert_into): Chain the inserted hash element at
5388 (vn_nary_op_insert_pieces): Adjust.
5389 (vn_nary_op_insert): Likewise.
5390 (vn_nary_op_insert_stmt): Likewise.
5391 (vn_phi_insert): Adjust, chain the inserted hash element at
5393 (process_scc): Remove clearing and copying the optimistic
5394 table. Instead remove elements inserted during an optimistic
5395 iteration from the single table we maintain.
5396 (init_scc_vn): Adjust.
5397 (free_scc_vn): Likewise.
5398 (sccvn_dom_walker::record_cond): Likewise.
5399 (sccvn_dom_walker::after_dom_children): Likewise.
5401 2018-07-19 Martin Sebor <msebor@redhat.com>
5403 PR tree-optimization/84047
5404 PR tree-optimization/83776
5405 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
5406 (check_array_bounds): Call it.
5408 2018-07-19 Martin Sebor <msebor@redhat.com>
5410 * align.h (align_flags): Use member initialization.
5412 2018-07-19 David Malcolm <dmalcolm@redhat.com>
5414 * Makefile.in (OBJS): Add optinfo.o.
5415 * coretypes.h (class symtab_node): New forward decl.
5416 (struct cgraph_node): New forward decl.
5417 (class varpool_node): New forward decl.
5418 * dump-context.h: New file.
5419 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
5421 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
5422 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
5423 (set_alt_dump_file): Likewise.
5424 (dump_context::~dump_context): New dtor.
5425 (dump_gimple_stmt): Move implementation to...
5426 (dump_context::dump_gimple_stmt): ...this new member function.
5427 Add the stmt to any pending optinfo, creating one if need be.
5428 (dump_gimple_stmt_loc): Move implementation to...
5429 (dump_context::dump_gimple_stmt_loc): ...this new member function.
5430 Start a new optinfo and add the stmt to it.
5431 (dump_gimple_expr): Move implementation to...
5432 (dump_context::dump_gimple_expr): ...this new member function.
5433 Add the stmt to any pending optinfo, creating one if need be.
5434 (dump_gimple_expr_loc): Move implementation to...
5435 (dump_context::dump_gimple_expr_loc): ...this new member function.
5436 Start a new optinfo and add the stmt to it.
5437 (dump_generic_expr): Move implementation to...
5438 (dump_context::dump_generic_expr): ...this new member function.
5439 Add the tree to any pending optinfo, creating one if need be.
5440 (dump_generic_expr_loc): Move implementation to...
5441 (dump_context::dump_generic_expr_loc): ...this new member
5442 function. Add the tree to any pending optinfo, creating one if
5444 (dump_printf): Move implementation to...
5445 (dump_context::dump_printf_va): ...this new member function. Add
5446 the text to any pending optinfo, creating one if need be.
5447 (dump_printf_loc): Move implementation to...
5448 (dump_context::dump_printf_loc_va): ...this new member function.
5449 Start a new optinfo and add the stmt to it.
5450 (dump_dec): Move implementation to...
5451 (dump_context::dump_dec): ...this new member function. Add the
5452 value to any pending optinfo, creating one if need be.
5453 (dump_context::dump_symtab_node): New member function.
5454 (dump_context::get_scope_depth): New member function.
5455 (dump_context::begin_scope): New member function.
5456 (dump_context::end_scope): New member function.
5457 (dump_context::ensure_pending_optinfo): New member function.
5458 (dump_context::begin_next_optinfo): New member function.
5459 (dump_context::end_any_optinfo): New member function.
5460 (dump_context::s_current): New global.
5461 (dump_context::s_default): New global.
5462 (dump_scope_depth): Delete global.
5463 (dumpfile_ensure_any_optinfo_are_flushed): New function.
5464 (dump_symtab_node): New function.
5465 (get_dump_scope_depth): Reimplement in terms of dump_context.
5466 (dump_begin_scope): Likewise.
5467 (dump_end_scope): Likewise.
5468 (selftest::temp_dump_context::temp_dump_context): New ctor.
5469 (selftest::temp_dump_context::~temp_dump_context): New dtor.
5470 (selftest::verify_item): New function.
5471 (ASSERT_IS_TEXT): New macro.
5472 (ASSERT_IS_TREE): New macro.
5473 (ASSERT_IS_GIMPLE): New macro.
5474 (selftest::test_capture_of_dump_calls): New test.
5475 (selftest::dumpfile_c_tests): Call it.
5476 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
5477 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
5478 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
5479 descriptive comment.
5480 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
5481 (dump_node, dump_bb): Move these unrelated decls.
5482 (class dump_manager): Add leading comment.
5483 * optinfo.cc: New file.
5484 * optinfo.h: New file.
5486 2018-07-19 Michael Collison <michael.collison@arm.com>
5487 Richard Henderson <rth@redhat.com>
5489 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
5490 (subti): Handle op1 zero.
5491 (subvti4, usub4ti4): New.
5492 (*sub<GPI>3_compare1_imm): New.
5493 (sub<GPI>3_carryinCV): New.
5494 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
5495 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
5497 2018-07-19 Michael Collison <michael.collison@arm.com>
5498 Richard Henderson <rth@redhat.com>
5500 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
5501 (addti3): Create simpler code if low part is already known to be 0.
5502 (addvti4, uaddvti4): New.
5503 (*add<GPI>3_compareC_cconly_imm): New.
5504 (*add<GPI>3_compareC_cconly): New.
5505 (*add<GPI>3_compareC_imm): New.
5506 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
5507 handle constants within this pattern..
5508 (*add<GPI>3_compareV_cconly_imm): New.
5509 (*add<GPI>3_compareV_cconly): New.
5510 (*add<GPI>3_compareV_imm): New.
5511 (add<GPI>3_compareV): New.
5512 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
5513 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
5514 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
5515 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
5516 with 'comparison' operator.
5517 (*add<GPI>3_compareV_cconly_imm): Ditto.
5518 (*add<GPI>3_compareV_cconly): Ditto.
5519 (*add<GPI>3_compareV_imm): Ditto.
5520 (add<GPI>3_compareV): Ditto.
5521 (add<mode>3_carryinC): Ditto.
5522 (*add<mode>3_carryinC_zero): Ditto.
5523 (*add<mode>3_carryinC): Ditto.
5524 (add<mode>3_carryinV): Ditto.
5525 (*add<mode>3_carryinV_zero): Ditto.
5526 (*add<mode>3_carryinV): Ditto.
5528 2018-07-19 Michael Collison <michael.collison@arm.com>
5529 Richard Henderson <rth@redhat.com>
5531 * config/aarch64/aarch64-modes.def (CC_V): New.
5532 * config/aarch64/aarch64-protos.h
5533 (aarch64_addti_scratch_regs): Declare
5534 (aarch64_subvti_scratch_regs): Declare.
5535 (aarch64_expand_subvti): Declare.
5536 (aarch64_gen_unlikely_cbranch): Declare
5537 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
5538 for signed overflow using CC_Vmode.
5539 (aarch64_get_condition_code_1): Handle CC_Vmode.
5540 (aarch64_gen_unlikely_cbranch): New function.
5541 (aarch64_addti_scratch_regs): New function.
5542 (aarch64_subvti_scratch_regs): New function.
5543 (aarch64_expand_subvti): New function.
5545 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5547 * config/aarch64/aarch64-option-extensions.def: New entry for profile
5549 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
5550 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
5553 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5556 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5557 address check not strict.
5559 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
5561 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
5563 (store_pair_lanes<mode>): Likewise.
5564 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
5565 enum value 'ADDR_QUERY_LDP_STP_N'.
5566 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
5567 (aarch64_print_address_internal): Add declaration.
5568 (aarch64_print_ldpstp_address): Remove.
5569 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
5570 (aarch64_print_operand): Change printing of 'y'.
5571 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
5572 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
5573 'true' rather than '1'.
5574 * config/aarch64/constraints.md (Uml): Likewise.
5575 (Uml): Rename to Umn.
5578 2018-07-19 Richard Biener <rguenther@suse.de>
5580 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
5582 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
5583 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
5584 (vn_reference_hasher): Likewise.
5585 (struct vn_tables_s): Remove obstack and alloc-pool members.
5586 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
5587 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
5588 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
5589 (vn_reference_insert_pieces): Likewise.
5590 (alloc_vn_nary_op_noinit): Adjust.
5591 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
5592 (vn_phi_eq): Adjust.
5593 (shared_lookup_phiargs): Remove.
5594 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
5595 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
5596 (visit_reference_op_call): Likewise.
5597 (copy_nary, copy_phi, copy_reference): Remove.
5598 (process_scc): Rewind the obstack when iterating. Do not
5599 copy the elements to valid_info but just move them from one
5600 hashtable to the other.
5601 (allocate_vn_table): Adjust.
5602 (free_vn_table): Likewise.
5603 (init_scc_vn): Likewise.
5604 (free_scc_vn): Likewise.
5606 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
5609 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
5610 indirect_return as function type attribute.
5611 (ix86_attribute_table): Change indirect_return to function
5613 * doc/extend.texi: Update indirect_return attribute.
5615 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
5617 * wide-int.h (widest2_int): New.
5618 * gimple-fold.c (arith_overflowed_p): Use it.
5619 * tree.h (widest2_int_cst): New.
5620 * tree-vrp.c (wide_int_binop_overflow): Rename from
5621 vrp_int_const_binop.
5622 Rewrite to work on trees.
5623 (extract_range_from_multiplicative_op_1): Abstract code to...
5624 (wide_int_range_min_max): ...here.
5625 (wide_int_range_cross_product): ...and here.
5626 (extract_range_from_binary_expr_1): Abstract overflow code to...
5627 (wide_int_range_mult_wrapping): ...here.
5628 * tree-vrp.h (wide_int_range_cross_product): New.
5629 (wide_int_range_mult_wrapping): New.
5631 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
5632 Julia Koval <julia.koval@intel.com>
5634 * config/i386/x86-tune-costs.h (skylake_memcpy,
5635 skylake_memset): Replace rep_prefix with unrolling for size 512.
5637 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
5640 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
5641 comparision with EQ_EXPR in last stmt.
5643 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
5645 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
5646 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
5647 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
5648 previously known as "PowerPC AltiVec Built-in Functions". Move
5649 some material to new subsubsections "PowerPC AltiVec Built-in
5650 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
5652 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
5653 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
5654 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
5655 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
5657 2018-07-18 Richard Biener <rguenther@suse.de>
5659 PR tree-optimization/86557
5660 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
5663 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
5665 * config/s390/s390.c (s390_function_profiler): Generate CFI.
5667 2018-07-17 Jeff Law <law@redhat.com>
5669 * config/arm/arm.c (get_label_padding): Update for recent
5670 changes to label_to_alignment.
5672 PR tree-optimization/86010
5673 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
5675 * config/mips/mips.c (vr4130_align_insns): Update for recent
5676 changes to label_to_alignment.
5678 * config/frv/frv.c (frv_label_align): Update for recent changes
5679 to label_to_alignment.
5681 * config/nios2/nios2.c (nios2_label_align): Update for recent
5682 changes which dropped ALIGN_LABELS_LOG.
5684 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
5686 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
5687 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
5689 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
5691 * config/arc/arc.c (arc_label_align): Use align_labels instead of
5692 deprecated align_labels_log.
5694 2018-07-17 Richard Biener <rguenther@suse.de>
5697 * dwarf2out.c (init_sections_and_labels): Always generate
5698 a debug_line_str_section for early LTO debug.
5699 (dwarf2out_finish): Reset debug_line_str_hash output early.
5700 Bump counter for extra dwarf5 .debug_loc labels to not conflict
5702 (dwarf2out_early_finish): Output debug_line_str.
5704 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
5706 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
5707 index register on z196 or later.
5709 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
5711 * config/s390/s390.c (s390_default_align): Set default function
5713 (s390_override_options_after_change): Call s390_default align.
5714 (s390_option_override_internal): Call s390_default_align.
5715 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
5717 2018-07-17 Jakub Jelinek <jakub@redhat.com>
5720 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
5721 to _looptemp_ clauses, other than the first two.
5723 2018-07-17 Martin Liska <mliska@suse.cz>
5725 * opts.c: Do not enable OPT_falign_* for -Os.
5727 2018-07-17 Martin Liska <mliska@suse.cz>
5729 * align.h (MAX_CODE_ALIGN): New.
5730 (MAX_CODE_ALIGN_VALUE): New.
5731 * common/config/i386/i386-common.c (ix86_handle_option):
5732 (MAX_CODE_ALIGN): Moved to align.h.
5733 * final.c (MAX_CODE_ALIGN): Likewise.
5734 * opts.c (parse_and_check_align_values):
5735 (MAX_CODE_ALIGN): Likewise.
5736 (MAX_CODE_ALIGN_VALUE): Likewise.
5738 2018-07-17 Martin Liska <mliska@suse.cz>
5740 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
5741 in order to fulfil coding style.
5742 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
5743 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5744 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5745 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
5746 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
5747 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
5748 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
5749 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
5750 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5752 2018-07-17 Martin Liska <mliska@suse.cz>
5754 * align.h: New file.
5755 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
5757 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
5758 align_flags of label_to_alignment.
5759 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
5761 * config/m68k/m68k.c: Do not use removed align_labels_value and
5763 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
5764 (LOOP_ALIGN): Likewise.
5765 (LABEL_ALIGN): Likewise.
5766 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
5767 Remove not used macro.
5768 (rs6000_loop_align): Change return type to align_flags.
5769 (rs6000_loop_align_max_skip): Remove.
5770 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
5771 Change return type to align_flags.
5772 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
5773 Remove not used macro.
5774 (rs6000_loop_align): Change return type to align_flags.
5775 (rs6000_loop_align_max_skip): Remove.
5776 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
5777 * config/rx/rx-protos.h (rx_align_for_label): Make it
5779 * config/rx/rx.c (rx_align_for_label): Change return type
5781 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
5783 into align_flags class.
5784 (LABEL_ALIGN): Likewise.
5785 (LOOP_ALIGN): Likewise.
5786 * config/s390/s390.c (s390_label_align): Use align_flags
5788 (s390_asm_output_function_label): Likewise.
5789 * config/sh/sh.c (sh_override_options_after_change):
5790 Use align_flags class directly without macros.
5791 (find_barrier): Likewise.
5792 (barrier_align): Likewise.
5793 (sh_loop_align): Likewise.
5794 * config/spu/spu.c (spu_option_override):
5795 Use align_flags_tuple::get_value instead of removed macros.
5796 (spu_sched_init): Likewise.
5797 * config/spu/spu.h (GTY): Likewise.
5798 * config/visium/visium.c (visium_option_override):
5799 Set "8" as default secondary alignment.
5800 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
5801 in order to guarantee secondary alignment of 8.
5802 * coretypes.h: Include align.h header file.
5803 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
5804 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
5805 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
5806 * doc/tm.texi.in: Likewise.
5807 * final.c (struct label_alignment): Remove not used structure.
5808 (LABEL_ALIGN): Change type to align_flags.
5809 (LOOP_ALIGN): Likewise.
5810 (JUMP_ALIGN): Likewise.
5811 (default_loop_align_max_skip): Remove.
5812 (default_label_align_max_skip): Likewise.
5813 (default_jump_align_max_skip): Likewise.
5814 (default_label_align_after_barrier_max_skip):
5815 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
5816 (LABEL_TO_MAX_SKIP): Remove.
5817 (label_to_alignment): Return align_flags type instead of integer.
5818 (label_to_max_skip): Remove.
5819 (align_fuzz): Use align_flags type.
5820 (compute_alignments): Use align_flags type and use align_flags::max
5821 to combine multiple alignments.
5822 (grow_label_align): Grow vec instead of C array.
5823 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
5824 (shorten_branches): Use align_flags type and use align_flags::max
5825 to combine multiple alignments.
5826 (final_scan_insn_1): Remove usage of secondary alignment that comes
5827 from label alignment, but instead use proper secondary alignment
5828 which is computed in grow_label_align.
5829 * flags.h (struct align_flags_tuple): Move to align.h.
5830 (struct align_flags): Likewise.
5831 (state_align_loops): Rename to align_loops.
5832 (state_align_jumps): Rename to align_jumps.
5833 (state_align_labels): Rename to align_labels.
5834 (state_align_functions): Rename to align_functions.
5835 (align_loops_log): Remove.
5836 (align_jumps_log): Remove.
5837 (align_labels_log): Remove.
5838 (align_functions_log): Remove.
5839 (align_loops_max_skip): Remove.
5840 (align_jumps_max_skip): Remove.
5841 (align_labels_max_skip): Remove.
5842 (align_functions_max_skip): Remove.
5843 (align_loops_value): Remove.
5844 (align_jumps_value): Remove.
5845 (align_labels_value): Remove.
5846 (align_functions_value): Remove.
5847 * output.h (label_to_alignment): Change return type to align_flags.
5848 (label_to_max_skip): Remove.
5849 * target.def: Remove loop_align_max_skip, label_align_max_skip,
5850 jump_align_max_skip macros.
5851 * targhooks.h (default_loop_align_max_skip): Remove.
5852 (default_label_align_max_skip): Likewise.
5853 (default_jump_align_max_skip): Likewise.
5854 (default_label_align_after_barrier_max_skip): Remove.
5855 * toplev.c (read_log_maxskip): Use ::normalize function.
5856 (parse_N_M): Remove not used argument and also call ::normalize.
5857 (parse_alignment_opts): Do not pass unused arguments.
5858 * varasm.c (assemble_start_function): Use directly align_functions
5859 instead of removed macros.
5860 * system.h: Do not poison removed macros.
5862 2018-07-17 Jakub Jelinek <jakub@redhat.com>
5865 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
5866 and cond temporaries don't have reference type if iterator has
5867 pointer type. For init use &for_pre_body instead of pre_p if
5868 for_pre_body is non-empty.
5870 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5872 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
5873 double-double modes to SFmode directly directly.
5874 (trunc<mode>sf2_fprs): Delete.
5876 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5878 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
5879 for conversions between IFmode and the decimal floating point modes.
5880 (init_float128_ieee): Use the correct names for conversions between
5881 KFmode and the decimal floating point modes.
5883 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
5885 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
5886 for the conversions between TDmode and IFmode.
5887 (init_float128_ieee): Use more correct names for the conversions
5888 between TDmode and KFmode.
5890 2018-07-16 Jakub Jelinek <jakub@redhat.com>
5892 PR tree-optimization/86526
5893 * builtins.c (expand_builtin_memcmp): Formatting fixes.
5894 (inline_expand_builtin_string_cmp): Likewise.
5895 (inline_string_cmp): Likewise. Use c_readstr instead of
5896 builtin_memcpy_read_str. Add unit_mode temporary.
5898 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
5901 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
5902 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
5904 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5906 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
5907 Alphabetize prototypes of built-in functions, separating out
5908 built-in functions that are listed in this section but should be
5909 described elsewhere.
5911 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
5914 * expmed.c (emit_store_flag): Do not emit setcc followed by a
5915 conditional move when trapping comparison was split to a
5916 non-trapping one (and vice versa).
5918 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5920 * config/s390/s390.c (s390_function_profiler): Generate nops
5921 instead of profiler call sequences.
5922 * config/s390/s390.opt: Add the new option.
5924 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5926 * config/s390/s390.c (s390_function_profiler): Generate
5927 __mcount_loc section.
5928 * config/s390/s390.opt: Add the new option.
5930 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
5932 * common.opt: Add the new warning.
5933 * config/s390/s390.c (s390_function_profiler): Emit "brasl
5934 %r0,__fentry__" when -mfentry is specified.
5935 (s390_option_override_internal): Disallow -mfentry for 31-bit
5937 * config/s390/s390.opt: Add the new option.
5939 2018-07-16 Richard Biener <rguenther@suse.de>
5942 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
5943 for function-local FUNCTION_DECL and RESULT_DECL immediately.
5945 2018-07-16 Martin Liska <mliska@suse.cz>
5948 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
5951 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5953 * config/arc/arcHS.md: Update ARCHS scheduling rules.
5955 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
5957 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
5959 * config/arc/arc-cpus.def (hs4x): New cpu.
5961 * config/arc/arc-tables.opt: Regenerate.
5962 * config/arc/arc.c (arc_sched_issue_rate): New function.
5963 (TARGET_SCHED_ISSUE_RATE): Define.
5964 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
5965 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
5967 (attr tune): Add ARCHS4x tune values.
5968 (attr tune_dspmpy): Define.
5969 (*tst): Correct instruction type.
5970 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
5971 * config/arc/arcHS4x.md: New file.
5972 * config/arc/fpu.md: Update instruction type attributes.
5973 * config/arc/t-multilib: Regenerate.
5975 2018-07-16 Tom de Vries <tdevries@suse.de>
5978 * var-tracking.c (vt_initialize): Fix pre_dec handling.
5980 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5982 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
5985 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
5987 PR tree-optimization/86514
5988 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
5989 conversion to a boolean type from a type with greater precision.
5991 2018-07-16 Tom de Vries <tdevries@suse.de>
5993 * var-tracking.c (vt_initialize): Print adjusted insn slim if
5994 dump_flags request TDF_SLIM.
5996 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
5998 * fold-const.c (int_const_binop_1): Abstract...
5999 (wide_int_binop): ...wide int code here.
6000 (poly_int_binop): ...poly int code here.
6001 Abstract the rest of int_const_binop_1 into int_const_binop.
6002 * fold-const.h (wide_int_binop): New.
6003 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
6004 Remove useless PLUS/MINUS_EXPR case.
6005 (zero_nonzero_bits_from_vr): Move wide int code...
6006 (zero_nonzero_bits_from_bounds): ...here.
6007 (extract_range_from_binary_expr_1): Move mask optimization code...
6008 (range_easy_mask_min_max): ...here.
6009 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
6010 (range_easy_mask_min_max): New.
6012 2018-07-15 Jeff Law <law@redhat.com>
6015 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
6018 2018-07-14 Jim Wilson <jimw@sifive.com>
6020 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
6022 2018-07-14 Paul Koning <ni1d@arrl.net>
6024 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
6026 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
6028 * lto-streamer-out.c (copy_function_or_variable): Dump info about
6031 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
6032 Steve Munroe <munroesj52@gmail.com>
6034 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
6035 (_mm_andnot_si128): Likewise.
6036 (_mm_or_si128): Likewise.
6037 (_mm_xor_si128): Likewise.
6039 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
6042 * builtins.c (expand_builtin_memcmp): Inline the calls first
6043 when result_eq is false.
6044 (expand_builtin_strcmp): Inline the calls first.
6045 (expand_builtin_strncmp): Likewise.
6046 (inline_string_cmp): New routine. Expand a string compare
6047 call by using a sequence of char comparison.
6048 (inline_expand_builtin_string_cmp): New routine. Inline expansion
6049 a call to str(n)cmp/memcmp.
6050 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
6052 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
6054 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
6056 * config/arm/driver-arm.c: Include arm-native.h.
6057 (host_detect_local_cpu): Use auto-generated data tables.
6058 (vendors, arm_cpu_table): Delete. Move part information to ...
6059 * config/arm/arm-cpus.in: ... here.
6060 * config/arm/parsecpu.awk (gen_native): New function.
6061 (vendor, part): New CPU fields.
6062 (END): Add support for building the native CPU detection tables.
6063 * config/arm/t-arm (arm-native.h): Add build rule.
6064 (driver-arm.o): Add dependency on arm-native.h.
6066 2018-07-13 Richard Biener <rguenther@suse.de>
6069 * match.pd (addr1 - addr2): Allow either of the operand to
6072 2018-07-13 Tom de Vries <tdevries@suse.de>
6074 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
6077 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
6079 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
6080 arrays instead of numbered variables.
6082 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
6084 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
6085 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
6087 2018-07-13 Richard Biener <rguenther@suse.de>
6090 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
6091 instead of get_context_die.
6093 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
6094 Richard Biener <rguenther@suse.de>
6097 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
6098 that the loop latch destination where phi is defined.
6100 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
6102 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
6103 (riscv_expand_epilogue): Add assertion to check interrupt mode.
6104 (riscv_set_current_function): Extract getting interrupt type to new
6106 (riscv_get_interrupt_type): New function.
6107 (riscv_merge_decl_attributes): New function, checking interrupt type is
6109 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
6111 2018-07-12 Paul Koning <ni1d@arrl.net>
6113 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
6116 2018-07-12 Paul Koning <ni1d@arrl.net>
6118 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
6119 zero reference, add doloop_end instead.
6120 * doc/md.texi (decrement_and_branch_until_zero): Remove.
6121 (Looping patterns): Remove decrement_and_branch_until_zero. Add
6122 detail for doloop_end.
6124 2018-07-12 Martin Sebor <msebor@redhat.com>
6127 * attribs.c (decl_attributes): Reject conflicting attributes before
6128 calling attribute handlers.
6130 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6132 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
6134 (gcc::dump_manager::get_dump_file_name): likewise.
6135 (dump_begin): Likewise.
6136 * dumpfile.h (dump_begin): Update prototype.
6137 (gcc::dump_manager::get_dump_file_name,
6138 gcc::dump_manager::get_dump_file_name): Update prototype.
6140 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6142 * internal-fn.h (vectorizable_internal_fn_p): New function.
6143 * tree-vect-slp.c (compatible_calls_p): Likewise.
6144 (vect_build_slp_tree_1): Remove nops argument. Handle calls
6145 to internal functions.
6146 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
6148 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6150 * fold-const.h (inverse_conditions_p): Declare.
6151 * fold-const.c (inverse_conditions_p): New function.
6152 * match.pd: Use inverse_conditions_p. Add folds of view_converts
6153 that test the inverse condition of a conditional internal function.
6154 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
6155 * internal-fn.c (internal_fn_mask_index): Handle conditional
6157 (vectorized_internal_fn_supported_p): New function.
6158 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
6159 (any_pred_load_store): Replace with...
6160 (need_to_predicate): ...this new variable.
6161 (redundant_ssa_names): New variable.
6162 (ifcvt_can_use_mask_load_store): Move initial checks to...
6163 (ifcvt_can_predicate): ...this new function. Handle tree codes
6164 for which a conditional internal function exists.
6165 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
6166 instead of ifcvt_can_use_mask_load_store. Update after variable
6168 (predicate_load_or_store): New function, split out from
6169 predicate_mem_writes.
6170 (check_redundant_cond_expr): New function.
6171 (value_available_p): Likewise.
6172 (predicate_rhs_code): Likewise.
6173 (predicate_mem_writes): Rename to...
6174 (predicate_statements): ...this. Use predicate_load_or_store
6175 and predicate_rhs_code.
6176 (combine_blocks, tree_if_conversion): Update after above name changes.
6177 (ifcvt_local_dce): Handle redundant_ssa_names.
6178 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
6179 general conditional functions.
6180 * tree-vect-stmts.c (vectorizable_call): Likewise.
6182 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6183 Alan Hayward <alan.hayward@arm.com>
6184 David Sherwood <david.sherwood@arm.com>
6186 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
6187 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
6188 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
6189 plus and minus and convert them into IFN_COND_FMA-based sequences.
6190 (convert_mult_to_fma): Handle conditional plus and minus.
6192 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6194 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
6195 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
6196 (cond_fnms_optab): New optabs.
6197 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
6199 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
6200 * internal-fn.h (get_conditional_internal_fn): Declare.
6201 (get_unconditional_internal_fn): Likewise.
6202 * internal-fn.c (cond_ternary_direct): New macro.
6203 (expand_cond_ternary_optab_fn): Likewise.
6204 (direct_cond_ternary_optab_supported_p): Likewise.
6205 (FOR_EACH_COND_FN_PAIR): Likewise.
6206 (get_conditional_internal_fn): New function.
6207 (get_unconditional_internal_fn): Likewise.
6208 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
6209 (gimple_match_op::gimple_match_op): Add a new overload for 5
6211 (gimple_match_op::set_op): Likewise.
6212 (gimple_resimplify5): Declare.
6213 * genmatch.c (decision_tree::gen): Generate simplifications for
6215 * gimple-match-head.c (gimple_simplify): Define an overload for
6216 5 operands. Handle calls with 5 arguments in the top-level overload.
6217 (convert_conditional_op): Handle conversions from unconditional
6218 internal functions to conditional ones.
6219 (gimple_resimplify5): New function.
6220 (build_call_internal): Pass a fifth operand.
6221 (maybe_push_res_to_seq): Likewise.
6222 (try_conditional_simplification): Try converting conditional
6223 internal functions to unconditional internal functions.
6224 Handle 3-operand unconditional forms.
6225 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
6226 Define ternary equivalents of the current rules for binary conditional
6228 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
6230 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
6231 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
6232 (optab): Handle them.
6233 (SVE_COND_FP_TERNARY): New int iterator.
6234 (sve_fmla_op, sve_fmad_op): New int attributes.
6235 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
6236 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
6237 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
6239 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
6241 * target.def (preferred_else_value): New target hook.
6242 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
6243 * doc/tm.texi: Regenerate.
6244 * targhooks.h (default_preferred_else_value): Declare.
6245 * targhooks.c (default_preferred_else_value): New function.
6246 * internal-fn.h (conditional_internal_fn_code): Declare.
6247 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
6248 (get_conditional_internal_fn): Use it.
6249 (conditional_internal_fn_code): New function.
6250 * gimple-match.h (gimple_match_cond): New struct.
6251 (gimple_match_op): Add a cond member function.
6252 (gimple_match_op::gimple_match_op): Update all forms to take a
6254 * genmatch.c (expr::gen_transform): Use the same condition as res_op
6255 for the suboperation, but don't specify a particular else_value.
6256 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
6257 (visit_nary_op, visit_reference_op_load): Pass
6258 gimple_match_cond::UNCOND to the gimple_match_op constructor.
6259 * gimple-match-head.c: Include tree-eh.h
6260 (convert_conditional_op): New function.
6261 (maybe_resimplify_conditional_op): Likewise.
6262 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
6263 (gimple_resimplify2): Likewise.
6264 (gimple_resimplify3): Likewise.
6265 (gimple_resimplify4): Likewise.
6266 (maybe_push_res_to_seq): Return null for conditional operations.
6267 (try_conditional_simplification): New function.
6268 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
6270 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
6272 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
6274 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
6276 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
6278 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6280 (hash_tree): Do not hash DECL_FCONTEXT
6281 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
6282 Do not stream DECL_FCONTEXT.
6283 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
6284 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
6286 2018-07-12 Richard Biener <rguenther@suse.de>
6289 * dwarf2out.c (gen_block_die): Only output blocks when they have
6290 at least one !DECL_IGNORED_P variable.
6292 2018-07-12 Richard Biener <rguenther@suse.de>
6295 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
6296 Remove -mieee-fp handling.
6298 2018-07-12 Richard Biener <rguenther@suse.de>
6300 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
6301 left-over from last patch.
6303 2018-07-12 Jakub Jelinek <jakub@redhat.com>
6305 PR tree-optimization/86492
6306 * gimple-ssa-store-merging.c
6307 (imm_store_chain_info::coalesce_immediate_stores): Call
6308 check_no_overlap even for the merge_overlapping case. Formatting fix.
6310 2018-07-12 Richard Biener <rguenther@suse.de>
6313 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
6314 move possibly trapping operations into the conditional.
6316 2018-07-12 Richard Biener <rguenther@suse.de>
6318 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
6319 (vn_lookup_simplify_result): Remove recursion limit applied
6321 (vn_nary_build_or_lookup_1): Adjust.
6322 (try_to_simplify): Likewise.
6323 * gimple-match-head.c (gimple_resimplify1): Instead apply one
6325 (gimple_resimplify2): Likewise.
6326 (gimple_resimplify3): Likewise.
6327 (gimple_resimplify4): Likewise.
6329 2018-07-11 Jakub Jelinek <jakub@redhat.com>
6331 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
6332 Use __mmask64 type instead of __mmask8 for __M argument.
6333 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
6334 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
6336 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
6337 __mmask16 for __M argument.
6338 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
6339 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
6340 to __mmask16 instead of __mmask8.
6341 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
6342 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
6343 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
6344 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
6345 instead of __mmask16 for __U argument.
6346 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
6347 __mmask16 instead of __mmask8 for __U argument.
6348 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
6350 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
6352 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
6354 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
6356 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
6358 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
6360 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
6362 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
6364 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
6366 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
6368 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
6369 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
6370 return type as well as __M argument type and all casts from __mmask8
6372 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
6373 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
6374 return type as well as __M argument type and all casts from __mmask8
6376 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
6377 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
6378 return type as well as __M argument type and all casts from __mmask8
6380 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
6381 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
6382 return type as well as __M argument type and all casts from __mmask8
6384 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
6385 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
6388 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
6390 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
6391 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
6394 2018-07-11 Paul Koning <ni1d@arrl.net>
6396 * doc/md.texi (define_subst): Document how multiple occurrences of
6397 the same argument in the replacement pattern are handled.
6399 2018-07-11 Paul Koning <ni1d@arrl.net>
6401 * doc/extend.texi (Common Variable Attributes): Move "mode" into
6403 (Common Type Attributes): Add "mode" attribute.
6405 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
6407 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
6408 stream DECL_ORIGINAL_TYPE.
6409 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
6410 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
6411 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
6412 Do not walk original type.
6413 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
6415 (write_ts_decl_non_common_tree_pointers): Do not stream
6417 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
6418 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
6420 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6422 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
6423 thread twice from the same starting edge.
6425 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
6427 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
6428 * gimple.c (gimple_call_nonnull_result_p): ...here...
6429 (gimple_call_nonnull_arg): ...and here.
6430 * gimple.h (gimple_call_nonnull_result_p): New.
6431 (gimple_call_nonnull_arg): New.
6433 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
6435 * config/arm/arm-cpus.in: Move information from fpu field of each
6436 cpu definition to the isa field.
6437 * config/arm/parsecpu.awk (fpu): Delete match rule.
6438 (gen_comm_data): Don't add bits from the CPU's FPU entry.
6440 2018-07-11 Richard Biener <rguenther@suse.de>
6443 * dwarf2out.c (init_sections_and_labels): Use
6444 output_asm_line_debug_info consistently.
6445 (dwarf2out_early_finish): Likewise.
6446 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
6449 2018-07-11 Richard Biener <rguenther@suse.de>
6451 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
6452 Rework father_bb setting in a way to avoid propagating constants
6453 multiple times on a loop body.
6455 2018-07-10 Mark Wielaard <mark@klomp.org>
6458 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
6460 2018-07-10 Richard Biener <rguenther@suse.de>
6462 * hash-map.h (hash_map::iterator::operator*): Return
6463 references to key and value.
6465 2018-07-10 Jakub Jelinek <jakub@redhat.com>
6468 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
6469 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
6470 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
6471 (gimplify_omp_for): For composite loops, move outer
6472 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
6473 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
6474 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
6475 TREE_LIST for both the original class iterator and the "last" helper
6476 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
6477 loop, remember has_decl_expr from outer composite loops for the
6478 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
6480 2018-07-09 Martin Sebor <msebor@redhat.com>
6484 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
6485 accessing implicitly initialized array elements.
6486 * expr.c (string_constant): Handle string initializers of
6487 character arrays within aggregates.
6488 * gimple-fold.c (fold_array_ctor_reference): Add argument.
6489 Store element offset. As a special case, handle zero size.
6490 (fold_nonarray_ctor_reference): Same.
6491 (fold_ctor_reference): Add argument. Store subobject offset.
6492 * gimple-fold.h (fold_ctor_reference): Add argument.
6494 2018-07-09 Paul Koning <ni1d@arrl.net>
6496 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
6497 (pdp11_insn_cost): New function.
6498 (pdp11_md_asm_adjust): New function.
6499 (TARGET_INVALID_WITHIN_DOLOOP): Define.
6500 (pdp11_rtx_costs): Update to match machine better.
6501 (output_addr_const_pdp11): Correct format mismatch warnings.
6502 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
6503 * config/pdp11/pdp11.md: General change to add base_cost and/or
6504 length attributes for use by new pdp11_insn_cost function.
6505 (MIN_BRANCH): Correct definition.
6507 (doloop_end): Use standard pattern name for looping pattern.
6508 (doloop_end_nocc): New.
6509 (movsf): Add another constraint alternative.
6510 (zero_extendqihi2): Add constraint alternatives for not in place
6512 (zero_extendhisi2): Remove.
6513 (shift patterns): Add CC handling variants.
6517 (define_peephole2): New peephole to recognize mov that sets CC for
6520 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6522 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
6523 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
6526 2018-07-09 Jakub Jelinek <jakub@redhat.com>
6529 * real.c (real_nextafter): Return true if result is denormal.
6531 2018-07-09 Martin Liska <mliska@suse.cz>
6533 * common.opt: Add back wrongly removed attribute.
6535 2018-07-09 Richard Biener <rguenther@suse.de>
6538 * dwarf2out.c (gen_block_die): For an early generated DIE
6539 always output high/low PC attributes.
6541 2018-07-09 Tom de Vries <tdevries@suse.de>
6543 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
6544 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
6545 onto VAR_DECL with abstract origin.
6547 2018-07-07 Jim Wilson <jimw@sifive.com>
6549 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
6551 2018-07-07 Tom de Vries <tdevries@suse.de>
6553 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
6555 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
6557 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
6559 (combine_bound): Use wide-int overflow calculation instead of
6561 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
6563 * fold-const.c (int_const_binop_2): Same.
6564 (extract_muldiv_1): Same.
6565 (fold_div_compare): Same.
6566 (fold_abs_const): Same.
6568 * poly-int.h (add): Same.
6572 * predict.c (predict_iv_comparison): Same.
6573 * profile-count.c (slow_safe_scale_64bit): Same.
6574 * simplify-rtx.c (simplify_const_binary_operation): Same.
6575 * tree-chrec.c (tree_fold_binomial): Same.
6576 * tree-data-ref.c (split_constant_offset_1): Same.
6577 * tree-if-conv.c (idx_within_array_bound): Same.
6578 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
6579 * tree-ssa-phiopt.c (minmax_replacement): Same.
6580 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
6581 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
6582 * vr-values.c (vr_values::adjust_range_with_scev): Same.
6583 * wide-int.cc (wi::add_large): Same.
6584 (wi::mul_internal): Same.
6585 (wi::sub_large): Same.
6586 (wi::divmod_internal): Same.
6587 * wide-int.h: Change overflow type to overflow_type for neg, add,
6588 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
6589 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
6590 mul_internal, divmod_internal.
6591 (overflow_type): New enum.
6592 (accumulate_overflow): New.
6594 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
6596 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
6597 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
6599 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6601 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
6602 argument is checked for zero before entering loop, avoid checking again.
6604 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
6606 * gimplify.h (generic_expr_could_trap_p): Set as global function.
6607 * gimplify.h (generic_expr_could_trap_p): Likwise.
6608 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
6610 2018-07-06 Jakub Jelinek <jakub@redhat.com>
6612 PR tree-optimization/86401
6613 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
6614 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
6615 (fold_bit_and_mask): ... here. New helper function for match.pd.
6616 * fold-const.h (fold_bit_and_mask): Declare.
6617 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
6619 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
6622 * target.def (translate_mode_attribute): New hook.
6623 * targhooks.h (default_translate_mode_attribute): Declare.
6624 * targhooks.c (default_translate_mode_attribute): New function.
6625 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
6626 * doc/tm.texi: Regenerate.
6627 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
6628 (rs6000_translate_mode_attribute): New function.
6630 2018-07-06 Paul Koning <ni1d@arrl.net>
6632 * doc/md.texi (define_split): Document DONE and FAIL.
6633 (define_peephole2): Ditto.
6635 2018-07-05 Jeff Law <law@redhat.com>
6637 PR tree-optimization/86010
6638 * tree-ssa-dse.c (compute_trims): More aggressively trim at
6639 both the head and tail of mem* and str* calls.
6641 2018-07-05 Jim Wilson <jimw@sifive.com>
6643 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
6645 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
6647 * config/aarch64/aarch64-simd.md: correct flags text for
6648 MIN_EXPR replacement.
6650 2018-07-05 James Clarke <jrtc27@jrtc27.com>
6652 * configure: Regenerated.
6654 2018-07-05 Carl Love <cel@us.ibm.com>
6656 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
6657 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
6658 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
6659 VSX_BUILTIN_DOUBLEL_V4SF.
6661 2018-07-05 Martin Sebor <msebor@redhat.com>
6664 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
6665 than its domain to compute its the upper bound of a char array.
6667 2018-07-05 Nathan Sidwell <nathan@acm.org>
6669 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
6670 * doc/cpp.texi: Update comment.
6671 * doc/tm.texi: Rebuilt.
6672 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
6673 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
6674 * doc/extend.texi (Backwards Compatibility): Clarify it is system
6675 headers affected by extern "C".
6676 * system.h: Poison NO_IMPLICIT_EXTERN_C.
6677 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
6678 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
6679 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
6680 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
6681 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
6682 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
6683 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
6684 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
6685 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
6686 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
6687 config/sparc/sp64-elf.h, config/spu/spu.h,
6688 config/stormy16/stormy16.h, config/v850/v850.h,
6689 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
6690 define NO_IMPLICIT_EXTERN_C.
6691 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
6693 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6696 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6697 instead of GET_MODE_SIZE when comparing Units.
6699 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6702 * rtlanal.c (set_noop_p): Constrain on mode change,
6703 include hard-reg-set.h
6705 2018-07-05 Tamar Christina <tamar.christina@arm.com>
6707 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
6709 2018-07-05 Jakub Jelinek <jakub@redhat.com>
6712 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
6715 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6717 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6719 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
6722 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6724 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6726 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
6729 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
6731 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
6733 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
6734 neg_*_op* variables.
6736 2018-07-04 Martin Liska <mliska@suse.cz>
6738 * tree-switch-conversion.c: Define
6739 max_ratio_for_speed and max_ratio_for_size constants.
6741 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
6742 Martin Liska <mliska@suse.cz>
6747 * common.opt: Rename align options with 'str_' prefix.
6748 * common/config/i386/i386-common.c (set_malign_value): New
6750 (ix86_handle_option): Use it to set -falign-* options/
6751 * config/aarch64/aarch64-protos.h (struct tune_params): Change
6752 type from int to string.
6753 * config/aarch64/aarch64.c: Update default values from int
6755 * config/alpha/alpha.c (alpha_override_options_after_change):
6757 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
6758 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6759 max skip conditionally.
6760 * config/i386/freebsd.h (SUBALIGN_LOG): New.
6761 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6762 max skip conditionally.
6763 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6764 max skip conditionally.
6765 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
6766 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6767 max skip conditionally.
6768 * config/i386/i386.c (struct ptt): Change type from int to
6770 (ix86_default_align): Set default values.
6771 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
6772 max skip conditionally.
6773 * config/i386/iamcu.h (SUBALIGN_LOG): New.
6774 (ASM_OUTPUT_MAX_SKIP_ALIGN):
6775 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
6776 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6777 max skip conditionally.
6778 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
6779 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
6780 * config/i386/x86-64.h (SUBALIGN_LOG): New.
6781 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6782 max skip conditionally.
6783 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6784 * config/ia64/ia64.c (ia64_option_override): Set default values
6785 for alignment options.
6786 * config/m68k/m68k.c: Handle new str_align_* options.
6787 * config/mips/mips.c (mips_set_compression_mode): Change
6789 (mips_option_override): Set default values for options.
6790 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
6792 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6794 * config/rx/rx.c (rx_option_override): Likewise.
6795 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
6796 (LABEL_ALIGN): Use align_labels_log.
6797 (LOOP_ALIGN): Use align_loops_align.
6798 * config/s390/s390.c (s390_asm_output_function_label): Use new
6800 * config/sh/sh.c (sh_override_options_after_change):
6801 Change type of constants.
6802 * config/spu/spu.c (spu_sched_init): Likewise.
6803 * config/sparc/sparc.c (sparc_option_override): Set default
6805 * config/visium/visium.c (visium_option_override): Likewise.
6806 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
6807 emit p2align format with last argument if it's not needed.
6808 * doc/invoke.texi: Document extended format of -falign-*.
6809 * final.c: Use align_labels alignment.
6810 * flags.h (struct target_flag_state): Change type to use
6812 (struct align_flags_tuple): New.
6813 (struct align_flags): Likewise.
6814 (align_loops_log): Redefine macro to use new types.
6815 (align_loops_max_skip): Redefine macro to use new types.
6816 (align_jumps_log): Redefine macro to use new types.
6817 (align_jumps_max_skip): Redefine macro to use new types.
6818 (align_labels_log): Redefine macro to use new types.
6819 (align_labels_max_skip): Redefine macro to use new types.
6820 (align_functions_log): Redefine macro to use new types.
6821 (align_loops): Redefine macro to use new types.
6822 (align_jumps): Redefine macro to use new types.
6823 (align_labels): Redefine macro to use new types.
6824 (align_functions): Redefine macro to use new types.
6825 (align_functions_max_skip): Redefine macro to use new types.
6826 (align_loops_value): New macro.
6827 (align_jumps_value): New macro.
6828 (align_labels_value): New macro.
6829 (align_functions_value): New macro.
6830 * function.c (invoke_set_current_function_hook): Propagate
6831 alignment values from flags to global variables default in
6833 * ipa-icf.c (sem_function::equals_wpa): Use
6834 cl_optimization_option_eq instead of memcmp.
6835 * lto-streamer.h (cl_optimization_stream_out): Support streaming
6837 (cl_optimization_stream_in): Likewise.
6838 * optc-save-gen.awk: Support strings in cl_optimization.
6839 * opth-gen.awk: Likewise.
6840 * opts.c (finish_options): Remove error checking of invalid
6842 (MAX_CODE_ALIGN): Remove.
6843 (MAX_CODE_ALIGN_VALUE): Likewise.
6844 (parse_and_check_align_values): New function.
6845 (check_alignment_argument): Likewise.
6846 (common_handle_option): Use check_alignment_argument.
6847 * opts.h (parse_and_check_align_values): Declare.
6848 * toplev.c (init_alignments): Remove.
6849 (read_log_maxskip): New.
6850 (parse_N_M): Likewise.
6851 (parse_alignment_opts): Likewise.
6852 (backend_init_target): Remove usage of init_alignments.
6853 * toplev.h (parse_alignment_opts): Declare.
6854 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
6856 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
6857 * tree.c (cl_option_hasher::equal): New.
6858 * varasm.c: Use new global macros.
6860 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
6862 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
6863 Use a simpler align directive also if MAXSKIP = ALIGN-1.
6864 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6865 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6866 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6867 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6868 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
6869 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
6870 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
6871 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6872 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6873 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6874 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6876 2018-07-04 Martin Liska <mliska@suse.cz>
6877 Jonathan Wakely <jwakely@redhat.com>
6879 * coverage.c: Use correct type.
6880 * doc/invoke.texi: Language correction.
6882 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
6885 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
6886 ENDBRANCH for non-tail call which may return via indirect branch.
6887 * doc/extend.texi: Document indirect_return attribute.
6889 2018-07-03 Martin Sebor <msebor@redhat.com>
6891 PR tree-optimization/86274
6892 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
6894 (format_floating): Correct handling of infinities and NaNs.
6896 2018-07-03 Martin Sebor <msebor@redhat.com>
6898 * print-tree.c (print_real_cst): New function.
6899 (print_node_brief): Call it.
6900 (print_node): Ditto.
6902 2018-07-03 Jeff Law <law@redhat.com>
6904 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
6905 into a single pattern.
6907 * config/h8300/h8300.md (ors code_iterator): New.
6908 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
6909 a single pattern and single splitter.
6910 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
6911 (iorqi3_1, xorqi3_1): Likewise.
6912 (iorqi3, xorqi3 expanders): Similarly.
6914 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
6915 (movmd_internal) into a single pattern using the P mode iterator.
6916 (movmd splitters): Similarly.
6917 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
6918 (movsd splitters): Similarly.
6920 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
6921 ADDB, ADDW and ADDL into a single ADD attribute which selects the
6922 right table based on the size of the operand.
6923 * config/h8300/h8300.md (length_table): Corresponding changes. All
6924 references to "addb", "addw" and "addl" changed to "add".
6925 (btst patterns): Merge two variants into a single pattern.
6926 (tstqi, tsthi): Likewise.
6927 (addhi3_incdec, addsi3_incdec): Likewise.
6928 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
6929 (mulhi3, mulsi3): Likewise.
6930 (udivhi3, udivsi3): Likewise.
6931 (divhi3, divsi3): Likewise.
6932 (andorqi3, andorhi3, andorsi3): Likewise.
6934 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
6937 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
6938 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
6940 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6942 PR tree-optimization/85694
6943 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
6944 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
6946 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
6947 (<u>avg<mode>3_ceil): New patterns.
6949 2018-07-03 David Malcolm <dmalcolm@redhat.com>
6951 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
6952 scan-tree-dump directive.
6953 * gcc.dg/vect/slp-perm-2.c: Likewise.
6954 * gcc.dg/vect/slp-perm-3.c: Likewise.
6955 * gcc.dg/vect/slp-perm-5.c: Likewise.
6956 * gcc.dg/vect/slp-perm-6.c: Likewise.
6957 * gcc.dg/vect/slp-perm-7.c: Likewise.
6958 * gcc.dg/vect/slp-perm-8.c: Likewise.
6960 2018-07-03 Marek Polacek <polacek@redhat.com>
6963 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
6965 2018-07-03 Richard Biener <rguenther@suse.de>
6968 * tree-ssa-structalias.c (find_func_clobbers): Properly
6969 handle indirect calls.
6971 2018-07-03 Jeff Law <law@redhat.com>
6973 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
6974 (shifts): New code iterator.
6975 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
6976 expander. Fix HImode handling on H8/SX.
6977 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
6978 (subqi3, subhi3, subsi3 expanders): Likewise.
6979 (andqi3, andhi3, andsi3 expanders): Likewise.
6980 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
6981 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
6982 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
6983 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
6984 (zero_extendqihi2, zero_extendqisi2): Likewise.
6985 (extendqihi2, extendqisi2): Likewise.
6986 (rotlqi3, rotlhi3, rotlsi3): Likewise.
6987 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
6988 (rotlqi3_1, rotlhi3_1): Likewise.
6989 (logicalhi3_sn, logicalsi3_sn): Likewise.
6990 (logicalhi3, logicalsi3): Likewise.
6992 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
6994 * tree-vect-patterns.c (vect_recog_rotate_pattern)
6995 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
6996 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
6997 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
6998 type to append_pattern_def_seq instead of creating a stmt_vec_info
7000 (build_mask_conversion): Likewise. Remove vinfo argument.
7001 (vect_add_conversion_to_patterm): Likewise, renaming to...
7002 (vect_add_conversion_to_pattern): ...this.
7003 (vect_recog_mask_conversion_pattern): Update call to
7004 build_mask_conversion. Pass the vector type to
7005 append_pattern_def_seq here too.
7006 (vect_recog_gather_scatter_pattern): Update call to
7007 vect_add_conversion_to_pattern.
7009 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7011 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
7012 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7013 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
7014 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
7015 STMT_VINFO_PATTERN_DEF_SEQ to null here.
7016 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
7017 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
7018 append_pattern_def_seq instead of new_pattern_def_seq.
7019 (vect_recog_divmod_pattern): Do both of the above.
7020 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
7023 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7025 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
7026 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
7027 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
7028 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
7029 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
7030 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7031 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
7032 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7033 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7034 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
7035 parameter with a single stmt_vec_info.
7036 (vect_recog_func_ptr): Likewise.
7037 (vect_recog_gather_scatter_pattern): Likewise, folding in...
7038 (vect_try_gather_scatter_pattern): ...this.
7039 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
7040 the stmt_vec_info of the statement to be matched. Don't clear
7041 STMT_VINFO_RELATED_STMT.
7042 (vect_pattern_recog): Update call accordingly.
7044 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7046 PR tree-optimization/85694
7047 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
7048 (uavgM3_ceil): Document new optabs.
7049 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
7050 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
7052 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
7053 (savg_ceil_optab): New optabs.
7054 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
7055 (vect_vect_recog_func_ptrs): Add it.
7056 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
7057 constant directly from the associated lhs.
7059 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7061 * tree-vect-patterns.c (vect_split_statement): New function.
7062 (vect_convert_input): Use it to try to split an existing cast.
7064 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7066 * poly-int.h (print_hex): New function.
7067 * dumpfile.h (dump_dec, dump_hex): Declare.
7068 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
7069 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
7070 min_input_precision, operation_precision and operation_sign.
7071 * tree-vect-patterns.c (vect_get_range_info): New function.
7072 (vect_same_loop_or_bb_p, vect_single_imm_use)
7073 (vect_operation_fits_smaller_type): Delete.
7074 (vect_look_through_possible_promotion): Add an optional
7075 single_use_p parameter.
7076 (vect_recog_over_widening_pattern): Rewrite to use new
7077 stmt_vec_info infomration. Handle one operation at a time.
7078 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
7079 (vect_truncatable_operation_p, vect_set_operation_type)
7080 (vect_set_min_input_precision): New functions.
7081 (vect_determine_min_output_precision_1): Likewise.
7082 (vect_determine_min_output_precision): Likewise.
7083 (vect_determine_precisions_from_range): Likewise.
7084 (vect_determine_precisions_from_users): Likewise.
7085 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
7086 (vect_vect_recog_func_ptrs): Put over_widening first.
7088 (vect_pattern_recog): Call vect_determine_precisions.
7090 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
7092 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
7093 statements that have been replaced by further pattern statements.
7094 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
7096 2018-07-03 Richard Biener <rguenther@suse.de>
7098 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
7099 always set *dt. Dump vectype in vectype overload.
7100 * dumpfile.h (dump_gimple_expr): New function.
7101 (dump_gimple_expr_loc): Likewise.
7102 * dumpfile.c (dump_gimple_expr): New function.
7103 (dump_gimple_expr_loc): Likewise.
7105 2018-07-02 Jeff Law <law@redhat.com>
7107 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
7108 the H8/300, H8/300H and H8/S variants into a single pattern.
7109 (movhi_h8300, movqi_h8300hs): Similarly.
7110 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
7111 (QHI mode iterator): New.
7113 * config/h8300/h8300.md: Remove trailing whitespace.
7115 2018-07-02 Jim Wilson <jimw@sifive.com>
7117 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
7118 instead of emit_insn for interrupt returns.
7119 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
7120 (riscv_sret, riscv_uret): Likewise.
7122 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7124 * pretty-print.c (selftest::test_pp_format): Move save and restore
7125 of quotes to class auto_fix_quotes, and add an instance.
7126 * selftest.c: Include "intl.h".
7127 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
7128 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
7129 * selftest.h (selftest::auto_fix_quotes): New class.
7131 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
7133 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
7134 (aarch64_sve_prepare_conditional_op): Remove.
7135 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
7136 Allow aarch64_simd_reg_or_zero as select operand; remove
7137 the aarch64_sve_prepare_conditional_op call.
7138 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
7139 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7140 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
7141 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
7142 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
7143 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
7144 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
7145 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
7146 and a splitters to match all of the *_any patterns.
7147 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
7149 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
7150 (SVE_COND_FP_BINARY_REV): Remove.
7151 (sve_int_op_rev, sve_fp_op_rev): New.
7152 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
7153 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
7154 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
7155 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
7156 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
7157 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
7158 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
7159 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
7160 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
7162 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
7163 Remove match_dup 1 from the inner unspec.
7164 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7166 * config/aarch64/aarch64.md (movprfx): New attr.
7167 (length): Default movprfx to 8.
7168 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
7169 (*madd<SVE_I>, *msub<SVE_I): Likewise.
7170 (*<su>mul<SVE_I>3_highpart): Likewise.
7171 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
7172 (*v<ASHIFT><SVE_I>3): Likewise.
7173 (*<su><MAXMIN><SVE_I>3): Likewise.
7174 (*<su><MAXMIN><SVE_F>3): Likewise.
7175 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
7176 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
7177 (*div<SVE_F>4): Likewise.
7179 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
7181 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
7184 2018-07-02 Richard Biener <rguenther@suse.de>
7186 PR tree-optimization/86363
7187 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
7188 memset argument refers to a non-variable address.
7190 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7192 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
7193 {PLUS,MINUS}_EXPR code to...
7194 (adjust_symbolic_bound): ...here,
7195 (combine_bound): ...here,
7196 (set_value_range_with_overflow): ...and here.
7198 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
7200 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
7202 (extract_range_from_abs_expr): ...here.
7204 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
7206 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
7207 -fno-omit-frame-pointer when not optimizing.
7209 2018-07-02 Martin Liska <mliska@suse.cz>
7212 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
7213 (propagate_nothrow): Likewise.
7215 2018-07-02 Martin Liska <mliska@suse.cz>
7218 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
7220 2018-07-02 David Malcolm <dmalcolm@redhat.com>
7222 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
7223 function in r262149, changing "loc" param from source_location to
7224 const dump_location_t &.
7225 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
7226 declaration, as above.
7228 2018-07-01 Paul Koning <ni1d@arrl.net>
7230 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7231 -munit-asm, -mgnu-asm, -mdec-asm.
7232 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
7233 (pdp11_output_labelref): New.
7234 (pdp11_output_def): New.
7235 (pdp11_output_addr_vec_elt): New.
7236 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
7237 %# and %@ format codes.
7238 (pdp11_option_override): New.
7239 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
7240 (pdp11_output_ident): New.
7241 (pdp11_asm_named_section): New.
7242 (pdp11_asm_init_sections): New.
7243 (pdp11_file_start): New.
7244 (pdp11_file_end): New.
7245 (output_ascii): Use .ascii/.asciz for -mdec-asm.
7246 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
7247 %o, like %c but octal.
7248 (pdp11_option_override): New.
7249 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
7251 (DATA_SECTION_ASM_OP): Ditto.
7252 (READONLY_DATA_SECTION_ASM_OP): New.
7253 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
7254 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
7255 (ASM_OUTPUT_LABELREF): Ditto.
7256 (ASM_OUTPUT_DEF): Ditto.
7257 (ASM_OUTPUT_EXTERNAL): New.
7258 (ASM_OUTPUT_SOURCE_FILENAME): New.
7259 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
7260 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
7261 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
7262 %# and %@ format codes.
7263 * config/pdp11/pdp11.opt (mgnu-asm): New.
7264 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
7265 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
7266 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
7268 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
7270 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
7271 dereferencing path[] beyond its length.
7273 (debug_all_paths): New.
7274 (rewire_first_differing_edge): New.
7275 (adjust_paths_after_duplication): New.
7276 (duplicate_thread_path): Call adjust_paths_after_duplication.
7278 (thread_through_all_blocks): Add new argument to
7279 duplicate_thread_path.
7281 2018-06-30 Jim Wilson <jimw@sifive.com>
7283 * config/riscv/predicates.md (p2m1_shift_operand): New.
7284 (high_mask_shift_operand): New.
7285 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
7286 pattern using p2m1_shift_operand.
7287 (lshsi3_zero_extend_3+2): New combiner pattern using
7288 high_mask_shift_operand.
7290 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7292 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
7294 (vect_recog_rotate_pattern): ...here.
7295 (vect_convert_input): Try to insert casts of invariants in the
7297 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
7298 preheader to be empty.
7300 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7302 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
7303 vector type. If given, install it in the new statement's
7305 (vect_element_precision): New function.
7306 (vect_unpromoted_value): New struct.
7307 (vect_unpromoted_value::vect_unpromoted_value): New function.
7308 (vect_unpromoted_value::set_op): Likewise.
7309 (vect_look_through_possible_promotion): Likewise.
7310 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
7311 (vect_widened_op_tree, vect_convert_input): Likewise.
7312 (vect_convert_inputs, vect_convert_output): Likewise.
7313 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
7314 to handle the optional cast of the multiplication result and
7315 vect_widened_op_tree to detect the widened multiplication itself.
7316 Do not require the input and output of promotion casts to have
7317 the same sign, but base the signedness of the operation on the
7318 input rather than the result. If the pattern includes two
7319 promotions, check that those promotions have the same sign.
7320 Do not restrict the MULT_EXPR handling to a double-width result;
7321 handle quadruple-width results and wider. Use vect_convert_inputs
7322 to convert the inputs to the common type.
7323 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
7324 to handle the optional cast of the ABS result. Also allow a sign
7325 change or a sign extension between the ABS and MINUS.
7326 Use vect_widened_op_tree to detect the widened subtraction and use
7327 vect_convert_inputs to convert the inputs to the common type.
7328 (vect_handle_widen_op_by_const): Delete.
7329 (vect_recog_widen_op_pattern): New function.
7330 (vect_recog_widen_mult_pattern): Use it.
7331 (vect_recog_widen_shift_pattern): Likewise.
7332 (vect_recog_widen_sum_pattern): Use
7333 vect_look_through_possible_promotion to handle the promoted
7336 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7338 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
7339 the containing gimple_seq *.
7340 * gimple-iterator.h (gsi_for_stmt): Declare it.
7341 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
7342 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
7343 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7344 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7345 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
7347 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
7349 (vect_mark_pattern_stmts): ...here. Handle cases in which the
7350 statement being replaced is part of an existing pattern
7351 definition sequence, inserting the new pattern statements before
7353 (vect_pattern_recog_1): Don't return a bool. If the statement
7354 is already part of a pattern, instead apply pattern matching
7355 to the pattern definition statements. Don't clear the
7356 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
7357 (vect_pattern_recog): Don't break after the first match;
7358 continue processing the pattern definition statements instead.
7359 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
7361 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7363 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7364 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7365 (vect_recog_widen_sum_pattern): Use it.
7367 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7369 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7370 phi is not a pattern statement and has not been replaced by
7371 a pattern statement.
7372 * tree-vect-patterns.c (type_conversion_p): Don't check
7373 STMT_VINFO_IN_PATTERN_P.
7374 (vect_recog_vector_vector_shift_pattern): Likewise.
7375 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
7376 the pattern statement rather than the original statement; check
7377 directly for a WIDEN_MULT_EXPR here.
7378 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
7379 vect_is_simple_use to return the pattern statement rather
7380 than the original statement; use is_pattern_stmt_p to check
7381 for such a pattern statement.
7382 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
7383 to return the pattern statement rather than the original statement;
7384 don't do the same transformation here.
7385 (vect_is_simple_use): If the defining statement has been replaced
7386 by a pattern statement, return the pattern statement instead.
7387 Remove the corresponding (local) transformation from the vectype
7390 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
7392 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
7393 end and default to null.
7394 * tree-vect-loop.c (vect_create_epilog_for_reduction)
7395 (vectorizable_reduction): Update calls accordingly, dropping the
7396 gimple ** argument if the passed-back statement isn't needed.
7397 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
7398 (vect_recog_rotate_pattern): Likewise.
7399 (vect_recog_mask_conversion_pattern): Likewise.
7400 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7401 (vect_mask_constant_operand_p): Likewise.
7402 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
7403 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
7404 (get_group_load_store_type, get_load_store_type): Likewise.
7405 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
7406 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
7407 (vectorizable_conversion, vectorizable_assignment): Likewise.
7408 (vectorizable_shift, vectorizable_operation): Likewise.
7409 (vectorizable_store, vect_is_simple_cond): Likewise.
7410 (vectorizable_condition, vectorizable_comparison): Likewise.
7411 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
7412 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
7413 and move it to the end. Cope with null def_stmt_outs.
7415 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7417 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
7419 2018-06-29 Jeff Law <law@redhat.com>
7421 * config/v850/v850.c (v850_legitimate_address_p): Handle large
7422 displacements for TARGET_V850E2V3 and newer.
7423 (TARGET_LRA_P): Remove. Defaults to LRA now.
7424 * config/v850/v850.md (sign23byte_load): Remove.
7425 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
7426 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
7428 2018-06-29 Martin Liska <mliska@suse.cz>
7431 * coverage.c (coverage_init): Mangle full path name.
7432 * doc/invoke.texi: Document the change.
7433 * gcov-io.c (mangle_path): New.
7434 * gcov-io.h (mangle_path): Likewise.
7435 * gcov.c (mangle_name): Use mangle_path for path mangling.
7437 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7439 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
7440 if starting source register is not even.
7442 2018-06-29 Martin Liska <mliska@suse.cz>
7444 PR tree-optimization/86263
7445 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
7446 Make edge redirection.
7448 2018-06-29 David Malcolm <dmalcolm@redhat.com>
7450 * dumpfile.c (dump_loc): Add indentation based on scope depth.
7451 (dump_scope_depth): New variable.
7452 (get_dump_scope_depth): New function.
7453 (dump_begin_scope): New function.
7454 (dump_end_scope): New function.
7455 * dumpfile.h (get_dump_scope_depth): New declaration.
7456 (dump_begin_scope): New declaration.
7457 (dump_end_scope): New declaration.
7458 (class auto_dump_scope): New class.
7459 (AUTO_DUMP_SCOPE): New macro.
7460 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
7463 2018-06-29 Richard Biener <rguenther@suse.de>
7465 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
7466 compute_all_dependences succeeds.
7467 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
7468 exceed --param loop-max-datarefs-for-datadeps.
7470 2018-06-29 Jakub Jelinek <jakub@redhat.com>
7472 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
7474 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
7477 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
7478 alternative 0 in preferred_for_speed attribute.
7480 2018-06-28 Paul Koning <ni1d@arrl.net>
7482 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
7483 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
7484 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
7485 * config/pdp11/pdp11.md: Correct "length" attribute calculation
7486 for shift insn patterns.
7488 2018-06-28 David Malcolm <dmalcolm@redhat.com>
7490 * cgraph.c (cgraph_node::get_body): Replace assignments to
7491 "dump_file" with calls to set_dump_file.
7492 * dumpfile.c (alt_dump_file): Make static, and group with...
7493 (alt_flags): ...this definition.
7494 (dumps_are_enabled): New variable.
7495 (refresh_dumps_are_enabled): New function.
7496 (set_dump_file): New function.
7497 (set_alt_dump_file): New function.
7498 (gcc::dump_manager::dump_start): Replace assignments to
7499 "dump_file" and "alt_dump_file" with calls to set_dump_file and
7501 (gcc::dump_manager::dump_finish): Likewise.
7502 * dumpfile.h (alt_dump_file): Delete decl.
7503 (dumps_are_enabled): New variable decl.
7504 (set_dump_file): New function decl.
7505 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
7507 * tree-nested.c (lower_nested_functions): Replace assignments to
7508 "dump_file" with calls to set_dump_file.
7510 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7512 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
7513 goto_locus of each outgoing edge of each basic block.
7515 2018-06-28 Richard Biener <rguenther@suse.de>
7517 * dwarf2out.c (decl_scope_table): Remove.
7518 (push_decl_scope): Likewise.
7519 (pop_decl_scope): Likewise.
7520 (gen_type_die_for_member): Do not call push/pop_decl_scope.
7521 (gen_struct_or_union_type_die): Likewise.
7522 (gen_tagged_type_die): Likewise.
7523 (dwarf2out_init): Do not initialize decl_scope_table.
7524 (dwarf2out_c_finalize): Do not free it.
7526 2018-06-28 Richard Biener <rguenther@suse.de>
7528 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
7529 deciding whether to not re-use a DIE.
7531 2018-06-28 Richard Biener <rguenther@suse.de>
7533 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
7534 DW_AT_abstract_origin attribute.
7536 2018-06-28 Martin Liska <mliska@suse.cz>
7538 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7539 Use newly introduced constants.
7540 * tree-switch-conversion.h (struct jump_table_cluster):
7541 Define max_ratio_for_size and max_ratio_for_speed.
7543 2018-06-28 Martin Liska <mliska@suse.cz>
7545 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7546 Add new checking assert to catch invalid state.
7547 (jump_table_cluster::can_be_handled): Handle single case
7549 (jump_table_cluster::is_beneficial): Bail out for such case.
7550 (bit_test_cluster::find_bit_tests):
7551 Add new checking assert to catch invalid state.
7552 (bit_test_cluster::can_be_handled): Handle single case
7554 (bit_test_cluster::is_beneficial): Bail out for such case.
7555 (switch_decision_tree::analyze_switch_statement):
7558 2018-06-28 Martin Liska <mliska@suse.cz>
7560 * common.opt: Introduce -completion option.
7561 * gcc.c (driver_handle_option): Handle it.
7562 (driver::main): Print completions if completion
7564 * opt-suggestions.c (option_proposer::get_completions):
7566 (option_proposer::suggest_completion): Likewise.
7567 (option_proposer::find_param_completions): Likewise.
7568 (verify_autocompletions): Likewise.
7569 (test_completion_valid_options): Likewise.
7570 (test_completion_valid_params): Likewise.
7571 (in_completion_p): Likewise.
7572 (empty_completion_p): Likewise.
7573 (test_completion_partial_match): Likewise.
7574 (test_completion_garbage): Likewise.
7575 (opt_proposer_c_tests): Likewise.
7576 * opt-suggestions.h: Declare new functions.
7577 * opts.c (common_handle_option): Handle OPT__completion_.
7578 * selftest-run-tests.c (selftest::run_tests): Add
7579 opt_proposer_c_tests.
7580 * selftest.c (assert_str_startswith): New.
7581 * selftest.h (assert_str_startswith): Likewise.
7582 (opt_proposer_c_tests): New.
7583 (ASSERT_STR_STARTSWITH): Likewise.
7585 2018-06-28 Martin Liska <mliska@suse.cz>
7587 * Makefile.in: Add opt-suggestions.o.
7588 * gcc-main.c: Include opt-suggestions.h.
7589 * gcc.c (driver::driver): Likewise.
7590 (driver::~driver): Remove m_option_suggestions.
7591 (driver::build_option_suggestions): Moved to option_proposer.
7592 (driver::suggest_option): Likewise.
7593 (driver::handle_unrecognized_options): Use option_proposer.
7594 * gcc.h (class driver): Add new memver m_option_proposer.
7595 * opt-suggestions.c: New file.
7596 * opt-suggestions.h: New file.
7598 2018-06-28 Martin Liska <mliska@suse.cz>
7600 * vec.h (class auto_string_vec): New (moved from auto_argvec).
7601 (auto_string_vec::~auto_string_vec): Likewise.
7603 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
7605 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
7606 prevent_decl_creation_for_types fields up and add reset_location field.
7607 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
7608 statement if id->reset_location is true.
7609 (copy_edges_for_bb): Do not set goto_locus on the new edges if
7610 id->reset_location is true.
7611 (copy_phis_for_bb): Force input_location on the arguments if
7612 id->reset_location is true.
7613 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
7614 is set on the function to be inlined.
7616 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
7618 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
7620 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
7622 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
7623 registers for Pmode.
7624 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
7625 hard registers for the clobbered pseudo.
7627 2018-06-27 Paul Koning <ni1d@arrl.net>
7629 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7630 mutually exclusive options.
7631 * config/pdp11/constraints.md (h): New constraint.
7632 (O): Update definition to match shift code generation.
7633 (D): New constraint.
7634 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
7636 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
7638 (output_jump): Change arguments.
7639 (pdp11_fixed_cc_regs): New function.
7640 (pdp11_cc_mode): Ditto.
7641 (pdp11_expand_shift): Ditto.
7642 (pdp11_assemble_shift): Ditto.
7643 (pdp11_small_shift): Ditto.
7644 (pdp11_branch_cost): Remove.
7645 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
7647 (pdp11_register_move_cost): Update for CC registers.
7648 (pdp11_rtx_costs): Add case for LSHIFTRT.
7649 (pdp11_output_jump): Add CCNZ mode conditional branches.
7650 (notice_update_cc_on_set): Remove.
7651 (pdp11_cc_mode): New function.
7652 (simple_memory_operand): Correct pre/post decrement case.
7653 (no_side_effect_operand): New function.
7654 (pdp11_regno_reg_class): Add CC_REGS class.
7655 (pdp11_fixed_cc_regs): New function.
7656 (pdp11_small_shift): New function.
7657 (pdp11_expand_shift): New function to expand shift insns.
7658 (pdp11_assemble_shift): New function to output shifts.
7659 (pdp11_branch_cost): Remove.
7660 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
7661 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
7662 (WCHAR_TYPE): Ditto.
7663 (PTRDIFF_TYPE): Ditto.
7664 (ADJUST_INSN_LENGTH): New macro.
7665 (FIXED_REGISTERS): Add CC registers.
7666 (CALL_USED_REGISTERS): Ditto.
7668 (REG_CLASS_NAMES): Ditto.
7669 (REG_CLASS_CONTENTS): Ditto.
7670 (SELECT_CC_MODE): Use new function.
7671 (TARGET_FLAGS_REGNUM): New macro.
7672 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
7673 (cc0_reg_rtx): Remove.
7674 (CC_STATUS_MDEP): Remove.
7675 (CC_STATUS_MDEFP_INIT): Remove.
7676 (CC_IN_FPU): Remove.
7677 (NOTICE_UPDATE_CC): Remove.
7678 (REGISTER_NAMES): Add CC registers.
7679 (BRANCH_COST): Change to constant 1.
7680 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
7682 * config/pdp11/pdp11.opt (mbcopy): Remove.
7683 (mbcopy-builtin): Remove.
7684 (mbranch-cheap): Remove.
7685 (mbranch-expensive): Remove.
7686 * config/pdp11/predicates.md (expand_shift_operand): Update to
7687 match shift code generation.
7688 (ccnz_operator): New predicate.
7689 * doc/invoke.texi (PDP-11 Options): Remove deleted options
7690 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
7691 Remove non-existent option -mabshi, -mno-abshi. Document mutually
7693 * doc/md.texi (PDP-11): Document new D and h constraints. Update
7694 description of O constraint.
7696 2018-06-27 Jeff Law <law@redhat.com>
7697 Austin Law <austinklaw@gmail.com>
7699 * config/v850/v850.md (addsi3_set_flags): New pattern.
7700 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
7701 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
7702 (zero_extendhisi2_v850_set_flags): Likewise.
7703 (zero_extendqisi2_v850_set_flags): Likewise.
7704 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
7705 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
7706 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
7708 * config/v850/v850-protos.h (notice_update_cc): Remove.
7709 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
7710 (v850_print_operand): Handle 'D' and "d".
7711 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
7712 Add handling of arithmetic/logical operations compared against zero.
7713 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
7714 Do not look at v850_compare_op, instead get mode from last argument.
7715 (v850_gen_compare): Remove
7716 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
7717 after reload for prologue insns.
7718 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
7720 (construct_save_jarl): Likewise.
7721 (TARGET_FLAGS_REGNUM): Define.
7722 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
7723 (NOTICE_UPDATE_CC): Remove.
7724 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
7725 than cc0. Conditionalize on reload_completed.
7726 (cmpsi_insn, setfcc_insn): Likewise.
7727 (tst1 splitter): Turn into define_and_split which sets the flags
7729 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
7730 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
7731 (cstoresf4, cstoredf4): Clobber the flags.
7732 (cmpsi, cmpsf, cmpdf): Remove expanders.
7733 (setf_insn): Remove pattern.
7734 (addsi3): Turn into define_and_split which clobbers the flags after
7735 reload and a suitable pattern (addsi3_clobber_flags) for use after
7737 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
7738 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
7739 (ashrsi3, ashrsi3_v850e2): Likewise.
7740 (bins): Clobber the flags.
7741 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
7742 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
7743 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
7744 (call_value_internal_short, call_value_internal_long): Likewise.
7745 (callt_save_interrupt, callt_return_interrupt): Likewise.
7746 (save_interrupt, return_interrupt): Likewise.
7747 (callt_save_all_interrupt, save_all_interrupt): Likewise.
7748 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
7749 (restore_all_interrupt, _restore_all_interrupt): Likewise.
7750 (All FP comparisons): Only allow after reload has completed.
7752 (divh, divhu): Tweak output template.
7753 (branch_z_normal, branch_z_invert): Remove
7754 (branch_nz_normal, branch_nz_invert): Likewise.
7755 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
7757 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
7758 * config/v850/v850.c (notice_update_cc): Remove.
7759 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
7760 (CC_NO_CARRY): Likewise.
7761 (NOTICE_UPDATE_CC): Define to nothing.
7762 * config/v850/v850.md: Remove block comment on cc0 handling
7763 Remove "cc" attribute from all patterns. Remove cc_status handling
7764 from all patterns. Minor formatting fixes.
7766 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7768 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
7769 (cortex-a76.cortex-a55): Likewise.
7770 * config/aarch64/aarch64-tune.md: Regenerate.
7771 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
7772 cortex-a76.cortex-a55.
7774 2018-06-27 Jeff Law <law@redhat.com>
7776 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
7777 (MULTILIB_DIRNAMES): Similarly.
7779 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
7781 * gimple.h (gimple_return_retbnd): Delete.
7782 (gimple_return_set_retbnd): Likewise.
7783 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
7784 gimple_return_set_retbnd.
7785 * gimple-pretty-print.c (dump_gimple_return): Remove call to
7786 gimple_return_retbnd and adjust.
7787 * tree-inline.h (struct copy_body_data): Remove retbnd field.
7788 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
7789 Explicitly return NULL in a couple more cases. Move assertion
7790 on debug statements and remove unreachable code.
7791 (reset_debug_binding): Do not test id->retbnd.
7792 (expand_call_inline): Do not set it.
7794 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
7796 * configure.ac: Add --disable-gcov option.
7797 * configure: Regenerate.
7798 * Makefile.in: Honour @enable_gcov@.
7799 * doc/install.texi: Document --disable-gcov.
7801 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7803 * config/arm/arm-cpus.in (cortex-a76): New entry.
7804 (cortex-a76.cortex-a55): Likewise.
7805 * config/arm/arm-tables.opt: Regenerate.
7806 * config/arm/arm-tune.md: Likewise.
7807 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
7808 * doc/invoke.texi (ARM Options): Document cortex-a76 and
7809 cortex-a76.cortex-a55.
7811 2018-06-27 Tamar Christina <tamar.christina@arm.com>
7814 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
7816 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
7818 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
7820 (EPILOGUE_USES): Likewise.
7822 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
7824 * tree-inline.c (remap_location): New function extracted from...
7825 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
7826 (copy_phis_for_bb): ...here. Call remap_location.
7827 (copy_cfg_body): Adjust call to copy_edges_for_bb.
7829 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
7831 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
7832 unaligned vsx for 16B memset.
7834 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
7837 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
7838 ieee128_float_type_node to long_double_type_node unless
7839 TARGET_LONG_DOUBLE_128 is set.
7841 2018-06-26 David Malcolm <dmalcolm@redhat.com>
7843 * cfgloop.c (get_loop_location): Convert return type from
7844 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
7845 by implicit construction from rtx_insn *, and using
7846 dump_user_location_t::from_function_decl for the fallback case.
7847 * cfgloop.h (get_loop_location): Convert return type from
7848 location_t to dump_user_location_t.
7849 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
7850 dump_printf_loc to pass in a dump_location_t rather than a
7851 location_t, via the gimple stmt.
7852 * coverage.c (get_coverage_counts): Update calls to
7853 dump_printf_loc to pass in dump_location_t rather than a
7855 * doc/optinfo.texi (Dump types): Convert example of
7856 dump_printf_loc from taking "locus" to taking "insn". Update
7857 description of the "_loc" calls to cover dump_location_t.
7858 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
7860 (dump_user_location_t::dump_user_location_t): New constructors,
7861 from gimple *stmt and rtx_insn *.
7862 (dump_user_location_t::from_function_decl): New function.
7863 (dump_loc): Make static.
7864 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
7865 const dump_location_t &.
7866 (dump_generic_expr_loc): Delete.
7867 (dump_printf_loc): Convert param "loc" from location_t to
7868 const dump_location_t &.
7869 (selftest::test_impl_location): New function.
7870 (selftest::dumpfile_c_tests): New function.
7871 * dumpfile.h: Include "profile-count.h".
7872 (class dump_user_location_t): New class.
7873 (struct dump_impl_location_t): New struct.
7874 (class dump_location_t): New class.
7875 (dump_printf_loc): Convert 2nd param from source_location to
7876 const dump_location_t &.
7877 (dump_generic_expr_loc): Delete.
7878 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
7879 const dump_location_t &.
7880 * gimple-fold.c (fold_gimple_assign): Update call to
7881 dump_printf_loc to pass in a dump_location_t rather than a
7882 location_t, via the gimple stmt.
7883 (gimple_fold_call): Likewise.
7884 * gimple-loop-interchange.cc
7885 (loop_cand::analyze_iloop_reduction_var): Update for change to
7886 check_reduction_path.
7887 (tree_loop_interchange::interchange): Update for change to
7889 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
7890 change in return-type of find_loop_location.
7891 (graphite_regenerate_ast_isl): Likewise.
7892 * graphite-optimize-isl.c (optimize_isl): Likewise.
7893 * graphite.c (graphite_transform_loops): Likewise.
7894 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
7895 pass in a dump_location_t rather than a location_t, via the
7897 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7898 * ipa.c (walk_polymorphic_call_targets): Likewise.
7899 * loop-unroll.c (report_unroll): Convert "locus" param from
7900 location_t to dump_location_t.
7901 (decide_unrolling): Update for change to get_loop_location's
7903 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
7904 location_t to dump_user_location_t.
7905 (grid_find_single_omp_among_assignments_1): Updates calls to
7906 dump_printf_loc to pass in a dump_location_t rather than a
7907 location_t, via the gimple stmt.
7908 (grid_parallel_clauses_gridifiable): Convert "tloc" from
7909 location_t to dump_location_t. Updates calls to dump_printf_loc
7910 to pass in a dump_location_t rather than a location_t, via the
7912 (grid_inner_loop_gridifiable_p): Likewise.
7913 (grid_dist_follows_simple_pattern): Likewise.
7914 (grid_gfor_follows_tiling_pattern): Likewise.
7915 (grid_target_follows_gridifiable_pattern): Likewise.
7916 (grid_attempt_target_gridification): Convert initialization
7917 of local "grid" from memset to zero-initialization; FIXME: does
7918 this require C++11? Update call to dump_printf_loc to pass in a
7919 optinfo_location rather than a location_t, via the gimple stmt.
7920 * profile.c (read_profile_edge_counts): Updates call to
7921 dump_printf_loc to pass in a dump_location_t rather than a
7923 (compute_branch_probabilities): Likewise.
7924 * selftest-run-tests.c (selftest::run_tests): Call
7926 * selftest.h (dumpfile_c_tests): New decl.
7927 * tree-loop-distribution.c (pass_loop_distribution::execute):
7928 Update for change in return type of find_loop_location.
7929 * tree-parloops.c (parallelize_loops): Likewise.
7930 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
7931 "locus" from location_t to dump_user_location_t.
7932 (canonicalize_loop_induction_variables): Likewise.
7933 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
7934 for change in return type of find_loop_location.
7935 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
7936 to dump_printf_loc to pass in a dump_location_t rather than a
7937 location_t, via the stmt.
7938 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
7940 * tree-vect-loop-manip.c (find_loop_location): Convert return
7941 type from source_location to dump_user_location_t.
7942 (vect_do_peeling): Update for above change.
7943 (vect_loop_versioning): Update for change in type of
7945 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
7946 from location_t to dump_user_location_t.
7947 (vect_estimate_min_profitable_iters): Update for change in type
7949 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
7950 location_t to dump_location_t.
7951 (vect_slp_bb): Update for change in type of vect_location.
7952 * tree-vectorizer.c (vect_location): Convert from source_location
7953 to dump_user_location_t.
7954 (try_vectorize_loop_1): Update for change in vect_location's type.
7955 (vectorize_loops): Likewise.
7956 (increase_alignment): Likewise.
7957 * tree-vectorizer.h (vect_location): Convert from source_location
7958 to dump_user_location_t.
7959 (find_loop_location): Convert return type from source_location to
7960 dump_user_location_t.
7961 (check_reduction_path): Convert 1st param from location_t to
7962 dump_user_location_t.
7963 * value-prof.c (check_counter): Update call to dump_printf_loc to
7964 pass in a dump_user_location_t rather than a location_t; update
7965 call to error_at for change in type of "locus".
7966 (check_ic_target): Update call to dump_printf_loc to
7967 pass in a dump_user_location_t rather than a location_t, via the
7970 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
7972 * config/s390/s390.h (enum processor_flags): Do not use
7973 default tune parameter when -march was specified.
7975 2018-06-26 Jakub Jelinek <jakub@redhat.com>
7978 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
7979 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
7982 2018-06-26 Richard Biener <rguenther@suse.de>
7984 PR tree-optimization/86287
7986 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
7987 (vect_analyze_loop): Initialize n_stmts.
7989 2018-06-26 Richard Biener <rguenther@suse.de>
7992 * fold-const.c (fold_convertible_p): Pointer extension
7995 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
7998 * dwarf2out.c (loc_list_has_views): Adjust comments.
7999 (dw_loc_list): Split single cross-partition range with
8002 2018-06-25 Jeff Law <law@redhat.com>
8004 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
8005 on -mbig-switch by default.
8007 * config/v850/predicates.md (const_float_1_operand): Fix match_code
8009 (const_float_0_operand): Remove unused predicate.
8010 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
8011 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
8012 (recipsf2): New expander. Original pattern now called
8014 (recipdf2, recipdf2_insn): Similarly.
8015 (rsqrtsf2, rsqrtsf2_insn): Similarly
8016 (rsqrtdf2, rsqrtdf2_insn): Similarly
8018 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
8020 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
8021 Simplify logic for FreeBSD (twice).
8023 2018-06-25 Martin Sebor <msebor@redhat.com>
8025 PR tree-optimization/86204
8026 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
8027 a strnlen result if it's less than the length of the string.
8029 2018-06-25 Martin Sebor <msebor@redhat.com>
8031 PR tree-optimization/85700
8032 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
8033 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
8034 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
8036 2018-06-25 Martin Sebor <msebor@redhat.com>
8038 * doc/extend.texi (Zero-length arrays): Update and clarify.
8040 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
8042 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
8043 added IEEE/IBM long double multilib support on PowerPC little
8044 endian Linux systems.
8045 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
8046 (MULTILIB_DEFAULTS): Likewise.
8047 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8049 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
8050 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
8051 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
8053 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
8056 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
8057 (REORDER_45): Likewise.
8059 2018-06-25 Jeff Law <law@redhat.com>
8061 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
8062 dividend to 32 bits. Adjust length.
8063 (udivmodhi4): Cleanup output template. Fix length.
8065 2018-06-25 Carl Love <cel@us.ibm.com>
8067 * config/rs6000/vsx.md: Change word selector to prefered location.
8069 2018-06-25 Richard Biener <rguenther@suse.de>
8071 PR tree-optimization/86304
8072 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
8073 epilogue-if-converted loops as well.
8075 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
8077 * lto-section-out.c (lto_begin_section): Do not print section
8078 name for noaddr and unnumbered dumps.
8080 2018-06-25 Richard Biener <rguenther@suse.de>
8082 * tree-vectorizer.h (struct vec_info_shared): New structure
8083 with parts split out from struct vec_info and loop_nest from
8084 struct _loop_vec_info.
8085 (struct vec_info): Adjust accordingly.
8086 (struct _loop_vec_info): Likewise.
8087 (LOOP_VINFO_LOOP_NEST): Adjust.
8088 (LOOP_VINFO_DATAREFS): Likewise.
8089 (LOOP_VINFO_DDRS): Likewise.
8090 (struct _bb_vec_info): Likewise.
8091 (BB_VINFO_DATAREFS): Likewise.
8092 (BB_VINFO_DDRS): Likewise.
8093 (struct _stmt_vec_info): Add dr_aux member.
8094 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
8095 (DR_MISALIGNMENT_UNINITIALIZED): New.
8096 (set_dr_misalignment): Adjust.
8097 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
8098 (vect_analyze_loop): Adjust prototype.
8099 (vect_analyze_loop_form): Likewise.
8100 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8101 Compute dependences lazily.
8102 (vect_record_base_alignments): Use shared datarefs/ddrs.
8103 (vect_verify_datarefs_alignment): Likewise.
8104 (vect_analyze_data_refs_alignment): Likewise.
8105 (vect_analyze_data_ref_accesses): Likewise.
8106 (vect_analyze_data_refs): Likewise.
8107 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
8108 constructor parameter for shared part.
8109 (vect_analyze_loop_form): Pass in shared part and adjust.
8110 (vect_analyze_loop_2): Pass in storage for the number of
8111 stmts. Move loop nest finding to the caller. Compute
8113 (vect_analyze_loop): Pass in shared part.
8114 (vect_transform_loop): Verify shared datarefs are unchanged.
8115 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
8116 constructor parameter for shared part.
8117 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
8118 (vect_slp_bb): Verify shared datarefs are unchanged before
8120 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
8122 (new_stmt_vec_info): Initialize DR_AUX misalignment to
8123 DR_MISALIGNMENT_UNINITIALIZED.
8124 * tree-vectorizer.c (vec_info::vec_info): Add constructor
8125 parameter for shared part.
8126 (vec_info::~vec_info): Adjust.
8127 (vec_info_shared::vec_info_shared): New.
8128 (vec_info_shared::~vec_info_shared): Likewise.
8129 (vec_info_shared::save_datarefs): Likewise.
8130 (vec_info_shared::check_datarefs): Likewise.
8131 (try_vectorize_loop_1): Construct shared part live for analyses
8132 of a single loop for multiple vector sizes.
8133 * tree-parloops.c (gather_scalar_reductions): Adjust.
8135 2018-06-25 Richard Biener <rguenther@suse.de>
8137 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
8138 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
8139 (vect_analyze_data_refs): Remove similar code from here and
8140 simplify accordingly.
8142 2018-06-25 Richard Biener <rguenther@suse.de>
8144 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
8145 for reverse storage order accesses rather than asserting
8146 they cannot happen here.
8148 2018-06-25 Tom de Vries <tdevries@suse.de>
8151 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
8152 Use data16 instead of .byte for insn prefix.
8154 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
8157 * parser.c (make_char_string_pack): Pass this literal chars
8158 through cpp_interpret_string.
8159 (cp_parser_userdef_numeric_literal): Check the result of
8160 make_char_string_pack.
8162 2018-06-24 Maya Rashish <coypu@sdf.org>
8164 * ginclude/stddef.h: Simplify conditions around avoiding
8165 re-definition of __size_t.
8167 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8169 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
8170 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
8172 2018-06-22 Maya Rashish <coypu@sdf.org>
8174 * doc/invoke.texi (mno-fancy-math-387): Update for changes
8175 made to OpenBSD and NetBSD through the years.
8177 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
8179 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8180 behavior of vec_pack (vector double, vector double) to match
8181 behavior of vec_float2 (vector double, vector double).
8183 2018-06-22 Olivier Hainque <hainque@adacore.com>
8185 * gimplify.c (gimplify_function_tree): Prevent creation
8186 of a trampoline for the address of the current function
8187 passed to entry/exit instrumentation hooks.
8189 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
8192 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
8195 2018-06-22 Martin Liska <mliska@suse.cz>
8197 PR tree-optimization/86263
8198 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8199 Bail out if is_enabled is false.
8200 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
8202 (jump_table_cluster::is_enabled): New function.
8204 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8206 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8207 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
8208 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
8209 (lto_input_ts_binfo_tree_pointers): Likewise.
8210 * tree-streamer-out.c (streamer_write_tree_bitfields,
8211 write_ts_binfo_tree_pointers): Likewise.
8212 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
8214 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
8216 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
8218 2018-06-22 Martin Liska <mliska@suse.cz>
8220 * symbol-summary.h (get): Make it pure and inline move
8221 functionality from ::get function.
8222 (get): Remove and inline into ::get and ::get_create.
8223 (get_create): Move code from ::get function.
8225 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8228 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
8229 -x assembler-with-cpp.
8231 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8233 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
8234 _FILE_OFFSET_BITS=64 for C++.
8236 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
8238 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
8239 conversion insn that shows up when pr85657-3.c is compiled using
8240 IEEE 128-bit long double.
8241 (neg<mode>2_internal): Use the correct mode to check whether the
8242 mode is IBM extended.
8243 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
8244 multiply and divide external functions from being created more
8247 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8249 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
8251 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
8252 the edge can be forwarded.
8253 (cfg_layout_merge_blocks): Likewise.
8255 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8257 * except.c (finish_eh_generation): Commit edge insertions only after
8258 the EH edges have been redirected from post-landing to landing pads.
8260 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8262 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
8263 create_tmp_var_for to create the FRAME decl.
8264 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
8266 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8268 * tree-inline.c (copy_edges_for_bb): Minor tweak.
8269 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
8270 debug statement when resetting its value.
8271 (expand_call_inline): Copy the locus of the call onto the assignment
8272 of the return value, if any. Use local variable in more cases.
8274 2018-06-21 Martin Liska <mliska@suse.cz>
8276 * ipa-pure-const.c (propagate_nothrow): Use
8277 funct_state_summaries->get.
8278 (dump_malloc_lattice): Likewise.
8279 (propagate_malloc): Likewise.
8281 2018-06-21 Richard Biener <rguenther@suse.de>
8283 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
8284 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
8285 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
8287 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
8288 BLOCK_ABSTRACT_ORIGIN unconditionally.
8290 2018-06-21 David Malcolm <dmalcolm@redhat.com>
8292 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
8294 * ipa-reference.c (ipa_reference_c_finalize): Delete
8295 ipa_ref_opt_sum_summaries and set it to NULL.
8297 2018-06-21 Tom de Vries <tdevries@suse.de>
8299 PR tree-optimization/85859
8300 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
8301 test with comment from bb_no_side_effects_p.
8303 2018-06-21 Richard Biener <rguenther@suse.de>
8305 PR tree-optimization/86232
8306 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
8307 max for constant niter.
8309 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8311 * config/aarch64/aarch64-simd.md
8312 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
8314 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
8316 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
8317 Make opernads of the unspec commutative.
8319 2018-06-21 Richard Biener <rguenther@suse.de>
8321 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
8322 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
8323 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
8324 (vect_analyze_data_ref_dependence): Re-order checks to deal with
8326 (vect_record_base_alignments): Do not record base alignment
8327 for gathers or scatters.
8328 (vect_compute_data_ref_alignment): Drop return value that is always
8329 true. Bail out early for gathers or scatters.
8330 (vect_enhance_data_refs_alignment): Bail out early for gathers
8332 (vect_find_same_alignment_drs): Likewise.
8333 (vect_analyze_data_refs_alignment): Remove dead code.
8334 (vect_slp_analyze_and_verify_node_alignment): Likewise.
8335 (vect_analyze_data_refs): For possible gathers or scatters do
8336 not create an alternate DR, just check their possible validity
8337 and mark them. Adjust DECL_NONALIASED handling to not rely
8339 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8340 update inits of gathers or scatters.
8341 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
8342 Also copy gather/scatter flag to pattern vinfo.
8344 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
8346 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8347 behavior of vec_packsu (vector unsigned long long, vector unsigned
8348 long long) to match behavior of vec_packs with same signature.
8350 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
8351 Thomas Schwinge <thomas@codesourcery.com>
8352 Cesar Philippidis <cesar@codesourcery.com>
8354 * gimplify.c (gimplify_scan_omp_clauses): Add support for
8355 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8356 (gimplify_adjust_omp_clauses): Likewise.
8357 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
8358 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
8359 (gimplify_omp_target_update): Update handling of acc update and
8361 * omp-low.c (install_var_field): Remove unused parameter
8362 base_pointers_restrict.
8363 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
8364 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
8366 (omp_target_base_pointers_restrict_p): Delete.
8367 (scan_omp_target): Update call to scan_sharing_clauses.
8368 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
8370 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
8371 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8372 (convert_local_omp_clauses): Likewise.
8373 * tree-pretty-print.c (dump_omp_clause): Likewise.
8374 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
8376 (omp_clause_code_name): Likewise.
8378 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8381 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
8384 PR tree-optimization/86231
8385 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
8386 anti-range don't overwrite *vr0min before using it to compute *vr0max.
8388 2018-06-20 Tom de Vries <tdevries@suse.de>
8390 PR tree-optimization/86097
8391 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
8392 iv type if signedness of iv type is not the same as that of *nit.
8394 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8396 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
8397 EDGE_EH edges, verify they are all EDGE_EH.
8399 2018-06-20 Maya Rashish <coypu@sdf.org>
8401 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
8403 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8405 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
8406 * config/aarch64/aarch64.c (xgene1_tunings): Add
8407 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
8408 (aarch64_mode_valid_for_sched_fusion_p):
8409 Allow 16-byte modes.
8410 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
8411 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
8413 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
8415 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
8416 * config/aarch64/iterators.md (VQ2): New mode iterator.
8418 2018-06-20 Martin Liska <mliska@suse.cz>
8420 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8421 Change default ratio from 10 to 8.
8423 2018-06-20 Martin Liska <mliska@suse.cz>
8425 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8427 (bit_test_cluster::find_bit_tests): Likewise.
8428 (switch_decision_tree::analyze_switch_statement): Find clusters.
8429 * tree-switch-conversion.h (struct jump_table_cluster): Document
8432 2018-06-20 Martin Liska <mliska@suse.cz>
8434 * tree-switch-conversion.c (switch_conversion::collect):
8435 Record m_uniq property.
8436 (switch_conversion::expand): Bail out for special conditions.
8437 (group_cluster::~group_cluster): New.
8438 (group_cluster::group_cluster): Likewise.
8439 (group_cluster::dump): Likewise.
8440 (jump_table_cluster::emit): New.
8441 (switch_decision_tree::fix_phi_operands_for_edges): New.
8442 (struct case_node): Remove struct.
8443 (jump_table_cluster::can_be_handled): New.
8444 (case_values_threshold): Moved to header.
8445 (reset_out_edges_aux): Likewise.
8446 (jump_table_cluster::is_beneficial): New.
8447 (bit_test_cluster::can_be_handled): Likewise.
8448 (add_case_node): Remove.
8449 (bit_test_cluster::is_beneficial): New.
8450 (case_bit_test::cmp): New.
8451 (bit_test_cluster::emit): New.
8452 (expand_switch_as_decision_tree_p): Remove.
8453 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
8454 (fix_phi_operands_for_edge): Likewise.
8455 (switch_decision_tree::analyze_switch_statement): New.
8456 (compute_cases_per_edge): Move ...
8457 (switch_decision_tree::compute_cases_per_edge): ... here.
8458 (try_switch_expansion): Likewise.
8459 (switch_decision_tree::try_switch_expansion): Likewise.
8460 (record_phi_operand_mapping): Likewise.
8461 (switch_decision_tree::record_phi_operand_mapping): Likewise.
8462 (emit_case_decision_tree): Likewise.
8463 (switch_decision_tree::emit): Likewise.
8464 (balance_case_nodes): Likewise.
8465 (switch_decision_tree::balance_case_nodes): Likewise.
8466 (dump_case_nodes): Likewise.
8467 (switch_decision_tree::dump_case_nodes): Likewise.
8468 (emit_jump): Likewise.
8469 (switch_decision_tree::emit_jump): Likewise.
8470 (emit_cmp_and_jump_insns): Likewise.
8471 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
8472 (emit_case_nodes): Likewise.
8473 (switch_decision_tree::emit_case_nodes): Likewise.
8474 (conditional_probability): Remove.
8475 * tree-switch-conversion.h (enum cluster_type): New.
8477 (struct cluster): Likewise.
8478 (cluster::cluster): Likewise.
8479 (struct simple_cluster): Likewise.
8480 (simple_cluster::simple_cluster): Likewise.
8481 (struct group_cluster): Likewise.
8482 (struct jump_table_cluster): Likewise.
8483 (struct bit_test_cluster): Likewise.
8484 (struct min_cluster_item): Likewise.
8485 (struct case_tree_node): Likewise.
8486 (case_tree_node::case_tree_node): Likewise.
8487 (jump_table_cluster::case_values_threshold): Likewise.
8488 (struct case_bit_test): Likewise.
8489 (struct switch_decision_tree): Likewise.
8490 (struct switch_conversion): Likewise.
8491 (switch_decision_tree::reset_out_edges_aux): Likewise.
8493 2018-06-20 Martin Liska <mliska@suse.cz>
8495 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
8496 (hoist_edge_and_branch_if_true): Likewise.
8497 (expand_switch_using_bit_tests_p): Likewise.
8498 (struct case_bit_test): Likewise.
8499 (case_bit_test_cmp): Likewise.
8500 (emit_case_bit_tests): Likewise.
8501 (switch_conversion::switch_conversion): New class.
8502 (struct switch_conv_info): Remove old struct.
8503 (collect_switch_conv_info): More to ...
8504 (switch_conversion::collect): ... this.
8505 (check_range): Likewise.
8506 (switch_conversion::check_range): Likewise.
8507 (check_all_empty_except_final): Likewise.
8508 (switch_conversion::check_all_empty_except_final): Likewise.
8509 (check_final_bb): Likewise.
8510 (switch_conversion::check_final_bb): Likewise.
8511 (create_temp_arrays): Likewise.
8512 (switch_conversion::create_temp_arrays): Likewise.
8513 (free_temp_arrays): Likewise.
8514 (gather_default_values): Likewise.
8515 (switch_conversion::gather_default_values): Likewise.
8516 (build_constructors): Likewise.
8517 (switch_conversion::build_constructors): Likewise.
8518 (constructor_contains_same_values_p): Likewise.
8519 (switch_conversion::contains_same_values_p): Likewise.
8520 (array_value_type): Likewise.
8521 (switch_conversion::array_value_type): Likewise.
8522 (build_one_array): Likewise.
8523 (switch_conversion::build_one_array): Likewise.
8524 (build_arrays): Likewise.
8525 (switch_conversion::build_arrays): Likewise.
8526 (gen_def_assigns): Likewise.
8527 (switch_conversion::gen_def_assigns): Likewise.
8528 (prune_bbs): Likewise.
8529 (switch_conversion::prune_bbs): Likewise.
8530 (fix_phi_nodes): Likewise.
8531 (switch_conversion::fix_phi_nodes): Likewise.
8532 (gen_inbound_check): Likewise.
8533 (switch_conversion::gen_inbound_check): Likewise.
8534 (process_switch): Use the newly created class.
8535 (switch_conversion::expand): New.
8536 (switch_conversion::~switch_conversion): New.
8537 * tree-switch-conversion.h: New file.
8539 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8541 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
8542 tree-vect-patterns.c.
8543 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
8544 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
8545 (vect_recog_sad_pattern): Likewise.
8546 (vect_recog_widen_sum_pattern): Likewise.
8547 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
8548 (vect_recog_widen_shift_pattern): Remove the type_in argument.
8549 (vect_recog_rotate_pattern): Likewise.
8550 (vect_recog_mult_pattern): Likewise.
8551 (vect_recog_vector_vector_shift_pattern): Likewise.
8552 (vect_recog_divmod_pattern): Likewise.
8553 (vect_recog_mixed_size_cond_pattern): Likewise.
8554 (vect_recog_bool_pattern): Likewise.
8555 (vect_recog_mask_conversion_pattern): Likewise.
8556 (vect_try_gather_scatter_pattern): Likewise.
8557 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
8558 (vect_recog_over_widening_pattern): Likewise.
8559 (vect_recog_gather_scatter_pattern): Likewise.
8560 (vect_recog_func_ptr): Move from tree-vectorizer.h
8561 (vect_vect_recog_func_ptrs): Move further down the file.
8562 (vect_recog_func): Likewise. Remove the third argument.
8563 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
8564 (vect_pattern_recog_1): Expect the pattern function to do any
8565 necessary target tests. Also expect it to provide a vector type.
8566 Remove the type_in handling.
8568 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8570 * tree-vect-patterns.c (vect_pattern_detected): New function.
8571 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
8572 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
8573 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
8574 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
8575 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8576 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8577 (vect_recog_mask_conversion_pattern)
8578 (vect_try_gather_scatter_pattern): Likewise.
8580 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8582 * tree-vect-patterns.c (vect_get_internal_def): New function.
8583 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8584 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
8585 (search_type_for_mask_1): Use it.
8587 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8589 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8590 redundant WIDEN_SUM_EXPR handling.
8591 (vect_recog_sad_pattern): Likewise.
8593 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8595 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8596 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
8597 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
8598 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
8599 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
8601 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8603 * tree-vect-stmts.c (vectorizable_call): Make sure that we
8604 use the stmt_vec_info of the original bb statement for the
8605 new zero assignment, even if the call is part of a pattern.
8607 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8609 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
8610 that the sequence is attached to the original statement rather
8611 than the pattern statement.
8612 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
8613 PATTERN_DEF_SEQ from the original statement rather than
8614 the main pattern statement.
8615 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
8616 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
8617 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
8619 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
8621 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
8622 definition statements before the early exit for statements that aren't
8624 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
8626 (vect_transform_loop): ...here. Process pattern definition
8627 statements without first checking whether the main pattern
8628 statement is live or relevant.
8630 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
8632 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
8633 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
8635 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
8637 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
8638 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
8639 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
8640 (expand_block_compare): Change select_block_compare_mode call.
8641 (expand_strncmp_align_check): Use new functions, fix comment.
8642 (emit_final_str_compare_gpr): New function.
8643 (expand_strn_compare): Refactor and clean up code.
8644 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
8646 2018-06-19 Tony Reix <tony.reix@atos.com>
8647 Damien Bergamini <damien.bergamini@atos.com>
8648 David Edelsohn <dje.gcc@gmail.com>
8650 * collect2.c (static_obj): New variable.
8651 (static_libs): New variable.
8652 (is_in_list): Uncomment declaration.
8653 (main): Track AIX libraries linked statically.
8654 (is_in_list): Uncomment definition.
8655 (scan_prog_file): Don't add AIX shared libraries initializer
8656 to constructor list if linking statically.
8658 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
8660 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
8662 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
8664 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
8666 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
8669 2018-06-19 Martin Liska <mliska@suse.cz>
8671 * config/i386/i386.c (ix86_can_inline_p): Do not use
8672 ipa_fn_summaries::get_create.
8673 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
8675 (devirtualization_time_bonus): Likewise.
8676 (ipcp_propagate_stage): Likewise.
8677 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
8678 (edge_set_predicate): Likewise.
8679 (evaluate_conditions_for_known_args): Likewise.
8680 (evaluate_properties_for_edge): Likewise.
8681 (ipa_call_summary::reset): Tranform to ...
8682 (ipa_call_summary::~ipa_call_summary): ... this.
8683 (ipa_fn_summary::reset): Transform to ...
8684 (ipa_fn_summary::~ipa_fn_summary): ... this.
8685 (ipa_fn_summary_t::remove): Rename to ...
8686 (ipa_fn_summary_t::remove_callees): ... this.
8687 (ipa_fn_summary_t::duplicate): Use placement new
8688 instead of memory copy.
8689 (ipa_call_summary_t::duplicate): Likewise.
8690 (ipa_call_summary_t::remove): Remove.
8691 (dump_ipa_call_summary): Change get_create to get.
8692 (ipa_dump_fn_summary): Dump only when summary exists.
8693 (analyze_function_body): Use symbol_summary::get instead
8695 (compute_fn_summary): Likewise.
8696 (estimate_edge_devirt_benefit): Likewise.
8697 (estimate_edge_size_and_time): Likewise.
8698 (inline_update_callee_summaries): Likewise.
8699 (remap_edge_change_prob): Likewise.
8700 (remap_edge_summaries): Likewise.
8701 (ipa_merge_fn_summary_after_inlining): Likewise.
8702 (write_ipa_call_summary): Likewise.
8703 (ipa_fn_summary_write): Likewise.
8704 (ipa_free_fn_summary): Likewise.
8705 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
8706 (struct ipa_call_summary): Likewise.
8707 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
8709 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
8710 (estimate_size_after_inlining): Likewise.
8711 (estimate_growth): Likewise.
8712 (growth_likely_positive): Likewise.
8713 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
8714 (inline_call): Likewise.
8715 * ipa-inline.c (caller_growth_limits): Likewise.
8716 (can_inline_edge_p): Likewise.
8717 (can_inline_edge_by_limits_p): Likewise.
8718 (compute_uninlined_call_time): Likewise.
8719 (compute_inlined_call_time): Likewise.
8720 (want_inline_small_function_p): Likewise.
8721 (edge_badness): Likewise.
8722 (update_caller_keys): Likewise.
8723 (update_callee_keys): Likewise.
8724 (inline_small_functions): Likewise.
8725 (inline_to_all_callers_1): Likewise.
8726 (dump_overall_stats): Likewise.
8727 (early_inline_small_functions): Likewise.
8728 (early_inliner): Likewise.
8729 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
8730 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8731 * ipa-pure-const.c (malloc_candidate_p): Likewise.
8732 * ipa-split.c (execute_split_functions): Likewise.
8733 * symbol-summary.h: Likewise.
8734 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
8736 2018-06-19 Richard Biener <rguenther@suse.de>
8738 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
8739 (vectorize_loops): ... here. Fix dbgcnt handling.
8740 (try_vectorize_loop): Wrap try_vectorize_loop_1.
8742 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
8745 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
8746 ieee128 argument takes up only one (vector) register, not two (floating
8749 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
8751 * gimplify.c (gimplify_init_constructor): Really never clear for an
8752 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
8754 2018-06-19 Richard Biener <rguenther@suse.de>
8756 PR tree-optimization/86179
8757 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
8758 after failed recognition.
8760 2018-06-18 Martin Sebor <msebor@redhat.com>
8763 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
8764 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
8765 Handle integer subtraction.
8766 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
8767 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
8769 2018-06-18 David Malcolm <dmalcolm@redhat.com>
8771 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
8772 param from rtx to rtx_insn *.
8773 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
8775 (frv_ifcvt_modify_insn): Likwise.
8776 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
8777 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
8778 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
8779 as_a <rtx_insn *> cast to local "unprotected_region" once
8780 it's been established that it's not NULL or pc_rtx.
8781 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
8782 param "sethi" from rtx to rtx_insn *.
8783 (nds32_group_float_insns): Likewise for param "insn".
8784 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
8786 (vax_output_int_subtract): Likewise.
8787 * config/vax/vax.c (vax_output_int_add): Likewise for param
8789 (vax_output_int_subtract): Likewise.
8790 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
8791 (emit_pattern_after): Likewise for param "after".
8792 (emit_insn_after): Likewise.
8793 (emit_jump_insn_after): Likewise.
8794 (emit_call_insn_after): Likewise.
8795 (emit_debug_insn_after): Likewise.
8796 (emit_pattern_before): Likewise for param "before".
8797 (emit_insn_before): Likewise.
8798 (emit_jump_insn_before): Likewise.
8799 * final.c (get_insn_template): Likewise for param "insn", removing
8801 * output.h (get_insn_template): Likewise for 2nd param.
8802 * rtl.h (emit_insn_before): Likewise.
8803 (emit_jump_insn_before): Likewise.
8804 (emit_debug_insn_before_noloc): Likewise.
8805 (emit_insn_after): Likewise.
8806 (emit_jump_insn_after): Likewise.
8807 (emit_call_insn_after): Likewise.
8808 (emit_debug_insn_after): Likewise.
8809 (set_insn_deleted): Likewise for param.
8811 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
8814 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
8815 floating point modes, so that IFmode is numerically greater than
8816 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
8817 to declare the ordering. This prevents IFmode from being
8818 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
8819 machine. Include rs6000-modes.h to share the fractional values
8820 between genmodes* and the rest of the compiler.
8824 * config/rs6000/rs6000-modes.h: New file.
8825 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
8826 meaning of rs6000_long_double_size so that 126..128 selects an
8827 appropriate 128-bit floating point type.
8828 (rs6000_option_override_internal): Likewise.
8829 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
8830 (TARGET_LONG_DOUBLE_128): Change the meaning of
8831 rs6000_long_double_size so that 126..128 selects an appropriate
8832 128-bit floating point type.
8833 (LONG_DOUBLE_TYPE_SIZE): Update comment.
8834 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
8835 source and destination to match the standard usage.
8836 (truncifkf2): Likewise.
8837 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
8838 ISA 2.07 to use an explicit clobber, instead of passing in a
8840 (copysign<mode>3_soft): Likewise.
8842 2018-06-18 David Malcolm <dmalcolm@redhat.com>
8844 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8845 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
8846 (vect_slp_analyze_instance_dependence): Likewise.
8847 (vect_enhance_data_refs_alignment): Likewise.
8848 (vect_analyze_data_refs_alignment): Likewise.
8849 (vect_slp_analyze_and_verify_instance_alignment
8850 (vect_analyze_data_ref_accesses): Likewise.
8851 (vect_prune_runtime_alias_test_list): Likewise.
8852 (vect_analyze_data_refs): Likewise.
8853 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
8854 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8855 (vect_analyze_scalar_cycles_1): Likewise.
8856 (vect_get_loop_niters): Likewise.
8857 (vect_analyze_loop_form_1): Likewise.
8858 (vect_update_vf_for_slp): Likewise.
8859 (vect_analyze_loop_operations): Likewise.
8860 (vect_analyze_loop): Likewise.
8861 (vectorizable_induction): Likewise.
8862 (vect_transform_loop): Likewise.
8863 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
8864 * tree-vect-slp.c (vect_analyze_slp): Likewise.
8865 (vect_make_slp_decision): Likewise.
8866 (vect_detect_hybrid_slp): Likewise.
8867 (vect_slp_analyze_operations): Likewise.
8868 (vect_slp_bb): Likewise.
8869 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
8870 (vectorizable_bswap): Likewise.
8871 (vectorizable_call): Likewise.
8872 (vectorizable_simd_clone_call): Likewise.
8873 (vectorizable_conversion): Likewise.
8874 (vectorizable_assignment): Likewise.
8875 (vectorizable_shift): Likewise.
8876 (vectorizable_operation): Likewise.
8877 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
8879 2018-06-18 Martin Sebor <msebor@redhat.com>
8881 PR tree-optimization/81384
8882 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
8883 * builtins.c (expand_builtin_strnlen): New function.
8884 (expand_builtin): Call it.
8885 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
8886 * builtins.def (BUILT_IN_STRNLEN): New.
8887 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
8888 Warn for bounds in excess of maximum object size.
8889 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
8890 single-value ranges. Handle strnlen.
8891 (handle_builtin_strlen): Handle strnlen.
8892 (strlen_check_and_optimize_stmt): Same.
8893 * doc/extend.texi (Other Builtins): Document strnlen.
8895 2018-06-18 Maya Rashish <coypu@sdf.org>
8897 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
8898 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
8899 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
8901 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
8903 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
8905 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8907 * tree.c (escaped_string::escape): Replace cast to char * by
8908 const_cast<char *> (unescaped).
8910 2018-06-18 Nick Clifton <nickc@redhat.com>
8913 * tree.c (escaped_string): New class. Converts an unescaped
8914 string into its escaped equivalent.
8915 (warn_deprecated_use): Use the new class to convert the
8916 deprecation message, if present.
8917 (test_escaped_strings): New self test.
8918 (test_c_tests): Add test_escaped_strings.
8919 * doc/extend.texi (deprecated): Add a note that the
8920 deprecation message is affected by the -fmessage-length
8921 option, and that control characters will be escaped.
8922 (#pragma GCC error): Document this pragma.
8923 (#pragma GCC warning): Likewise.
8924 * doc/invoke.texi (-fmessage-length): Document this option's
8925 effect on the #warning and #error preprocessor directives and
8926 the deprecated attribute.
8928 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
8930 * tree.c (decl_value_expr_lookup): Revert latest change.
8931 (decl_value_expr_insert): Likewise.
8933 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
8935 * gimplify.c (nonlocal_vlas): Delete.
8936 (nonlocal_vla_vars): Likewise.
8937 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
8939 (gimplify_body): Do not create and destroy nonlocal_vlas.
8940 * tree-nested.c: Include diagnostic.h.
8941 (use_pointer_in_frame): Tweak.
8942 (lookup_field_for_decl): Add assertion and declare the transformation.
8943 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
8944 internal error when the reference is in a wrong context. Do not
8945 create a debug decl by default.
8946 (note_nonlocal_block_vlas): Delete.
8947 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
8948 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
8949 create a debug decl by default.
8950 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
8951 call is in a wrong context.
8952 (fixup_vla_decls): New function.
8953 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
8954 debug variables were created.
8955 * tree.c (decl_value_expr_lookup): Add checking assertion.
8956 (decl_value_expr_insert): Likewise.
8958 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8961 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
8962 * tree-scalar-evolution.c (interpret_expr): Likewise.
8963 (expression_expensive_p): Likewise.
8964 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
8965 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
8966 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
8967 (ssa_defined_by_minus_one_stmt_p): New.
8969 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
8972 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
8973 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
8974 * dojump.c (do_jump): Likewise.
8975 * expr.c (expand_expr_real_2): Check operand type's sign.
8976 * fold-const.c (const_unop): Handle ABSU_EXPR.
8977 (fold_abs_const): Likewise.
8978 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
8979 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
8980 (strip_sign_op_1): Likesise.
8981 * match.pd: Add new pattern to generate ABSU_EXPR.
8982 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
8983 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
8984 * tree-eh.c (operation_could_trap_helper_p): Likewise.
8985 * tree-inline.c (estimate_operator_cost): Likewise.
8986 * tree-pretty-print.c (dump_generic_node): Likewise.
8987 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
8988 * tree.def (ABSU_EXPR): New.
8990 2018-06-16 Jakub Jelinek <jakub@redhat.com>
8993 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
8994 documented as preserved for backward compatibility only.
8995 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
8997 PR rtl-optimization/86108
8998 * bb-reorder.c (create_forwarder_block): Renamed to ...
8999 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
9000 jump from new landing pad to the second part.
9001 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
9004 2018-06-15 Jakub Jelinek <jakub@redhat.com>
9007 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
9008 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
9009 Only call store_expr for halves if the mode is the same.
9012 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
9013 Fix up comment formatting.
9015 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
9017 * typed-splay-tree.h (typed_splay_tree::remove): New function.
9018 (typed_splay_tree::closure,
9019 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
9020 (typed_splay_tree::typed_splay_tree,
9021 typed_splay_tree::operator =): Declared private.
9022 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
9023 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
9024 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
9025 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
9026 typed_splay_tree::splay_tree_splay,
9027 typed_splay_tree::splay_tree_foreach_helper,
9028 typed_splay_tree::splay_tree_insert,
9029 typed_splay_tree::splay_tree_remove,
9030 typed_splay_tree::splay_tree_lookup,
9031 typed_splay_tree::splay_tree_predecessor,
9032 typed_splay_tree::splay_tree_successor,
9033 typed_splay_tree::splay_tree_min,
9034 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
9035 (typed_splay_tree::root, typed_splay_tree::comp,
9036 typed_splay_tree::delete_key,
9037 typed_splay_tree::delete_value): New data members.
9038 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
9039 typed_splay_tree::remove.
9041 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
9043 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
9044 -mginv and -mno-ginv to the assembler.
9045 * config/mips/mips.opt (-mcrc): New option.
9047 * doc/invoke.text (-mcrc): Document.
9050 2018-06-15 Nick Clifton <nickc@redhat.com>
9053 * tree.c (escaped_string): New class. Converts an unescaped
9054 string into its escaped equivalent.
9055 (warn_deprecated_use): Use the new class to convert the
9056 deprecation message, if present.
9057 (test_escaped_strings): New self test.
9058 (test_c_tests): Add test_escaped_strings.
9059 * doc/extend.texi (deprecated): Add a note that the
9060 deprecation message is affected by the -fmessage-length
9061 option, and that control characters will be escaped.
9062 (#pragma GCC error): Document this pragma.
9063 (#pragma GCC warning): Likewise.
9064 * doc/invoke.texi (-fmessage-length): Document this option's
9065 effect on the #warning and #error preprocessor directives and
9066 the deprecated attribute.
9068 2018-06-15 Richard Biener <rguenther@suse.de>
9070 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
9071 here, also noting vector size used.
9072 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
9073 size used in MSG_OPTIMIZED_LOCATIONS dump.
9074 (pass_slp_vectorize::execute): Adjust.
9076 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
9079 * config/arc/arc.c (arc_return_address_register): Fix
9082 2018-06-15 Richard Biener <rguenther@suse.de>
9085 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
9086 leave useless conversion stripping to force_gimple_operand_gsi.
9087 (gimplify_build2): Likewise.
9088 (gimplify_build1): Likewise.
9090 2018-06-15 Richard Biener <rguenther@suse.de>
9093 * tree-cfg.c (move_stmt_op): unshare invariant addresses
9094 before adjusting their block.
9096 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
9098 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
9099 multilibs for *-*-rtems*.
9100 * config/riscv/t-rtems: New file.
9102 2018-06-14 Jakub Jelinek <jakub@redhat.com>
9105 * match.pd ((A +- CST1) +- CST2): Punt if last resort
9106 unsigned_type_for returns NULL.
9109 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
9110 subregs of multi-word pseudos unless the float mode has word size.
9112 2018-06-14 Richard Biener <rguenther@suse.de>
9115 * tree-vect-generic.c (build_word_mode_vector_type): Remove
9116 duplicate and harmful type_hash_canon.
9117 * tree.c (type_hash_canon): Assert we didn't find ourselves.
9119 2018-06-14 Richard Biener <rguenther@suse.de>
9122 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
9125 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
9127 * config/rtems.h (STDINT_LONG32): Define.
9129 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
9130 Prachi Godbole <prachi.godbole@imgtec.com>
9132 * config/mips/mips-cpus.def: Define P6600.
9133 * config/mips/mips-tables.opt: Regenerate.
9134 * config/mips/mips.c (mips_ucbranch_type): New enum.
9135 (mips_rtx_cost_data): Add support for P6600.
9136 (mips_issue_rate): Likewise.
9137 (mips_multipass_dfa_lookahead): Likewise.
9138 (mips_avoid_hazard): Likewise.
9139 (mips_reorg_process_insns): Likewise.
9140 (mips_classify_branch_p6600): New function.
9141 * config/mips/mips.h (TUNE_P6600): New define.
9142 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
9143 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
9144 * config/mips/mips.md: Include p6600.md.
9145 (processor): Add p6600.
9146 * config/mips/p6600.md: New file.
9147 * doc/invoke.texi: Add p6600 to supported architectures.
9149 2018-06-13 Martin Sebor <msebor@redhat.com>
9151 PR tree-optimization/86114
9152 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
9154 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
9156 2018-06-13 Richard Biener <rguenther@suse.de>
9158 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
9159 Properly set vector type of the intermediate stmt.
9160 * tree-vect-stmts.c (vectorizable_operation): The destination
9161 var always has vectype_out type.
9163 2018-06-13 Jeff Law <law@redhat.com>
9165 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
9166 integer 0 for argument to print_rtl_with_bb.
9167 (rl78_reorg): Likewise.
9169 2018-06-13 David Malcolm <dmalcolm@redhat.com>
9171 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
9172 from rtx to rtx_insn *.
9173 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
9175 (add_sched_insns_for_speculation): Likewise for local "target",
9176 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9177 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
9178 from rtx_insn ** to rtx_code_label **.
9179 (reorg_emit_nops): Likewise.
9180 (c6x_reorg): Likewise for local "call_labels".
9181 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
9183 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
9184 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
9185 the loops over LABEL_REFS.
9186 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
9188 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9189 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
9190 (split_branches): Strengthen local "olabel" from rtx to
9191 rtx_insn *, adding a safe_as_a cast.
9192 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
9194 (add_insn_after): Likewise for first two params.
9195 (add_insn_before): Likewise.
9196 (remove_insn): Likewise for param.
9197 (emit_pattern_before_noloc): Likewise for second and third params.
9198 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
9199 (emit_call_insn_before_noloc): Likewise.
9200 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
9202 (emit_barrier_before): Likewise.
9203 (emit_label_before): Strengthen "label" param from "rtx" to
9204 "rtx_code_label *". Strengthen "before" param from "rtx" to
9206 (emit_insn_after_1): Strengthen "after" param from "rtx" to
9208 (emit_pattern_after_noloc): Likewise.
9209 (emit_insn_after_noloc): Likewise.
9210 (emit_jump_insn_after_noloc): Likewise.
9211 (emit_call_insn_after_noloc): Likewise.
9212 (emit_debug_insn_after_noloc): Likewise.
9213 (emit_barrier_after): Likewise.
9214 (emit_label_after): Likewise for both params.
9215 (emit_pattern_after_setloc): Likewise for "after" param. Convert
9216 "loc" param from "int" to "location_t".
9217 (emit_insn_after_setloc): Likewise.
9218 (emit_jump_insn_after_setloc): Likewise.
9219 (emit_call_insn_after_setloc): Likewise.
9220 (emit_debug_insn_after_setloc): Likewise.
9221 (emit_pattern_before_setloc): Likewise for "before" param. Convert
9222 "loc" param from "int" to "location_t".
9223 (emit_pattern_before): Convert NULL_RTX to NULL.
9224 (emit_insn_before_setloc): Convert "loc" param from "int" to
9226 (emit_jump_insn_before_setloc): Likewise.
9227 (emit_call_insn_before_setloc): Likewise.
9228 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
9229 rtx_insn *. Convert "loc" param from "int" to "location_t".
9230 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
9231 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
9232 Convert 3rd param from "int" to "location_t".
9233 (emit_barrier_before, emit_barrier_after, next_real_insn):
9234 Strengthen param from rtx to rtx_insn *.
9235 (emit_label_before): Strengthen 1st param from "rtx" to
9236 "rtx_code_label *". Strengthen 2nd param from "rtx" to
9238 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
9239 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
9240 Strengthen 2nd param from "rtx" to "rtx_insn *".
9241 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
9242 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
9243 Likewise. Convert 3rd param from "int" to "location_t".
9244 (emit_label_after): Strengthen 1st param from "rtx" to
9246 (next_real_insn, remove_insn): Strengthen param from "rtx" to
9248 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
9249 from "rtx" to "rtx_insn *".
9251 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
9253 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
9254 bodies streamed in with -Q.
9255 * dumpfile.c (dump_files): Add lto-stream-out dump file.
9256 * dumpfile.h (tree_dump_index): Add lto_stream_out.
9257 * gimple-streamer-out.c: Include gimple-pretty-print.h
9258 (output_bb): Dump stmts streamed.
9259 * lto-section-out.c: Include print-tree.h
9260 (lto_begin_section): Dump sections created.
9261 (lto_output_decl_index): Dump decl encoded.
9262 * lto-streamer-out.c: Include print-tree.h
9263 (create_output_block): Dump output block created.
9264 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
9265 (output_function): Dump function output.
9266 (output_constructor): Dump constructor streamed.
9267 (write_global_stream): Output indexes encoded.
9268 (produce_asm_for_decls): Dump streams encoded.
9269 * lto-streamer.c (streamer_dump_file): New global var.
9270 * lto-streamer.h (streamer_dump_file): Declare.
9271 * passes.c (ipa_write_summaries): Initialize streamer dump.
9272 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
9275 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
9278 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
9279 offsets for register save directives. Emit a second batch of save
9280 directives, if need be, when the function accesses prior frames.
9282 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9284 * config/arc/fpu.md (fmasf4): Force operand to register.
9285 (fnmasf4): Likewise.
9287 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9289 * config/arc/arc-protos.h (arc_pad_return): Remove.
9290 * config/arc/arc.c (machine_function): Remove force_short_suffix
9292 (arc_print_operand): Adjust printing of '&'.
9293 (arc_verify_short): Remove conditional printing of short suffix.
9294 (arc_final_prescan_insn): Remove reference to size_reason.
9295 (pad_return): New function.
9296 (arc_reorg): Call pad_return.
9297 (arc_pad_return): Remove.
9298 (arc_init_machine_status): Remove reference to force_short_suffix.
9299 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
9300 (attr length): When attribute iscompact is true force to 2
9301 regardless; in the case of maybe check if we want to force the
9302 instruction to have 4 bytes length.
9303 (nopv): Change it to generate 4 byte long nop as well.
9304 (blockage): New pattern.
9305 (simple_return): Remove call to arc_pad_return.
9306 (p_return_i): Likewise.
9308 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9310 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
9312 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9314 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
9317 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9319 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
9320 for ARC700 and ARCv2.
9322 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
9325 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
9326 operands[2] instead of operands[1].
9329 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9331 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
9332 case, check whether the outer register overlaps an unallocatable
9333 register, not just whether it fits the required class.
9335 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
9337 * poly-int.h (can_div_trunc_p): Add new overload in which all values
9339 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
9340 (memrefs_conflict_p): Likewise.
9341 (init_alias_analysis): Likewise.
9342 * cfgexpand.c (expand_debug_expr): Likewise.
9343 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
9344 * cse.c (fold_rtx): Likewise.
9345 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
9346 * expr.c (emit_block_move_hints): Likewise.
9347 (clear_storage_hints, push_block, emit_push_insn): Likewise.
9348 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
9349 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
9350 (emit_group_store): Likewise.
9351 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
9352 to read the PRE/POST_MODIFY increment.
9353 * calls.c (store_one_arg): Use strip_offset.
9354 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
9356 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
9358 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
9359 (simplify_binary_operation_1): Extend CONST_INT handling to
9361 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
9362 than a HOST_WIDE_INT.
9363 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
9365 (adjust_mems, add_stores): Update accodingly.
9366 (vt_canonicalize_addr): Track polynomial offsets.
9367 (emit_note_insn_var_location): Likewise.
9368 (vt_add_function_parameter): Likewise.
9369 (vt_initialize): Likewise.
9371 2018-06-12 Jeff Law <law@redhat.com>
9373 * config.gcc (alpha*-*-freebsd*): Remove.
9374 * config/alpha/freebsd.h: Remove.
9376 2018-06-12 David Malcolm <dmalcolm@redhat.com>
9379 * spellcheck-tree.c (levenshtein_distance): Rename to...
9380 (get_edit_distance): ...this, and update for underlying renaming.
9381 * spellcheck-tree.h (levenshtein_distance): Rename to...
9382 (get_edit_distance): ...this.
9383 * spellcheck.c (levenshtein_distance): Rename to...
9384 (get_edit_distance): ...this. Convert from Levenshtein distance
9385 to Damerau-Levenshtein distance by supporting transpositions of
9386 adjacent characters. Rename "v1" to "v_next" and "v0" to
9388 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
9389 (selftest::test_edit_distance_unit_test_oneway): ...this, and
9390 update for underlying renaming.
9391 (selftest::levenshtein_distance_unit_test): Rename to...
9392 (selftest::test_get_edit_distance_unit): ...this, and update for
9393 underlying renaming.
9394 (selftest::test_find_closest_string): Add example from PR 69968
9395 where transposition helps
9396 (selftest::test_metric_conditions): Update for renaming.
9397 (selftest::test_metric_conditions): Likewise.
9398 (selftest::spellcheck_c_tests): Likewise.
9399 * spellcheck.h (levenshtein_distance): Rename both overloads to...
9400 (get_edit_distance): ...this.
9401 (best_match::consider): Update for renaming.
9403 2018-06-12 Martin Sebor <msebor@redhat.com>
9405 PR tree-optimization/85259
9406 * builtins.c (compute_objsize): Handle constant offsets.
9407 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
9408 true iff a warning has been issued.
9409 * gimple.h (gimple_nonartificial_location): New function.
9410 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
9411 gimple_nonartificial_location and handle -Wno-system-headers.
9412 (handle_builtin_stxncpy): Same.
9414 2018-06-12 Martin Sebor <msebor@redhat.com>
9417 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
9419 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
9421 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
9422 BUILTIN_VEC_XST entries for pointer to double and long long.
9424 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
9427 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9429 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9432 2018-06-12 Martin Liska <mliska@suse.cz>
9434 * doc/options.texi: Document IntegerRange.
9436 2018-06-12 Martin Liska <mliska@suse.cz>
9438 * config/i386/i386.opt: Make MPX-related options as Deprecated.
9439 * opt-functions.awk: Handle Deprecated flag.
9440 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
9442 (read_cmdline_option): Report warning for a deprecated option.
9443 * opts.h (struct cl_option): Add new field cl_deprecated.
9444 (CL_ERR_DEPRECATED): New.
9446 2018-06-12 Martin Liska <mliska@suse.cz>
9448 * doc/options.texi: Document Deprecated option flag.
9450 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9452 * config/arc/arc-arch.h (arc_extras): New enum.
9453 (arc_cpu_t):Add field extra.
9454 (arc_cpu_types): Consider the extras.
9455 * config/arc/arc-cpus.def: Add extras info.
9456 * config/arc/arc-opts.h (processor_type): Consider extra field.
9457 * config/arc/arc.c (arc_override_options): Handle extra field.
9459 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
9461 * config/arc/arc-arch.h: Update ARC_OPTX macro.
9462 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
9464 * config/arc/arc.c (arc_init): Update pic warning.
9465 (irq_range): Update irq range parsing warnings.
9466 (arc_override_options): Update various warning messages.
9467 (arc_handle_aux_attribute): Likewise.
9469 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
9471 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
9473 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9475 * doc/sourcebuild.texi: Document usage of line number 0 in verify
9476 compiler messages directives.
9478 2018-06-12 Matthew Fortune <mfortune@gmail.com>
9480 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
9481 * config/mips/mips-tables.opt: Regenerate.
9482 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
9484 * doc/invoke.texi: Document -march=i6500.
9486 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
9488 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
9489 (i6400_gpmul): Add cpu_unit.
9490 (i6400_gpdiv): Likewise.
9491 (i6400_msa_add_d): Update reservations.
9492 (i6400_msa_int_add) Likewise.
9493 (i6400_msa_short_logic3) Likewise.
9494 (i6400_msa_short_logic2) Likewise.
9495 (i6400_msa_short_logic) Likewise.
9496 (i6400_msa_move) Likewise.
9497 (i6400_msa_cmp) Likewise.
9498 (i6400_msa_short_float2) Likewise.
9499 (i6400_msa_div_d) Likewise.
9500 (i6400_msa_long_logic1) Likewise.
9501 (i6400_msa_long_logic2) Likewise.
9502 (i6400_msa_mult) Likewise.
9503 (i6400_msa_long_float2) Likewise.
9504 (i6400_msa_long_float4) Likewise.
9505 (i6400_msa_long_float5) Likewise.
9506 (i6400_msa_long_float8) Likewise.
9507 (i6400_fpu_fadd): Include frint type.
9508 (i6400_fpu_store): New define_insn_reservation.
9509 (i6400_fpu_load): Likewise.
9510 (i6400_fpu_move): Likewise.
9511 (i6400_fpu_fcmp): Likewise.
9512 (i6400_fpu_fmadd): Likewise.
9513 (i6400_int_mult): Include imul3nc type and update reservation.
9514 (i6400_int_div): Include idiv3 type and update reservation.
9515 (i6400_int_load): Update to check type not move_type.
9516 (i6400_int_store): Likewise.
9517 (i6400_int_prefetch): Set zero latency.
9519 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
9521 * gcc.c: Document new %@{...} sequence.
9522 (LINK_COMMAND_SPEC): Use it for the -L switches.
9523 (cpp_unique_options): Use it for the -I switches.
9524 (at_file_argbuf): New global variable.
9525 (in_at_file): Likewise.
9526 (alloc_args): Create at_file_argbuf.
9527 (clear_args): Truncate at_file_argbuf.
9528 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
9529 (open_at_file): New function.
9530 (close_at_file): Likewise.
9531 (create_at_file): Delete.
9532 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
9535 (validate_switches_from_spec): Deal with %@{...} sequence.
9536 (validate_switches): Likewise.
9537 (driver::finalize): Call clear_args.
9539 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
9541 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
9543 2018-06-11 Martin Sebor <msebor@redhat.com>
9545 * doc/invoke.texi (-Wall): List -Wc++17-compat.
9546 (Wno-class-memaccess): Add @opindex.
9547 (Wno-templates, Wno-multiple-inheritance): Same.
9548 (Wno-virtual-inheritance, Wno-namespaces): Same.
9549 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
9550 (Wno-format-overflow, Wno-format-truncation): Same.
9551 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
9552 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
9553 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
9554 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
9555 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
9556 (Wno-misspelled-isr): Same.
9558 2018-06-11 Martin Sebor <msebor@redhat.com>
9560 * PR tree-optimization/86083
9561 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
9563 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
9565 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
9567 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
9570 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
9571 on the correct operand.
9572 (*movdi_internal64): Ditto.
9574 2018-06-11 Martin Liska <mliska@suse.cz>
9576 PR tree-optimization/86089
9577 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
9579 2018-06-11 Julia Koval <julia.koval@intel.com>
9581 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
9582 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
9583 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
9585 2018-06-11 Olivier Hainque <hainque@adacore.com>
9587 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
9588 for Ada with strict dwarf2.
9590 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
9593 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
9596 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
9598 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
9600 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
9602 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
9605 2018-06-08 Martin Liska <mliska@suse.cz>
9607 * tree-cfg.h (debug_function): Fix argument type to match
9610 2018-06-08 Martin Liska <mliska@suse.cz>
9612 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
9613 Remove usage of MPX-related (and removed) fields.
9614 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
9616 2018-06-08 David Malcolm <dmalcolm@redhat.com>
9618 * cfg.c (debug): Use TDF_NONE rather than 0.
9619 * cfghooks.c (debug): Likewise.
9620 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
9621 (struct dump_option_value_info): Convert to...
9622 (struct kv_pair): ...this template type.
9623 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
9625 (optinfo_verbosity_options): Likewise.
9626 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
9628 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
9629 than int for "optgroup_flags" param.
9630 (dump_generic_expr_loc): Use dump_flags_t rather than int for
9632 (dump_dec): Likewise.
9633 (dump_finish): Use TDF_NONE rather than 0.
9634 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
9635 rather than int for "optgroup_flags" param. Use TDF_NONE rather
9636 than 0. Update for change to option_ptr.
9637 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
9638 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
9639 0. Update for changes to optinfo_verbosity_options and
9641 (opt_info_switch_p): Convert optgroup_flags from int to
9643 (dump_basic_block): Use dump_flags_t rather than int
9644 for "dump_kind" param.
9645 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
9646 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
9647 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
9648 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
9649 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
9650 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
9651 TDF_NONE): Convert from macros to...
9652 (enum dump_flag): ...this new enum.
9653 (dump_flags_t): Update to use enum.
9654 (operator|, operator&, operator~, operator|=, operator&=):
9655 Implement for dump_flags_t.
9656 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
9657 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
9658 Convert from macros to...
9659 (enum optgroup_flag): ...this new enum.
9660 (optgroup_flags_t): New typedef.
9661 (operator|, operator|=): Implement for optgroup_flags_t.
9662 (struct dump_file_info): Convert field "alt_flags" to
9663 dump_flags_t. Convert field "optgroup_flags" to
9665 (dump_basic_block): Use dump_flags_t rather than int for param.
9666 (dump_generic_expr_loc): Likewise.
9667 (dump_dec): Likewise.
9668 (dump_register): Convert param "optgroup_flags" to
9670 (opt_info_enable_passes): Likewise.
9671 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
9673 * gimple-pretty-print.c (debug): Likewise.
9674 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
9675 (merged_store_group::apply_stores): Likewise.
9676 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
9677 * gimple.c (verify_gimple_pp): Likewise.
9678 * graphite-poly.c (print_pbb_body): Likewise.
9679 * passes.c (pass_manager::register_one_dump_file): Convert
9680 local "optgroup_flags" to optgroup_flags_t.
9681 * print-tree.c (print_node): Use TDF_NONE rather than 0.
9683 (debug_body): Likewise.
9684 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
9685 to optgroup_flags_t.
9686 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
9688 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
9689 (convert_mult_to_fma): Likewise.
9690 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
9691 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
9692 * tree-vect-data-refs.c (dump_lower_bound): Convert param
9693 "dump_kind" to dump_flags_t.
9695 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
9697 * config/rs6000/rs6000.c (min, max): Delete.
9699 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
9701 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
9702 -mabi=spe and -mabi=no-spe.
9704 2018-06-08 Martin Liska <mliska@suse.cz>
9706 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
9707 where we expect an existing summary.
9709 2018-06-08 Martin Liska <mliska@suse.cz>
9711 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
9712 * ipa-inline.h (estimate_edge_growth): Likewise.
9714 2018-06-08 Martin Liska <mliska@suse.cz>
9716 * cgraph.c (function_version_hasher::hash): Use
9717 cgraph_node::get_uid ().
9718 (function_version_hasher::equal):
9719 * cgraph.h (cgraph_node::get_uid): New method.
9720 * ipa-inline.c (update_caller_keys): Use
9721 cgraph_node::get_uid ().
9722 (update_callee_keys): Likewise.
9723 * ipa-utils.c (searchc): Likewise.
9724 (ipa_reduced_postorder): Likewise.
9725 * lto-cgraph.c (input_node): Likewise.
9726 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
9727 * symbol-summary.h (symtab_insertion): Likewise.
9728 (symtab_removal): Likewise.
9729 (symtab_duplication): Likewise.
9730 * tree-pretty-print.c (dump_function_header): Likewise.
9731 * tree-sra.c (convert_callers_for_node): Likewise.
9733 2018-06-08 Martin Liska <mliska@suse.cz>
9735 * cgraph.c (symbol_table::create_edge): Always assign a new
9737 (symbol_table::free_edge): Do not recycle numbers.
9738 * cgraph.h (cgraph_edge::get): New method.
9739 * symbol-summary.h (symtab_removal): Use it.
9740 (symtab_duplication): Likewise.
9741 (call_summary::hashable_uid): Remove.
9743 2018-06-08 Martin Liska <mliska@suse.cz>
9745 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
9746 (initialize_growth_caches): Remove.
9747 (free_growth_caches): Likewise.
9748 (do_estimate_edge_time): Use edge_growth_cache.
9749 (do_estimate_edge_size): Likewise.
9750 (do_estimate_edge_hints): Likewise.
9751 * ipa-inline.c (reset_edge_caches): Likewise.
9752 (recursive_inlining): Likewise.
9753 (inline_small_functions): Likewise.
9754 * ipa-inline.h (initialize_growth_caches): Remove.
9755 (estimate_edge_size): Likewise.
9756 (estimate_edge_time): Likewise.
9757 (estimate_edge_hints): Likewise.
9758 (reset_edge_growth_cache): Likewise.
9759 * symbol-summary.h (call_summary::remove): New method.
9761 2018-06-08 Martin Liska <mliska@suse.cz>
9763 * ipa-cp.c (class edge_clone_summary): New summary.
9764 (grow_edge_clone_vectors): Remove.
9765 (ipcp_edge_duplication_hook): Remove.
9766 (class edge_clone_summary_t): New call_summary class.
9767 (ipcp_edge_removal_hook): Remove.
9768 (edge_clone_summary_t::duplicate): New function.
9769 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
9770 (create_specialized_node): Likewise.
9771 (ipcp_driver): Initialize edge_clone_summaries and do not
9774 2018-06-08 Martin Liska <mliska@suse.cz>
9776 * symbol-summary.h (get): New function.
9777 (call_summary::m_initialize_when_cloning): New class member.
9779 2018-06-08 Martin Liska <mliska@suse.cz>
9781 * cgraph.c (cgraph_node::remove): Do not recycle uid.
9782 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
9783 (symbol_table::allocate_cgraph_symbol): Do not set uid.
9784 * passes.c (uid_hash_t): Record removed_nodes by their uids.
9785 (remove_cgraph_node_from_order): Use the removed_nodes set.
9786 (do_per_function_toporder): Likwise.
9787 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
9788 instead of summary_uid.
9789 (symtab_removal): Likewise.
9790 (symtab_duplication): Likewise.
9792 2018-06-08 Martin Liska <mliska@suse.cz>
9794 * ipa-cp.c (ipcp_store_bits_results): Use
9795 ipcp_transformation_sum.
9796 (ipcp_store_vr_results): Likewise.
9797 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
9799 (ipcp_transformation_initialize): ... this.
9800 (ipa_set_node_agg_value_chain):
9801 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
9802 (write_ipcp_transformation_info): Likewise.
9803 (read_ipcp_transformation_info): Likewise.
9804 (ipcp_update_bits): Likewise.
9805 (ipcp_update_vr): Likewise.
9806 (ipcp_transform_function): Likewise.
9807 * ipa-prop.h: Rename ipcp_transformation_summary to
9808 ipcp_transformation.
9809 (class ipcp_transformation_t): New function summary.
9810 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
9811 (ipa_get_agg_replacements_for_node): Likewise.
9813 2018-06-08 Martin Liska <mliska@suse.cz>
9815 * ipa-pure-const.c (struct funct_state_d): Do it class instead
9817 (class funct_state_summary_t): New function_summary class.
9818 (has_function_state): Remove.
9819 (get_function_state): Likewise.
9820 (set_function_state): Likewise.
9821 (add_new_function): Likewise.
9822 (funct_state_summary_t::insert): New function.
9823 (duplicate_node_data): Remove.
9824 (remove_node_data): Remove.
9825 (funct_state_summary_t::duplicate): New function.
9826 (register_hooks): Create new funct_state_summaries.
9827 (pure_const_generate_summary): Use it.
9828 (pure_const_write_summary): Likewise.
9829 (pure_const_read_summary): Likewise.
9830 (propagate_pure_const): Likewise.
9831 (propagate_nothrow): Likewise.
9832 (dump_malloc_lattice): Likewise.
9833 (propagate_malloc): Likewise.
9834 (execute): Do not register hooks, just remove summary
9836 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
9839 2018-06-08 Martin Liska <mliska@suse.cz>
9841 * ipa-reference.c (remove_node_data): Remove.
9842 (duplicate_node_data): Likewise.
9843 (class ipa_ref_var_info_summary_t): New class.
9844 (class ipa_ref_opt_summary_t): Likewise.
9845 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
9846 (get_reference_optimization_summary): Use
9847 ipa_ref_opt_sum_summaries.
9848 (set_reference_vars_info): Remove.
9849 (set_reference_optimization_summary): Likewise.
9850 (ipa_init): Create summaries.
9851 (init_function_info): Use function summary.
9852 (ipa_ref_opt_summary_t::duplicate): New function.
9853 (ipa_ref_opt_summary_t::remove): New function.
9854 (get_read_write_all_from_node): Fix GNU coding style.
9855 (propagate): Use function summary.
9856 (write_node_summary_p): Fix GNU coding style.
9857 (stream_out_bitmap): Likewise.
9858 (ipa_reference_read_optimization_summary): Use function summary.
9859 (ipa_reference_c_finalize): Do not release hooks.
9861 2018-06-08 Martin Liska <mliska@suse.cz>
9863 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
9864 (analyze_function_body): Extract multiple calls of get_create.
9865 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9866 * ipa-inline.c (recursive_inlining): Use ::get method.
9867 * ipa-inline.h (estimate_edge_growth): Likewise.
9869 2018-06-08 Martin Liska <mliska@suse.cz>
9871 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
9873 (hsa_function_summary::hsa_function_summary): Use the new enum
9875 (hsa_gpu_implementation_p): Use hsa_summaries::get.
9876 * hsa-gen.c (hsa_get_host_function): Likewise.
9877 (get_brig_function_name): Likewise.
9878 * ipa-hsa.c (process_hsa_functions): Likewise.
9879 (ipa_hsa_write_summary): Likewise.
9880 * symbol-summary.h (symtab_duplication): Use ::get function/
9881 (get): New function.
9883 2018-06-08 Martin Liska <mliska@suse.cz>
9885 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
9887 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
9888 (hsa_register_kernel): Likewise.
9889 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
9890 * hsa-gen.c (hsa_get_host_function): Likewise.
9891 (get_brig_function_name): Likewise.
9892 (generate_hsa): Likewise.
9893 (pass_gen_hsail::execute): Likewise.
9894 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
9895 (devirtualization_time_bonus): Likewise.
9896 (ipcp_propagate_stage): Likewise.
9897 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9898 (edge_set_predicate): Likewise.
9899 (evaluate_conditions_for_known_args): Likewise.
9900 (evaluate_properties_for_edge): Likewise.
9901 (ipa_fn_summary::reset): Likewise.
9902 (ipa_fn_summary_t::duplicate): Likewise.
9903 (dump_ipa_call_summary): Likewise.
9904 (ipa_dump_fn_summary): Likewise.
9905 (analyze_function_body): Likewise.
9906 (compute_fn_summary): Likewise.
9907 (estimate_edge_devirt_benefit): Likewise.
9908 (estimate_edge_size_and_time): Likewise.
9909 (estimate_calls_size_and_time): Likewise.
9910 (estimate_node_size_and_time): Likewise.
9911 (inline_update_callee_summaries): Likewise.
9912 (remap_edge_change_prob): Likewise.
9913 (remap_edge_summaries): Likewise.
9914 (ipa_merge_fn_summary_after_inlining): Likewise.
9915 (ipa_update_overall_fn_summary): Likewise.
9916 (read_ipa_call_summary): Likewise.
9917 (inline_read_section): Likewise.
9918 (write_ipa_call_summary): Likewise.
9919 (ipa_fn_summary_write): Likewise.
9920 (ipa_free_fn_summary): Likewise.
9921 * ipa-hsa.c (process_hsa_functions): Likewise.
9922 (ipa_hsa_write_summary): Likewise.
9923 (ipa_hsa_read_section): Likewise.
9924 * ipa-icf.c (sem_function::merge): Likewise.
9925 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9926 (do_estimate_edge_time): Likewise.
9927 (estimate_size_after_inlining): Likewise.
9928 (estimate_growth): Likewise.
9929 (growth_likely_positive): Likewise.
9930 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9931 (inline_call): Likewise.
9932 * ipa-inline.c (caller_growth_limits): Likewise.
9933 (can_inline_edge_p): Likewise.
9934 (can_inline_edge_by_limits_p): Likewise.
9935 (compute_uninlined_call_time): Likewise.
9936 (compute_inlined_call_time): Likewise.
9937 (want_inline_small_function_p): Likewise.
9938 (edge_badness): Likewise.
9939 (update_caller_keys): Likewise.
9940 (update_callee_keys): Likewise.
9941 (recursive_inlining): Likewise.
9942 (inline_small_functions): Likewise.
9943 (inline_to_all_callers_1): Likewise.
9944 (dump_overall_stats): Likewise.
9945 (early_inline_small_functions): Likewise.
9946 (early_inliner): Likewise.
9947 * ipa-inline.h (estimate_edge_growth): Likewise.
9948 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
9949 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9950 * ipa-prop.h (IPA_NODE_REF): Likewise.
9951 (IPA_EDGE_REF): Likewise.
9952 * ipa-pure-const.c (malloc_candidate_p): Likewise.
9953 (propagate_malloc): Likewise.
9954 * ipa-split.c (execute_split_functions): Likewise.
9955 * symbol-summary.h: Rename get to get_create.
9957 (get_create): Likewise.
9958 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
9960 2018-06-08 Martin Liska <mliska@suse.cz>
9962 * symbol-summary.h (release): Move definition out of class
9964 (symtab_removal): Likewise.
9965 (symtab_duplication): Likewise.
9967 2018-06-08 Martin Liska <mliska@suse.cz>
9969 * symbol-summary.h (function_summary): Move constructor
9970 implementation out of class declaration.
9971 (release): Likewise.
9972 (symtab_insertion): Likewise.
9973 (symtab_removal): Likewise.
9974 (symtab_duplication): Likewise.
9977 2018-06-08 Martin Liska <mliska@suse.cz>
9979 * Makefile.in: Remove support for MPX (macros, related functions,
9980 fields in cgraph_node, ...).
9981 * builtin-types.def (BT_BND): Likewise.
9982 (BT_FN_BND_CONST_PTR): Likewise.
9983 (BT_FN_CONST_PTR_BND): Likewise.
9984 (BT_FN_VOID_PTR_BND): Likewise.
9985 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
9986 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
9987 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
9988 (expand_builtin_mempcpy_with_bounds): Likewise.
9989 (expand_builtin_memset_with_bounds): Likewise.
9990 (expand_builtin_memset_args): Likewise.
9991 (std_expand_builtin_va_start): Likewise.
9992 (expand_builtin): Likewise.
9993 (expand_builtin_with_bounds): Likewise.
9994 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
9995 (DEF_LIB_BUILTIN_CHKP): Likewise.
9996 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
9997 (DEF_CHKP_BUILTIN): Likewise.
9998 (BUILT_IN_MEMCPY): Likewise.
9999 (BUILT_IN_MEMMOVE): Likewise.
10000 (BUILT_IN_MEMPCPY): Likewise.
10001 (BUILT_IN_MEMSET): Likewise.
10002 (BUILT_IN_STPCPY): Likewise.
10003 (BUILT_IN_STRCAT): Likewise.
10004 (BUILT_IN_STRCHR): Likewise.
10005 (BUILT_IN_STRCPY): Likewise.
10006 (BUILT_IN_STRLEN): Likewise.
10007 (BUILT_IN_MEMCPY_CHK): Likewise.
10008 (BUILT_IN_MEMMOVE_CHK): Likewise.
10009 (BUILT_IN_MEMPCPY_CHK): Likewise.
10010 (BUILT_IN_MEMSET_CHK): Likewise.
10011 (BUILT_IN_STPCPY_CHK): Likewise.
10012 (BUILT_IN_STRCAT_CHK): Likewise.
10013 (BUILT_IN_STRCPY_CHK): Likewise.
10014 * calls.c (store_bounds): Likewise.
10015 (emit_call_1): Likewise.
10016 (special_function_p): Likewise.
10017 (maybe_warn_nonstring_arg): Likewise.
10018 (initialize_argument_information): Likewise.
10019 (finalize_must_preallocate): Likewise.
10020 (compute_argument_addresses): Likewise.
10021 (expand_call): Likewise.
10022 * cfgexpand.c (expand_call_stmt): Likewise.
10023 (expand_return): Likewise.
10024 (expand_gimple_stmt_1): Likewise.
10025 (pass_expand::execute): Likewise.
10026 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
10027 (cgraph_node::remove): Likewise.
10028 (cgraph_node::dump): Likewise.
10029 (cgraph_node::verify_node): Likewise.
10030 * cgraph.h (chkp_function_instrumented_p): Likewise.
10031 (symtab_node::get_alias_target): Likewise.
10032 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
10033 (cgraph_local_p): Likewise.
10034 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
10035 (cgraph_edge::rebuild_references): Likewise.
10036 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
10037 (walk_polymorphic_call_targets): Likewise.
10038 (cgraph_node::expand_thunk): Likewise.
10039 (symbol_table::output_weakrefs): Likewise.
10040 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
10041 (ix86_handle_option): Likewise.
10042 * config/i386/constraints.md: Likewise.
10043 * config/i386/i386-builtin-types.def (BND): Likewise.
10047 * config/i386/i386-builtin.def (BDESC_END): Likewise.
10048 (BDESC_FIRST): Likewise.
10050 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
10051 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
10052 * config/i386/i386.c (enum reg_class): Likewise.
10053 (ix86_target_string): Likewise.
10054 (ix86_option_override_internal): Likewise.
10055 (ix86_conditional_register_usage): Likewise.
10056 (ix86_valid_target_attribute_inner_p): Likewise.
10057 (ix86_set_indirect_branch_type): Likewise.
10058 (ix86_set_current_function): Likewise.
10059 (ix86_function_arg_regno_p): Likewise.
10060 (init_cumulative_args): Likewise.
10061 (ix86_function_arg_advance): Likewise.
10062 (ix86_function_arg): Likewise.
10063 (ix86_pass_by_reference): Likewise.
10064 (ix86_function_value_regno_p): Likewise.
10065 (ix86_function_value_1): Likewise.
10066 (ix86_function_value_bounds): Likewise.
10067 (ix86_return_in_memory): Likewise.
10068 (ix86_setup_incoming_vararg_bounds): Likewise.
10069 (ix86_va_start): Likewise.
10070 (indirect_thunk_need_prefix): Likewise.
10071 (print_reg): Likewise.
10072 (ix86_print_operand): Likewise.
10073 (ix86_expand_call): Likewise.
10074 (ix86_output_function_return): Likewise.
10075 (reg_encoded_number): Likewise.
10076 (BDESC_VERIFYS): Likewise.
10077 (ix86_init_mpx_builtins): Likewise.
10078 (ix86_init_builtins): Likewise.
10079 (ix86_emit_cmove): Likewise.
10080 (ix86_emit_move_max): Likewise.
10081 (ix86_expand_builtin): Likewise.
10082 (ix86_builtin_mpx_function): Likewise.
10083 (ix86_get_arg_address_for_bt): Likewise.
10084 (ix86_load_bounds): Likewise.
10085 (ix86_store_bounds): Likewise.
10086 (ix86_load_returned_bounds): Likewise.
10087 (ix86_store_returned_bounds): Likewise.
10088 (ix86_class_likely_spilled_p): Likewise.
10089 (ix86_hard_regno_mode_ok): Likewise.
10090 (x86_order_regs_for_local_alloc): Likewise.
10091 (ix86_mitigate_rop): Likewise.
10092 (ix86_bnd_prefixed_insn_p): Likewise.
10093 (ix86_mpx_bound_mode): Likewise.
10094 (ix86_make_bounds_constant): Likewise.
10095 (ix86_initialize_bounds): Likewise.
10096 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
10097 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
10098 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
10099 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
10100 (TARGET_CHKP_BOUND_MODE): Likewise.
10101 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
10102 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
10103 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
10104 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
10105 * config/i386/i386.h (TARGET_MPX): Likewise.
10106 (TARGET_MPX_P): Likewise.
10107 (VALID_BND_REG_MODE): Likewise.
10108 (FIRST_BND_REG): Likewise.
10109 (LAST_BND_REG): Likewise.
10110 (enum reg_class): Likewise.
10111 (BND_REG_P): Likewise.
10112 (BND_REGNO_P): Likewise.
10113 (BNDmode): Likewise.
10114 (ADJUST_INSN_LENGTH): Likewise.
10115 * config/i386/i386.md: Likewise.
10116 * config/i386/i386.opt: Likewise.
10117 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
10118 (defined): Likewise.
10119 (LINK_MPX): Likewise.
10120 (MPX_SPEC): Likewise.
10121 (LIBMPX_SPEC): Likewise.
10122 (LIBMPXWRAPPERS_SPEC): Likewise.
10123 (CHKP_SPEC): Likewise.
10124 * config/i386/predicates.md: Likewise.
10125 * dbxout.c (dbxout_type): Likewise.
10126 * doc/extend.texi: Likewise.
10127 * doc/invoke.texi: Likewise.
10128 * doc/md.texi: Likewise.
10129 * doc/tm.texi: Likewise.
10130 * doc/tm.texi.in: Likewise.
10131 * dwarf2out.c (is_base_type): Likewise.
10132 (gen_formal_types_die): Likewise.
10133 (gen_subprogram_die): Likewise.
10134 (gen_type_die_with_usage): Likewise.
10135 (gen_decl_die): Likewise.
10136 (dwarf2out_late_global_decl): Likewise.
10137 * expr.c (expand_assignment): Likewise.
10138 (emit_storent_insn): Likewise.
10139 (store_expr_with_bounds): Likewise.
10140 (store_expr): Likewise.
10141 (expand_expr_real_1): Likewise.
10142 * expr.h (store_expr_with_bounds): Likewise.
10143 * function.c (use_register_for_decl): Likewise.
10144 (struct bounds_parm_data): Likewise.
10145 (assign_parms_augmented_arg_list): Likewise.
10146 (assign_parm_find_entry_rtl): Likewise.
10147 (assign_parm_is_stack_parm): Likewise.
10148 (assign_parm_load_bounds): Likewise.
10149 (assign_bounds): Likewise.
10150 (assign_parms): Likewise.
10151 (expand_function_start): Likewise.
10152 * gcc.c (CHKP_SPEC): Likewise.
10153 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
10154 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
10155 (wrestrict_dom_walker::check_call): Likewise.
10156 * gimple.c (gimple_build_call_from_tree): Likewise.
10157 * gimple.h (enum gf_mask): Likewise.
10158 (gimple_call_with_bounds_p): Likewise.
10159 (gimple_call_set_with_bounds): Likewise.
10160 * gimplify.c (gimplify_init_constructor): Likewise.
10161 * ipa-cp.c (initialize_node_lattices): Likewise.
10162 (propagate_constants_across_call): Likewise.
10163 (find_more_scalar_values_for_callers_subset): Likewise.
10164 * ipa-hsa.c (process_hsa_functions): Likewise.
10165 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
10166 * ipa-icf.c (sem_function::merge): Likewise.
10167 * ipa-inline.c (early_inliner): Likewise.
10168 * ipa-pure-const.c (warn_function_noreturn): Likewise.
10169 (warn_function_cold): Likewise.
10170 (propagate_pure_const): Likewise.
10171 * ipa-ref.h (enum GTY): Likewise.
10172 * ipa-split.c (find_retbnd): Likewise.
10173 (consider_split): Likewise.
10174 (split_function): Likewise.
10175 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10176 * ipa.c (walk_polymorphic_call_targets): Likewise.
10177 (symbol_table::remove_unreachable_nodes): Likewise.
10178 (process_references): Likewise.
10179 (cgraph_build_static_cdtor_1): Likewise.
10180 * lto-cgraph.c (lto_output_node): Likewise.
10181 (output_refs): Likewise.
10182 (compute_ltrans_boundary): Likewise.
10183 (input_overwrite_node): Likewise.
10184 (input_node): Likewise.
10185 (input_cgraph_1): Likewise.
10186 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
10187 * passes.c (pass_manager::execute_early_local_passes): Likewise.
10188 (class pass_chkp_instrumentation_passes): Likewise.
10189 (make_pass_chkp_instrumentation_passes): Likewise.
10190 * passes.def: Likewise.
10191 * rtl.h (struct GTY): Likewise.
10192 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
10193 * stor-layout.c (layout_type): Likewise.
10194 * symtab.c: Likewise.
10195 * target.def: Likewise.
10196 * targhooks.c (default_chkp_bound_type): Likewise.
10197 (default_chkp_bound_mode): Likewise.
10198 (default_builtin_chkp_function): Likewise.
10199 (default_chkp_function_value_bounds): Likewise.
10200 (default_chkp_make_bounds_constant): Likewise.
10201 (default_chkp_initialize_bounds): Likewise.
10202 * targhooks.h (default_chkp_bound_type): Likewise.
10203 (default_chkp_bound_mode): Likewise.
10204 (default_builtin_chkp_function): Likewise.
10205 (default_chkp_function_value_bounds): Likewise.
10206 (default_chkp_make_bounds_constant): Likewise.
10207 (default_chkp_initialize_bounds): Likewise.
10208 * toplev.c (compile_file): Likewise.
10209 (process_options): Likewise.
10210 * tree-core.h (DEF_BUILTIN): Likewise.
10211 (DEF_BUILTIN_CHKP): Likewise.
10212 * tree-inline.c (declare_return_variable): Likewise.
10213 (remap_gimple_stmt): Likewise.
10214 (copy_bb): Likewise.
10215 (initialize_inlined_parameters): Likewise.
10216 (expand_call_inline): Likewise.
10217 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
10218 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
10219 (make_pass_ipa_chkp_produce_thunks): Likewise.
10220 (make_pass_chkp): Likewise.
10221 (make_pass_chkp_opt): Likewise.
10222 (make_pass_chkp_instrumentation_passes): Likewise.
10223 * tree-pretty-print.c (dump_generic_node): Likewise.
10224 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
10225 * tree-ssa-dce.c (propagate_necessity): Likewise.
10226 (eliminate_unnecessary_stmts): Likewise.
10227 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
10228 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
10229 * tree-ssa-sccvn.h: Likewise.
10230 * tree-ssa-strlen.c (get_string_length): Likewise.
10231 (valid_builtin_call): Likewise.
10232 (adjust_last_stmt): Likewise.
10233 (handle_builtin_strchr): Likewise.
10234 (handle_builtin_strcpy): Likewise.
10235 (handle_builtin_stxncpy): Likewise.
10236 (handle_builtin_memcpy): Likewise.
10237 (handle_builtin_strcat): Likewise.
10238 (strlen_check_and_optimize_stmt): Likewise.
10239 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
10240 * tree-streamer-in.c: Likewise.
10241 * tree-streamer.c (record_common_node): Likewise.
10242 * tree.c (tree_code_size): Likewise.
10243 (wide_int_to_tree_1): Likewise.
10244 (type_contains_placeholder_1): Likewise.
10245 (build_common_tree_nodes): Likewise.
10246 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
10247 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
10248 (POINTER_BOUNDS_P): Likewise.
10249 (BOUNDED_TYPE_P): Likewise.
10250 (BOUNDED_P): Likewise.
10251 (CALL_WITH_BOUNDS_P): Likewise.
10252 (pointer_bounds_type_node): Likewise.
10253 * value-prof.c (gimple_ic): Likewise.
10254 * var-tracking.c (vt_add_function_parameters): Likewise.
10255 * varasm.c (make_decl_rtl): Likewise.
10256 (assemble_start_function): Likewise.
10257 (output_constant): Likewise.
10258 (maybe_assemble_visibility): Likewise.
10259 * varpool.c (ctor_for_folding): Likewise.
10260 * chkp-builtins.def: Remove.
10261 * ipa-chkp.c: Remove.
10262 * ipa-chkp.h: Remove.
10263 * rtl-chkp.c: Remove.
10264 * rtl-chkp.h: Remove.
10265 * tree-chkp-opt.c: Remove.
10266 * tree-chkp.c: Remove.
10267 * tree-chkp.h: Remove.
10269 2018-06-07 Carl Love <cel@us.ibm.com>
10271 * config/rs6000/vsx.md (vextract_fp_from_shorth,
10272 vextract_fp_from_shortl): Add BE support.
10274 2018-06-07 Paul Koning <ni1d@arrl.net>
10276 * compare-elim.c (try_merge_compare): Don't merge compare if
10277 address contains a side effect.
10278 (try_eliminate_compare): Likewise.
10280 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
10282 * config.gcc: Support "tremont".
10283 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
10284 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10286 * config/i386/i386.c (m_TREMONT): Define.
10287 (processor_target_table): Add "tremont".
10288 (PTA_TREMONT): Define.
10289 (ix86_lea_outperforms): Add TARGET_TREMONT.
10290 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
10291 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
10292 and M_INTEL_GOLDMONT_PLUS.
10293 (fold_builtin_cpu): Add "tremont".
10294 (ix86_add_stmt_cost): Add TARGET_TREMONT.
10295 (ix86_option_override_internal): Add "tremont".
10296 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
10297 (processor_type): Add PROCESSOR_TREMONT.
10298 * config/i386/x86-tune.def: Add m_TREMONT.
10299 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
10301 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10303 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
10304 symbol defined for msp430i* devices to be lower case.
10306 2018-06-07 Richard Biener <rguenther@suse.de>
10308 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
10309 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
10310 Properly wrap signed arithmetic if overflow wraps.
10312 2018-06-07 Jakub Jelinek <jakub@redhat.com>
10314 PR tree-optimization/69615
10315 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
10316 of a cast from a same precision integral SSA_NAME in a bb dominated
10317 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
10318 cast to utype if rhs2 has already a compatible type.
10320 2018-06-07 Richard Biener <rguenther@suse.de>
10322 PR tree-optimization/85935
10323 * graphite-scop-detection.c (find_params_in_bb): Analyze
10324 condition operands with respect to the correct loop. Assert
10325 the analysis doesn't fail.
10327 2018-06-04 Carl Love <cel@us.ibm.com>
10329 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
10330 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
10331 as it is slightly cheaper.
10332 (first_match_or_eos_index_<mode>):
10333 Calculate index using natural element order.
10334 (first_match_index_<mode>):
10335 Calculate index using natural element order.
10336 (first_match_or_eos_index_<mode>):
10337 Calculate index using natural order.
10338 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
10339 for BE and LE modes.
10340 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
10341 P9V_BUILTIN_VCLZLSBB_V16QI.
10342 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
10345 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10347 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
10348 indentation and line wrap for many prototypes. Add missing
10349 @smallexample directives around block of prototypes for vec_xl and
10352 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
10354 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
10355 track if we pass or return IEEE 128-bit floating point.
10356 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
10357 C++ mangling that is compatible with GCC 8.1.
10358 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
10359 (init_cumulative_args): Note if we pass or return IEEE 128-bit
10360 floating point types.
10361 (rs6000_function_arg_advance_1): Likewise.
10362 (rs6000_mangle_type): Optionally generate mangled names that match
10363 what GCC 8.1 generated for IEEE 128-bit floating point types.
10364 (rs6000_globalize_decl_name): If we have an external function that
10365 passes or returns IEEE 128-bit types, generate a weak reference
10366 from the mangled name used in GCC 8.1 to the current mangled
10368 (rs6000_init_builtins): Make __ibm128 use the long double type if
10369 long double is IBM extended double. Make __float128 use the long
10370 double type if long double is IEEE 128-bit.
10373 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
10374 macro for __ibm128 built-in functions.
10375 (PACK_IF): Add __ibm128 pack/unpack functions.
10376 (UNPACK_IF): Likewise.
10377 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
10378 enable long double built-in functions if long double is IEEE
10379 128-bit floating point.
10380 (rs6000_invalid_builtin): Update long double built-in function
10382 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
10383 functions, adjust the built-in function to use the long double
10384 built-in function if __ibm128 and long double are the same type.
10385 * doc/extend.texi (PowerPC builtins): Update documention for
10386 __builtin_{,un}pack_longdouble. Add documentation for
10387 __builtin_{,un}pack_ibm128.
10389 2018-06-06 Jim Wilson <jimw@sifive.com>
10391 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
10392 (struct machine_function): New field interrupt_mode.
10393 (riscv_handle_type_attribute): New function. Add forward declaration.
10394 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
10395 (riscv_expand_epilogue): Check interrupt_mode field.
10396 (riscv_set_current_function): Check interrupt attribute args and
10397 set interrupt_mode field.
10398 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
10399 (riscv_sret, riscv_uret): New.
10400 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
10401 new arguments to interrupt attribute.
10403 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
10406 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
10407 Don't handle -mcpu=power8 if -mpower9-vector is also used.
10409 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
10411 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
10412 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
10413 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10414 several redundant entries.
10416 2018-06-06 David Malcolm <dmalcolm@redhat.com>
10418 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
10419 type from "rtx" to "rtx_insn *".
10420 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
10421 for local "call_insn", removing cast.
10422 (ix86_expand_call): Likewise, introducing a "call_insn" local.
10424 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
10426 PR tree-optimization/86066
10427 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
10428 for BIT_INSERT_EXPR stores.
10430 2018-06-06 Richard Biener <rguenther@suse.de>
10432 PR tree-optimization/86062
10433 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10434 component refs ontop
10435 of to be offsetted base.
10437 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10439 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
10440 to be static and remove check on interrupt attribute name.
10442 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
10444 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10445 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
10447 2018-06-05 Steve Ellcey <sellcey@cavium.com>
10450 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
10452 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
10453 Remove second argument, change how error is called.
10454 (aarch64_layout_arg): Remove second argument from
10455 aarch64_err_no_fpadvsimd call.
10456 (aarch64_init_cumulative_args): Ditto.
10457 (aarch64_gimplify_va_arg_expr): Ditto.
10458 * config/aarch64/aarch64.md (mov<mode>): Ditto.
10460 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
10462 * config/i386/i386.md (simple_return_indirect_internal): New expander.
10463 (*simple_return_indirect_internal<mode>): Rename from
10464 simple_return_indirect_internal. Use W mode iterator.
10465 (rstorssp): New expander.
10466 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
10467 (clrssbsy): New expander.
10468 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
10470 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
10472 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
10474 (cmse_check_pointed_object): Likewise.
10476 2018-06-05 Martin Liska <mliska@suse.cz>
10478 PR gcov-profile/47618
10479 * doc/invoke.texi: Document how -fprofile-dir format
10482 2018-06-05 Richard Biener <rguenther@suse.de>
10484 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
10485 removal pretend DOM info isn't available so we do not update
10486 it and only remove edges, not dominated blocks. Actually free
10487 DOM info in case we removed something. Remove unreachable blocks.
10488 (mfb_keep_latches): Work with either DOM info or marked backedges.
10489 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
10490 first. Mark backedges if DOM info isn't available.
10491 (Re-)compute DOM info after cleanup_control_flow_pre.
10493 2018-06-05 Richard Biener <rguenther@suse.de>
10495 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
10496 (locus_discrim_hasher::hash): Adjust.
10497 (locus_discrim_hasher::equal): Likewise.
10498 (next_discriminator_for_locus): Work on line directly.
10499 (same_line_p): Pass in expanded locus1 as well.
10500 (assign_discriminators): Avoid redundant location expansions.
10502 2018-06-05 Richard Biener <rguenther@suse.de>
10504 PR tree-optimization/86046
10505 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
10506 if required after clearing TREE_ADDRESSABLE.
10508 2018-06-05 Richard Biener <rguenther@suse.de>
10510 PR tree-optimization/86047
10511 * tree-ssa-loop.c (for_each_index): Glob handling of all
10512 decls and constants and really handle all of them.
10514 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10517 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
10518 qualifier_void_pointer and qualifier_const_void_pointer.
10519 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
10520 (arm_init_builtins): Handle the above.
10521 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
10522 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
10523 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
10526 2018-06-05 Martin Liska <mliska@suse.cz>
10528 * auto-profile.c (read_autofdo_file): Do not use
10529 gcov_ctr_summary struct.
10530 (afdo_callsite_hot_enough_for_early_inline): Likewise.
10531 * coverage.c (struct counts_entry): Likewise.
10532 (read_counts_file): Read just single summary entry.
10533 (get_coverage_counts): Use gcov_summary struct.
10534 * coverage.h (get_coverage_counts): Likewise.
10535 * gcov-dump.c (dump_working_sets): Likewise.
10536 (tag_summary): Dump just single summary.
10537 * gcov-io.c (gcov_write_summary): Write just histogram
10539 (gcov_read_summary): Read just single summary.
10540 (compute_working_sets): Use gcov_summary struct.
10541 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
10542 of GCOV_COUNTERS_SUMMABLE.
10543 (GCOV_COUNTERS_SUMMABLE): Remove.
10544 (GCOV_FIRST_VALUE_COUNTER): Replace with
10545 GCOV_COUNTER_V_INTERVAL.
10546 (struct gcov_ctr_summary): Remove.
10547 (struct gcov_summary): Directly use fields of former
10549 (compute_working_sets): Use gcov_summary struct.
10550 * gcov.c (read_count_file): Do not use ctrs fields.
10551 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
10553 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
10555 * profile.c: Likewise.
10556 * profile.h: Likewise.
10558 2018-06-05 Martin Liska <mliska@suse.cz>
10560 PR gcov-profile/84846
10561 * gcov.c (output_lines): Print working directory only
10562 in intermediate format.
10564 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
10566 * config/s390/s390-builtin-types.def: Add void function type.
10567 * config/s390/s390-builtins.def: Use the function type for the
10570 2018-06-04 Jim Wilson <jimw@sifive.com>
10572 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
10574 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
10575 and maybe_eh_return. Change regno to unsigned int. Use new args to
10576 handle EH_RETURN_DATA_REGNO registers properly.
10577 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
10578 (riscv_expand_epilogue): Update comment. Change argument name and
10579 type. Update code to use new name and type. Pass new args to
10580 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
10582 * config/riscv/riscv.md (NORMAL_RETURN): New.
10583 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
10584 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
10585 (eh_return): Call gen_eh_return_internal and emit barrier.
10586 (eh_return_internal): Call riscv_expand_epilogue.
10588 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
10590 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
10591 bit_insertion field and declare can_be_merged_into method.
10592 (merged_store_group::can_be_merged_into): New method.
10593 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
10594 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
10595 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
10597 2018-06-04 Richard Biener <rguenther@suse.de>
10599 PR tree-optimization/85955
10600 * builtins.c (fold_builtin_sincos): Convert pointers to
10601 destination to appropriate type before dereferencing.
10603 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
10605 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
10607 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
10609 * expr.c (expand_expr_real_1): Force the operand into memory if
10610 its TYPE_MODE is BLKmode and if there is no integer mode for
10611 the number of bits being extracted.
10613 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10617 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10618 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
10620 2018-06-04 Richard Biener <rguenther@suse.de>
10622 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
10623 (cleanup_tree_cfg_noloop): ... single caller. Do
10624 start_recording_case_labels later.
10626 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
10628 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
10629 to _IMMINTRIN_H_INCLUDED.
10630 * config/i386/pconfigintrin.h: Ditto.
10631 * config/i386/waitpkgintrin.h: Ditto.
10632 * config/i386/immintrin.h: Add includes for sgxintrin.h,
10633 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
10634 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
10635 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
10636 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
10637 waitpkgintrin.h and cldemoteintrin.h.
10639 2018-06-04 Richard Biener <rguenther@suse.de>
10641 PR tree-optimization/86038
10642 * tracer.c (find_best_successor): Check probability for
10643 being initialized, bail out if not.
10645 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
10648 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
10649 of bits to ignore when comparing architectures.
10651 2018-06-04 Jakub Jelinek <jakub@redhat.com>
10653 PR tree-optimization/69615
10654 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
10655 maximum or minimum of the type, try to merge it also as if
10656 range1 is + [-, x - 1] or + [x + 1, -].
10659 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
10661 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
10663 PR tree-optimization/86034
10664 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
10665 the unsigned bitfield type in a bit insertion sequence if it does not
10666 have a larger precision than the bitfield size.
10667 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
10669 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
10671 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
10673 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
10675 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
10676 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
10677 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
10678 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
10680 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
10682 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
10683 Disable -fdelete-null-pointer-checks for ELF toolchain.
10685 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
10686 Kito Cheng <kito.cheng@gmail.com>
10688 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
10689 (nds32le-*-*, nds32be-*-*): Integrate checking process.
10690 (nds32*-*-*): Add glibc and uclibc conditions.
10691 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
10692 (TARGET_EXCEPT_UNWIND_INFO): Define.
10693 * config/nds32/elf.h: New file.
10694 * config/nds32/linux.h: New file.
10695 * config/nds32/nds32-elf.opt: New file.
10696 * config/nds32/nds32-linux.opt: New file.
10697 * config/nds32/nds32-fp-as-gp.c
10698 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
10699 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
10701 (nds32_asm_file_end): Ditto.
10702 (nds32_print_operand): Ditto.
10703 (nds32_insert_attributes): Ditto.
10704 (nds32_init_libfuncs): New function.
10705 (TARGET_HAVE_TLS): Define.
10706 (TARGET_INIT_LIBFUNCS): Define.
10707 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
10709 (TARGET_ELF): Apply different mcmodel setting.
10710 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
10711 been migrated into elf.h and linux.h files.
10712 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
10713 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
10714 (mcmodel): The content has been migrated into nds32-elf.opt and
10715 nds32-linux.opt files.
10716 * config/nds32/t-elf: New file.
10717 * config/nds32/t-linux: New file.
10719 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
10720 Shiva Chen <shiva0217@gmail.com>
10722 * config/nds32/constants.md (unspec_volatile_element): Add
10723 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
10724 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
10726 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
10727 (make_pass_nds32_fp_as_gp): Declare.
10728 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
10730 (nds32_asm_function_end_prologue): Remove unused asm output.
10731 (nds32_asm_function_begin_epilogue): Remove unused asm output.
10732 (nds32_asm_file_start): Output necessary fp_as_gp information.
10733 (nds32_option_override): Adjust register usage.
10734 (nds32_expand_prologue): Consider fp_as_gp situation.
10735 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
10736 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
10739 (simple_return): Ditto.
10740 (omit_fp_begin): Output special directive for fp_as_gp.
10741 (omit_fp_end): Output special directive for fp_as_gp.
10742 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
10743 mforbid-fp-as-gp): New options.
10745 2018-06-01 Mark Wielaard <mark@klomp.org>
10747 * dwarf2out.c (dwarf2out_finish): Remove generation of
10748 DW_AT_loclists_base.
10750 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
10752 * gimple-ssa-store-merging.c: Include gimple-fold.h.
10753 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
10754 (struct merged_store_group): Add bit_insertion field.
10755 (dump_char_array): Use standard hexadecimal format.
10756 (merged_store_group::merged_store_group): Set bit_insertion to false.
10757 (merged_store_group::apply_stores): Use optimal buffer size. Deal
10758 with BIT_INSERT_EXPR stores. Move up code updating the mask and
10759 also print the mask in the dump file.
10760 (pass_store_merging::gate): Minor tweak.
10761 (imm_store_chain_info::coalesce_immediate): Fix wrong association
10762 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
10763 stores with INTEGER_CST stores.
10764 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
10765 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
10766 and use it throughout. Generate bit insertion sequences if need be.
10767 (pass_store_merging::process_store): Remove redundant condition.
10768 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
10770 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
10772 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
10773 the 128-bit floating point types. Fix function comment.
10775 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10777 * config/aarch64/aarch64-simd.md
10778 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
10780 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
10783 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10785 PR tree-optimization/85989
10786 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
10788 (backprop::intersect_uses): Check it when deciding whether this
10789 is a backedge reference.
10790 (backprop::process_block): Add each phi to m_visited_phis
10791 after visiting it, then clear it at the end.
10793 2018-06-01 Richard Biener <rguenther@suse.de>
10795 * tree-vectorizer.h (vect_dr_stmt): New function.
10796 (vect_get_load_cost): Adjust.
10797 (vect_get_store_cost): Likewise.
10798 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
10799 Use vect_dr_stmt instead of DR_SMTT.
10800 (vect_record_base_alignments): Likewise.
10801 (vect_calculate_target_alignment): Likewise.
10802 (vect_compute_data_ref_alignment): Likewise and make static.
10803 (vect_update_misalignment_for_peel): Likewise.
10804 (vect_verify_datarefs_alignment): Likewise.
10805 (vector_alignment_reachable_p): Likewise.
10806 (vect_get_data_access_cost): Likewise. Pass down
10807 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
10808 (vect_get_peeling_costs_all_drs): Likewise.
10809 (vect_peeling_hash_get_lowest_cost): Likewise.
10810 (vect_enhance_data_refs_alignment): Likewise.
10811 (vect_find_same_alignment_drs): Likewise.
10812 (vect_analyze_data_refs_alignment): Likewise.
10813 (vect_analyze_group_access_1): Likewise.
10814 (vect_analyze_group_access): Likewise.
10815 (vect_analyze_data_ref_access): Likewise.
10816 (vect_analyze_data_ref_accesses): Likewise.
10817 (vect_vfa_segment_size): Likewise.
10818 (vect_small_gap_p): Likewise.
10819 (vectorizable_with_step_bound_p): Likewise.
10820 (vect_prune_runtime_alias_test_list): Likewise.
10821 (vect_analyze_data_refs): Likewise.
10822 (vect_supportable_dr_alignment): Likewise.
10823 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10824 (vect_gen_prolog_loop_niters): Likewise.
10825 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10826 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
10828 (vect_recog_mask_conversion_pattern): Likewise.
10829 (vect_try_gather_scatter_pattern): Likewise.
10830 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
10831 to vect_get_store_cost.
10832 (vect_get_store_cost): Get stmt_info instead of DR.
10833 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
10834 (vect_get_load_cost): Get stmt_info instead of DR.
10836 2018-06-01 Richard Biener <rguenther@suse.de>
10838 PR middle-end/86017
10839 * gimple-fold.c (var_decl_component_p): Also allow offsetted
10840 vars wrapped in MEM_REFs.
10842 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
10844 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10845 Fix subreg tests so that we only return a choice between
10846 GENERAL_REGS and FP_REGS if the original classes included both.
10848 2018-06-01 Richard Biener <rguenther@suse.de>
10851 * tree-ssa-structalias.c (get_function_part_constraint):
10852 Handle NULL fi->decl.
10853 (find_func_aliases_for_call): Properly handle indirect
10854 fi from direct call.
10855 (find_func_clobbers): Likewise.
10856 (ipa_pta_execute): Likewise.
10857 (create_variable_info_for): For functions that are ifunc_resolver
10858 resolve to a varinfo that contains the result of the resolver call.
10859 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
10862 2018-05-31 Michael Collison <michael.collison@arm.com>
10864 * config/aarch64/aarch64.md:
10865 (*fix_to_zero_extenddfdi2): New pattern.
10866 * gcc.target/aarch64/fix_extend1.c: New testcase.
10868 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
10870 PR middle-end/78809
10871 PR middle-end/83026
10872 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
10873 and BUILT_IN_STRNCMP_EQ.
10874 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
10875 BUILT_IN_STRNCMP_EQ.
10876 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
10877 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10878 (gimple_fold_builtin): Likewise.
10879 * tree-ssa-strlen.c (compute_string_length): New function.
10880 (determine_min_obsize): New function.
10881 (handle_builtin_string_cmp): New function to handle calls to
10882 string compare functions.
10883 (strlen_optimize_stmt): Add handling to builtin string compare
10885 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
10886 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10887 * tree.c (build_common_builtin_nodes): Add new defines of
10888 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
10890 2018-05-31 Jakub Jelinek <jakub@redhat.com>
10893 * bb-reorder.c (pass_partition_blocks::gate): Return false for
10894 functions with naked attribute.
10896 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
10898 * config/i386/sse.md (avx_vec_concat<mode>):
10899 Substitute concat_tg_mode mode attribute with xtg_mode.
10900 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
10901 (concat_tg_mode): Remove mode attribute.
10903 2018-05-31 Martin Sebor <msebor@redhat.com>
10906 * calls.c (alloc_max_size): Correct a logic error/typo.
10907 Treat excessive arguments as infinite. Warn for invalid arguments.
10908 * doc/invoke.texi (-Walloc-size-larger-than): Update.
10910 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
10913 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
10914 and movx for Haswell.
10916 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
10917 Cesar Philippidis <cesar@codesourcery.com>
10919 PR middle-end/85879
10920 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
10921 when emitting error on private/firstprivate reductions.
10922 * omp-low.c (lower_omp_target): Avoid reference-type processing
10923 on pointers for firstprivate clause.
10925 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
10927 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
10930 * config/aarch64/aarch64-simd.md
10931 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
10932 (aarch64_ld1_x3_<mode>): Likewise
10933 (aarch64_st1x2<VALLDIF:mode>): Likewise
10934 (aarch64_st1_x2_<mode>): Likewise
10935 (aarch64_st1x3<VALLDIF:mode>): Likewise
10936 (aarch64_st1_x3_<mode>): Likewise
10937 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
10938 (vld1_s8_x3): Likewise.
10939 (vld1_u16_x3): Likewise.
10940 (vld1_s16_x3): Likewise.
10941 (vld1_u32_x3): Likewise.
10942 (vld1_s32_x3): Likewise.
10943 (vld1_u64_x3): Likewise.
10944 (vld1_s64_x3): Likewise.
10945 (vld1_f16_x3): Likewise.
10946 (vld1_f32_x3): Likewise.
10947 (vld1_f64_x3): Likewise.
10948 (vld1_p8_x3): Likewise.
10949 (vld1_p16_x3): Likewise.
10950 (vld1_p64_x3): Likewise.
10951 (vld1q_u8_x3): Likewise.
10952 (vld1q_s8_x3): Likewise.
10953 (vld1q_u16_x3): Likewise.
10954 (vld1q_s16_x3): Likewise.
10955 (vld1q_u32_x3): Likewise.
10956 (vld1q_s32_x3): Likewise.
10957 (vld1q_u64_x3): Likewise.
10958 (vld1q_s64_x3): Likewise.
10959 (vld1q_f16_x3): Likewise.
10960 (vld1q_f32_x3): Likewise.
10961 (vld1q_f64_x3): Likewise.
10962 (vld1q_p8_x3): Likewise.
10963 (vld1q_p16_x3): Likewise.
10964 (vld1q_p64_x3): Likewise.
10965 (vst1_s64_x2): Likewise.
10966 (vst1_u64_x2): Likewise.
10967 (vst1_f64_x2): Likewise.
10968 (vst1_s8_x2): Likewise.
10969 (vst1_p8_x2): Likewise.
10970 (vst1_s16_x2): Likewise.
10971 (vst1_p16_x2): Likewise.
10972 (vst1_s32_x2): Likewise.
10973 (vst1_u8_x2): Likewise.
10974 (vst1_u16_x2): Likewise.
10975 (vst1_u32_x2): Likewise.
10976 (vst1_f16_x2): Likewise.
10977 (vst1_f32_x2): Likewise.
10978 (vst1_p64_x2): Likewise.
10979 (vst1q_s8_x2): Likewise.
10980 (vst1q_p8_x2): Likewise.
10981 (vst1q_s16_x2): Likewise.
10982 (vst1q_p16_x2): Likewise.
10983 (vst1q_s32_x2): Likewise.
10984 (vst1q_s64_x2): Likewise.
10985 (vst1q_u8_x2): Likewise.
10986 (vst1q_u16_x2): Likewise.
10987 (vst1q_u32_x2): Likewise.
10988 (vst1q_u64_x2): Likewise.
10989 (vst1q_f16_x2): Likewise.
10990 (vst1q_f32_x2): Likewise.
10991 (vst1q_f64_x2): Likewise.
10992 (vst1q_p64_x2): Likewise.
10993 (vst1_s64_x3): Likewise.
10994 (vst1_u64_x3): Likewise.
10995 (vst1_f64_x3): Likewise.
10996 (vst1_s8_x3): Likewise.
10997 (vst1_p8_x3): Likewise.
10998 (vst1_s16_x3): Likewise.
10999 (vst1_p16_x3): Likewise.
11000 (vst1_s32_x3): Likewise.
11001 (vst1_u8_x3): Likewise.
11002 (vst1_u16_x3): Likewise.
11003 (vst1_u32_x3): Likewise.
11004 (vst1_f16_x3): Likewise.
11005 (vst1_f32_x3): Likewise.
11006 (vst1_p64_x3): Likewise.
11007 (vst1q_s8_x3): Likewise.
11008 (vst1q_p8_x3): Likewise.
11009 (vst1q_s16_x3): Likewise.
11010 (vst1q_p16_x3): Likewise.
11011 (vst1q_s32_x3): Likewise.
11012 (vst1q_s64_x3): Likewise.
11013 (vst1q_u8_x3): Likewise.
11014 (vst1q_u16_x3): Likewise.
11015 (vst1q_u32_x3): Likewise.
11016 (vst1q_u64_x3): Likewise.
11017 (vst1q_f16_x3): Likewise.
11018 (vst1q_f32_x3): Likewise.
11019 (vst1q_f64_x3): Likewise.
11020 (vst1q_p64_x3): Likewise.
11022 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11024 * config/msp430/msp430.c (msp430_output_labelref): Prepend
11025 user_label_prefix to name.
11027 * tree-core.h: Update comment about the format of NAME string
11028 passed to handler in attribute_spec.
11030 * config/msp430/msp430.md: Remove erroneous subreg expression from
11031 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
11032 zero_extend{q,h}isi2.
11034 2018-05-30 Borislav Petkov <bp@suse.de>
11036 * doc/extend.texi: Document some architecture specific
11037 constraints and sort entries.
11039 2018-05-30 Martin Sebor <msebor@redhat.com>
11041 PR middle-end/85369
11042 * builtins.c (expand_builtin_stpcpy_1): New function.
11043 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
11044 only if the former succeeds.
11046 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
11048 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
11051 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11053 * doc/invoke.texi (-flinker-output): Document
11055 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11057 * passes.c (ipa_write_summaries): Only modify statements if body
11059 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
11060 incrementally linking.
11061 (ipa_passes): Likewise.
11062 * lto-cgraph.c (lto_output_node): When incrementally linking do not
11063 pass down resolution info.
11064 * common.opt (flag_incremental_link): Update info.
11065 * gcc.c (plugin specs): Turn flinker-output=* to
11066 -plugin-opt=-linker-output-known
11067 * toplev.c (compile_file): Also cut compilation when doing incremental
11069 * flag-types. (enum lto_partition_model): Add
11070 LTO_LINKER_OUTPUT_NOLTOREL.
11071 (invoke.texi): Add -flinker-output docs.
11072 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
11073 link same way as WPA; do not stream in dead initializers.
11075 * dwarf2out.c (dwarf2out_die_ref_for_decl,
11076 darf2out_register_external_decl): Support incremental link.
11078 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11080 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
11082 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
11084 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
11085 it down to simple_object_copy_lto_debug_sections.
11086 (run_gcc): Determine incremental LTO link time and configure
11087 lto1 into non-wpa mode, disable renaming of debug sections.
11089 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
11091 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11092 descriptions of various incorrectly documented functions.
11094 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
11097 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11098 address check not strict.
11100 2018-05-30 Richard Biener <rguenther@suse.de>
11102 PR tree-optimization/85964
11103 * tracer.c (better_p): Drop initialized count check, we only
11104 call the function with initialized counts now.
11105 (find_best_successor): Do find a best edge if one
11106 has uninitialized count.
11107 (find_best_predecessor): Likewise. Do BB frequency check only
11108 if count is initialized.
11110 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
11112 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
11113 (aarch64_ldrstr_offset_compare): New.
11114 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
11115 load/store orderings.
11116 (aarch64_gen_adjusted_ldpstp): Likewise.
11118 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
11120 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11121 Check for subset of GENERAL_REGS and FP_REGS.
11122 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
11125 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
11127 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
11128 and wi::to_poly_offset. Add the current offset and then check
11129 whether the sum fits, rather than using an unchecked addition of
11130 a checked term. Check for a shwi rather than a uhwi.
11131 * expr.c (get_bit_range): Use tree_to_poly_uint64.
11132 (store_constructor): Use poly_int_tree_p.
11133 (expand_expr_real_1): Likewise.
11134 * function.c (assign_temp): Likewise.
11135 * fold-const.c (const_binop): Use poly_int_tree_p and
11136 wi::to_poly_offset.
11137 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
11139 * ipa-icf-gimple.c (func_checker::compare_operand): Use
11140 to_poly_offset for MEM offsets.
11141 * ipa-icf.c (sem_variable::equals): Likewise.
11142 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
11143 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
11144 wi::to_poly_offset for BIT_FIELD_REF offsets.
11145 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
11146 wi::to_poly_offset.
11147 * var-tracking.c (emit_note_insn_var_location): Use
11148 tree_to_poly_uint64.
11150 2018-05-29 Jim Wilson <jimw@sifive.com>
11152 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
11154 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
11157 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
11158 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
11160 (sse4_1_round<mode>2): Use nonimmediate_operand
11161 for operand 1 predicate.
11163 2018-05-29 Martin Sebor <msebor@redhat.com>
11164 Richard Biener <rguenther@suse.de>
11167 * calls.c (get_size_range): Call determine_value_range instead
11168 of get_value_range..
11169 * tree-vrp.h (determine_value_range): Declared new function.
11170 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
11172 2018-05-29 Richard Biener <rguenther@suse.de>
11174 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
11175 sure to use non-pattern stmts for get_earlier_stmt arguments.
11176 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
11177 called on pattern stmts.
11178 (get_later_stmt): Likewise.
11180 2018-05-29 Martin Liska <mliska@suse.cz>
11182 PR gcov-profile/85759
11183 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
11186 2018-05-29 Jakub Jelinek <jakub@redhat.com>
11188 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
11190 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
11194 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11195 VEC_PACK_FLOAT_EXPR): New tree codes.
11196 * tree-pretty-print.c (op_code_prio): Handle
11197 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11198 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11199 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11200 * tree-inline.c (estimate_operator_cost): Likewise.
11201 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
11202 * fold-const.c (const_binop): Likewise.
11203 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
11204 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11205 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11206 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
11207 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11208 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11209 * expr.c (expand_expr_real_2): Likewise.
11210 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
11211 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
11212 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
11214 * optabs.c (expand_widen_pattern_expr): For
11215 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
11216 sign from result type rather than operand's type.
11217 (expand_binop_directly): For vec_packu_float_optab and
11218 vec_packs_float_optab allow result type to be different from operand's
11220 * optabs-tree.c (optab_for_tree_code): Handle
11221 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11222 VEC_PACK_FLOAT_EXPR. Formatting fixes.
11223 * tree-vect-generic.c (expand_vector_operations_1): Handle
11224 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11225 VEC_PACK_FLOAT_EXPR.
11226 * tree-vect-stmts.c (supportable_widening_operation): Handle
11228 (supportable_narrowing_operation): Handle FLOAT_EXPR.
11229 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
11230 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
11231 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
11232 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
11234 (vec_pack<floatprefix>_float_<mode>): New expander.
11235 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
11237 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
11238 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
11239 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
11240 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
11241 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
11243 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
11244 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
11245 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11246 VEC_PACK_FLOAT_EXPR): Document.
11248 2018-05-29 Richard Biener <rguenther@suse.de>
11250 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
11252 (stmt_vec_info_vec): Make pointer.
11253 (init_stmt_vec_info_vec): Remove.
11254 (free_stmt_vec_info_vec): Likewise.
11255 (set_stmt_vec_info_vec): New function.
11256 (free_stmt_vec_infos): Likewise.
11257 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
11258 (set_vinfo_for_stmt): Likewise.
11259 (get_earlier_stmt): Likewise.
11260 (get_later_stmt): Likewise.
11261 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
11262 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
11263 (vec_info::~vec_info): Free stmt_vec_infos.
11264 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
11265 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
11266 (pass_slp_vectorize::execute): Likewise.
11267 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
11268 (free_stmt_vec_info_vec): Likewise.
11269 (set_stmt_vec_info_vec): New function.
11270 (free_stmt_vec_infos): Likewise.
11271 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
11272 the global stmt_vec_info_vec.
11273 * tree-parloops.c (gather_scalar_reductions): Use
11274 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
11277 2018-05-29 Richard Biener <rguenther@suse.de>
11279 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
11281 2018-05-29 Martin Liska <mliska@suse.cz>
11282 David Malcolm <dmalcolm@redhat.com>
11284 * vec.c (test_reverse): New.
11285 (vec_c_tests): Add new test.
11286 * vec.h (vl_ptr>::reverse): New function.
11288 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
11290 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
11292 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
11295 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
11297 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
11299 2018-05-28 Richard Biener <rguenther@suse.de>
11301 PR tree-optimization/85933
11302 * tree-vect-data-refs.c (vect_record_base_alignments): Only
11303 look at stmts marked as vectorizable.
11305 2018-05-28 Richard Biener <rguenther@suse.de>
11307 PR tree-optimization/85934
11308 * tree-vect-generic.c (expand_vector_operations_1): Hoist
11309 vector boolean check before scalar optimization.
11311 2018-05-28 Jakub Jelinek <jakub@redhat.com>
11313 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
11316 2018-05-28 Mark Wielaard <mark@klomp.org>
11318 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
11319 if it is an expression containing a minus sign.
11321 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
11323 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
11325 2018-05-27 Paul Koning <ni1d@arrl.net>
11327 * config/pdp11/pdp11.md (truncsihi2): Remove.
11329 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
11330 Chung-Ju Wu <jasonwucj@gmail.com>
11332 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
11334 (unaligned_store_dw): Ditto.
11335 * config/nds32/nds32-memory-manipulation.c
11336 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
11337 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
11338 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
11339 (emit_setmem_word_loop): Rename to ...
11340 (emit_setmem_doubleword_loop): ... this.
11341 (nds32_gen_dup_4_byte_to_word_value): New function.
11342 (nds32_gen_dup_8_byte_to_double_word_value): New function.
11343 (nds32_expand_setmem_loop): Refine implementation.
11344 (nds32_expand_setmem_loop_v3m): Ditto.
11345 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
11348 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11350 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
11352 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
11354 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
11355 (nds32_init_machine_status): Initialize machine->attr_naked_p and
11356 machine->attr_no_prologue_p.
11357 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
11358 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
11359 (nds32_expand_epilogue): Consider attr_naked_p.
11360 (nds32_expand_epilogue_v3pop): Likewise.
11361 (nds32_can_use_return_insn): Likewise.
11362 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
11363 attr_no_prologue_p fields.
11364 * config/nds32/nds32.opt (mret-in-naked-func): New option.
11366 2018-05-27 Jakub Jelinek <jakub@redhat.com>
11369 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
11371 * config/i386/sse.md
11372 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
11374 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
11376 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
11378 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
11380 (*<floatsuffix>floatv2div2sf2): Rename to ...
11381 (*float<floatunssuffix>v2div2sf2): ... this.
11382 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
11383 (float<floatunssuffix>v2div2sf2_mask): ... this.
11384 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
11385 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
11386 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
11388 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
11390 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11392 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11394 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11396 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11398 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
11399 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
11400 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
11401 gen_ufix_truncv8dfv8si2.
11402 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
11403 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
11404 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
11405 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
11406 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
11407 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
11408 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
11409 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
11411 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11415 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
11417 2018-05-25 Jim Wilson <jimw@sifive.com>
11419 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
11420 * config/riscv/riscv.c (struct machine_function): Add
11421 interrupt_handler_p and attribute_checked_p fields.
11422 (riscv_attribute_table): Add interrupt.
11423 (riscv_interrupt_type_p): New.
11424 (riscv_save_reg_p): Save extra regs for interrupt handler.
11425 (riscv_use_save_libcall): Return false for interrupt handler.
11426 (riscv_first_stack_step): Add forward declaration.
11427 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
11428 for interrupt handler with large frame. Use it for saved reg list.
11429 (riscv_expand_prologue): Move flag_stack_usage_info support to
11430 eliminate duplication.
11431 (riscv_expand_epilogue): Generate mret for interrupt handler.
11432 (riscv_epilogue_uses): New.
11433 (riscv_can_use_return_insn): Return false for interrupt handler.
11434 (riscv_function_ok_for_sibcall): Likewise.
11435 (riscv_set_current_function): Add interrupt handler support.
11436 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
11437 * config/riscv/riscv.md (UNSPECV_MRET): New.
11439 (riscv_frflags, riscv_fsflags): Use tab after opcode.
11441 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
11443 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
11445 PR tree-optimization/85712
11446 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
11447 this candidate has already been replaced in-situ by a copy.
11449 2018-05-25 Jason Merrill <jason@redhat.com>
11451 PR c++/80485 - inline function non-zero address.
11452 * symtab.c (nonzero_address): Check DECL_COMDAT.
11454 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
11457 * config/alpha/alpha.md (ashlsi3): New insn pattern.
11458 (*ashlsi_se): Rename from *ashldi_se. Define as sign
11459 extension of SImode operation. Use const123_operand predicate.
11460 (*saddsi_1): Remove.
11461 (*saddl_se_1): Ditto.
11462 (*ssubsi_1): Ditto.
11463 (*ssubl_se_1): Ditto.
11464 * config/alpha/predicates.md (const123_operand): New predicate.
11465 * config/alpha/constraints.md (P): Use IN_RANGE.
11467 2018-05-25 Richard Biener <rguenther@suse.de>
11469 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
11471 (ref_maybe_used_by_stmt_p): Likewise.
11472 (stmt_may_clobber_ref_p): Likewise.
11473 (stmt_may_clobber_ref_p_1): Likewise.
11474 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
11476 (ref_maybe_used_by_stmt_p): Likewise.
11477 (stmt_may_clobber_ref_p): Likewise.
11478 (stmt_may_clobber_ref_p_1): Likewise.
11479 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
11480 the alias oracle to disambiguate DRs with stmts DR analysis
11482 (vect_analyze_data_refs): Do not give up on not analyzable
11483 DRs for BB vectorization. Remove code truncating the dataref
11486 2018-05-25 Jakub Jelinek <jakub@redhat.com>
11489 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11490 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
11491 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
11493 2018-05-25 Richard Biener <rguenther@suse.de>
11495 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
11496 function, combining stmt data ref gathering and fatal analysis
11498 (vect_analyze_data_refs): Remove now redudnant code and simplify.
11499 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
11500 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
11501 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
11502 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
11504 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11506 PR tree-optimization/85720
11507 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
11508 SCC if all partitions are builtins.
11509 (version_loop_by_alias_check): New parameter. Generate cancelable
11510 runtime alias check if all partitions are builtins.
11511 (distribute_loop): Update call to above function.
11513 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11515 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
11516 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
11517 (parm_default_def_partition_arg): Ditto.
11518 (set_parm_default_def_partition): Ditto.
11519 (get_parm_default_def_partitions): Ditto and make it static.
11520 (get_undefined_value_partitions): Ditto and make it static.
11521 (remove_ssa_form): Refactor call to init_var_map here.
11522 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
11523 computation for loop region.
11524 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
11525 (register_default_def): Delete.
11526 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
11527 (parm_default_def_partition_arg): Ditto.
11528 (set_parm_default_def_partition): Ditto.
11529 (get_parm_default_def_partitions): Ditto and make it static.
11530 (get_undefined_value_partitions): Ditto and make it static.
11531 (coalesce_with_default, coalesce_with_default): Update comment.
11532 (create_coalesce_list_for_region): New func factored out from
11533 create_outofssa_var_map.
11534 (populate_coalesce_list_for_outofssa): New func factored out from
11535 create_outofssa_var_map and coalesce_ssa_name.
11536 (create_outofssa_var_map): Delete.
11537 (coalesce_ssa_name): Refactor to support live range computation.
11538 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
11539 (get_parm_default_def_partitions): Delete.
11540 (get_undefined_value_partitions): Ditto.
11541 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
11542 computation for loop region.
11543 (new_tree_live_info, loe_visit_block): Ditto.
11544 (live_worklist, set_var_live_on_entry): Ditto.
11545 (calculate_live_on_exit, verify_live_on_entry): Ditto.
11546 * tree-ssa-live.h (struct _var_map): New fields.
11547 (init_var_map): Change decl.
11548 (region_contains_p): New.
11550 2018-05-25 Bin Cheng <bin.cheng@arm.com>
11552 * tree-ssa-live.h (live_merge_and_clear): Delete.
11554 2018-05-25 Richard Biener <rguenther@suse.de>
11557 * tree-dump.c (dequeue_and_dump): Remove access to removed
11558 operand 2 of a SWITCH_EXPR.
11560 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11562 * doc/sourcebuild.texi (vect_double_cond_arith): Include
11563 multiplication and division.
11564 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
11565 (cond_udiv@var{m}, cond_umod@var{m}): Document.
11566 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
11567 (cond_udiv_optab, cond_umod_optab): New optabs.
11568 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
11569 (IFN_COND_RDIV): New internal functions.
11570 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
11571 TRUNC_MOD_EXPR and RDIV_EXPR.
11572 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
11573 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
11575 (SVE_INT_BINARY): Include mult.
11576 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
11577 (optab, sve_int_op): Handle mult.
11578 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
11580 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
11581 for SVE_INT_BINARY_SD.
11583 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11585 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
11586 (optab, sve_int_op): Handle div and udiv.
11587 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
11588 for SVE_INT_BINARY_SD.
11589 (*<optab><mode>3): New insn for the same.
11591 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11593 * tree-vect-patterns.c: Include predict.h.
11594 (vect_recog_divmod_pattern): Restrict check for division support
11595 to when optimizing for size.
11597 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11599 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
11600 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
11601 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
11602 (gimple_match_op::set_op): Likewise.
11603 (gimple_resimplify4): Declare.
11604 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
11605 (expr::gen_transform): Likewise.
11606 (decision_tree::gen): Generate a simplification routine for 4 operands.
11607 * gimple-match-head.c (gimple_simplify): Add an overload for
11608 4 operands. In the top-level function, handle up to 4 call
11609 arguments and call gimple_resimplify4.
11610 (gimple_resimplify4): New function.
11611 (build_call_internal): Pass a fourth operand.
11612 (maybe_push_to_seq): Likewise.
11613 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
11614 Fold VEC_COND_EXPRs of an operation and a default value into
11615 an IFN_COND_* function if possible.
11616 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
11618 (SVE_COND_FP_BINARY): Include them.
11619 (optab, sve_fp_op): Handle them.
11620 (SVE_INT_BINARY_REV): New code iterator.
11621 (SVE_COND_FP_BINARY_REV): New int iterator.
11622 (commutative): New int attribute.
11623 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
11625 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
11627 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
11628 (*cond_<optab><mode>): New patterns for reversed operands.
11630 2018-05-25 Richard Biener <rguenther@suse.de>
11632 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
11633 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
11634 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
11635 (STMT_VINFO_GROUPED_ACCESS): Adjust.
11636 * tree-vect-data-refs.c (everywhere): Adjust users.
11637 * tree-vect-loop.c (everywhere): Likewise.
11638 * tree-vect-slp.c (everywhere): Likewise.
11639 * tree-vect-stmts.c (everywhere): Likewise.
11640 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11642 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11644 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
11646 (gcc_cv_as_section_exclude): ... this.
11647 Try Solaris as #exclude syntax.
11648 * configure: Regenerate.
11649 * config.in: Regenerate.
11650 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
11652 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
11653 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
11655 * varasm.c (default_elf_asm_named_section): Don't check if
11656 HAVE_GAS_SECTION_EXCLUDE is defined.
11658 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11660 * doc/md.texi: Update the documentation of the cond_* optabs
11661 to mention the new final operand. Fix GET_MODE_NUNITS call.
11662 Describe the scalar case too.
11663 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
11664 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
11666 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
11667 (get_conditional_internal_fn): Update comment.
11668 * tree-vect-loop.c (vectorizable_reduction): Pass the original
11669 accumulator value as a final argument to conditional functions.
11670 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
11671 a define_expand and add an "else" operand. Assert for now that
11672 the else operand is equal to operand 2. Use SVE_INT_BINARY and
11673 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
11674 (*cond_<optab><mode>): New patterns.
11675 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
11676 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11677 (UNSPEC_COND_EOR): Delete.
11678 (optab): Remove associated mappings.
11679 (SVE_INT_BINARY): New code iterator.
11680 (sve_int_op): Remove int attribute and add "minus" to the code
11682 (SVE_COND_INT_OP): Delete.
11683 (SVE_COND_FP_OP): Rename to...
11684 (SVE_COND_FP_BINARY): ...this.
11686 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
11688 * optabs.c (can_reuse_operands_p): New function.
11689 (maybe_legitimize_operands): Try to reuse the results for
11692 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
11694 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
11695 Add {q} suffix to insn mnemonic.
11697 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11699 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
11700 (msp430_warn_func_return): New.
11702 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
11704 * fold-const.c (tree_nonzero_bits): New function.
11705 * fold-const.h (tree_nonzero_bits): Likewise.
11706 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
11707 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
11709 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
11713 * varasm.c (assemble_alias): Check ifunc_resolver only on
11716 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
11719 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
11720 when memory input operand is handled.
11722 2018-05-24 Luis Machado <luis.machado@linaro.org>
11724 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
11726 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
11728 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
11730 * match.pd: Delay FMA folds until after vectorization.
11732 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
11735 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11736 address check not strict.
11738 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
11740 * gimple-match.h (gimple_match_op): New class.
11741 (mprts_hook): Replace parameters with a gimple_match_op *.
11742 (maybe_build_generic_op): Likewise.
11743 (gimple_simplified_result_is_gimple_val): Replace parameters with
11744 a const gimple_match_op *.
11745 (gimple_simplify): Replace code_helper * and tree * parameters with
11746 a gimple_match_op * parameter.
11747 (gimple_resimplify1): Replace code_helper *, tree and tree *
11748 parameters with a gimple_match_op * parameter.
11749 (gimple_resimplify2): Likewise.
11750 (gimple_resimplify3): Likewise.
11751 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
11752 parameters with a gimple_match_op * parameter.
11753 * gimple-match-head.c (gimple_simplify): Change prototypes of
11754 auto-generated functions to take a gimple_match_op * instead of
11755 separate code_helper * and tree * parameters. Make the same
11756 change in the top-level overload and update calls to the
11757 gimple_resimplify routines. Update calls to the auto-generated
11758 functions and to maybe_push_res_to_seq in the publicly-facing
11759 operation-specific gimple_simplify overloads.
11760 (gimple_match_op::MAX_NUM_OPS): Define.
11761 (gimple_resimplify1): Replace rcode and ops with a single res_op
11762 parameter. Update call to gimple_simplify.
11763 (gimple_resimplify2): Likewise.
11764 (gimple_resimplify3): Likewise.
11765 (mprts_hook): Replace parameters with a gimple_match_op *.
11766 (maybe_build_generic_op): Likewise.
11767 (build_call_internal): Replace type, nargs and ops with
11768 a gimple_match_op *.
11769 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
11770 with a single gimple_match_op *. Update calls to mprts_hook,
11771 build_call_internal and gimple_simplified_result_is_gimple_val.
11772 Factor out code that is common to the tree_code and combined_fn cases.
11773 * genmatch.c (expr::gen_transform): Replace tem_code and
11774 tem_ops with a gimple_match_op called tem_op. Update calls
11775 to the gimple_resimplify functions and maybe_push_res_to_seq.
11776 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
11777 res_ops. Update call to the gimple_resimplify functions.
11778 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
11779 (decision_tree::gen): Make the functions take a gimple_match_op *
11780 called res_op instead of separate res_code and res_ops parameters.
11781 Update call accordingly.
11782 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
11783 and ops with a single res_op parameter. Update calls to
11784 maybe_build_generic_op and maybe_push_res_to_seq.
11785 (fold_stmt_1): Update calls to gimple_simplify and
11786 replace_stmt_with_simplification.
11787 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
11788 and gimple_simplified_result_is_gimple_val.
11789 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
11791 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
11792 with a gimple_match_op *.
11793 (vn_nary_build_or_lookup): Likewise. Update call to
11794 vn_nary_build_or_lookup_1.
11795 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
11796 gimple_match_op *. Update calls to the gimple_resimplify routines
11797 and to gimple_simplified_result_is_gimple_val.
11798 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
11799 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
11800 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
11801 (visit_nary_op): Likewise.
11802 (visit_reference_op_load): Likewise.
11804 2018-05-23 Luis Machado <luis.machado@linaro.org>
11806 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
11807 modifier for printing the step amount.
11809 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
11812 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
11815 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
11817 * doc/sourcebuild.texi (Endianness): New subsubsection.
11819 2018-05-23 Luis Machado <luis.machado@linaro.org>
11821 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11822 <prefetch_dynamic_strides>: New const bool field.
11823 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11824 prefetch_dynamic_strides.
11825 (exynosm1_prefetch_tune): Likewise.
11826 (thunderxt88_prefetch_tune): Likewise.
11827 (thunderx_prefetch_tune): Likewise.
11828 (thunderx2t99_prefetch_tune): Likewise.
11829 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
11831 (aarch64_override_options_internal): Update to set
11832 PARAM_PREFETCH_DYNAMIC_STRIDES.
11833 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11834 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11835 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11836 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11837 prefetch-dynamic-strides setting.
11839 2018-05-23 Luis Machado <luis.machado@linaro.org>
11841 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11842 <minimum_stride>: New const int field.
11843 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11844 minimum_stride field defaulting to -1.
11845 (exynosm1_prefetch_tune): Likewise.
11846 (thunderxt88_prefetch_tune): Likewise.
11847 (thunderx_prefetch_tune): Likewise.
11848 (thunderx2t99_prefetch_tune): Likewise.
11849 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
11850 <default_opt_level>: Set to 3.
11851 (aarch64_override_options_internal): Update to set
11852 PARAM_PREFETCH_MINIMUM_STRIDE.
11853 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11854 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11855 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11856 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11857 stride is constant and is below the minimum stride threshold.
11859 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11861 * config/arm/arm-cpus.in (mode26): Delete.
11862 (armv4): Delete mode26 reference.
11863 * config/arm/arm.c (arm_configure_build_target): Delete use of
11866 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11868 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
11870 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
11871 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
11873 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
11874 (floatunsdidf2): Ditto.
11876 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11878 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
11879 (fixuns_trunc<mode>si2_avx512f): Ditto.
11880 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
11881 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
11882 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
11884 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11886 PR rtl-optimization/79985
11887 * df-scan.c (df_insn_refs_collect): Remove special case for
11888 global registers and asm statements.
11890 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
11892 * extend.texi (Global Register Variables): Rewrite the bullet list.
11893 Note that the register is available for allocation. Note that access
11894 via inline asm must use constraints. Add note about async-signal
11895 handlers. Remove paragraph about automagic register selection.
11897 2018-05-23 Richard Biener <rguenther@suse.de>
11899 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
11900 of fixed offset from memset VN.
11902 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
11904 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
11905 first_interp field.
11906 (alloc_cand_and_find_basis): Initialize first_interp field.
11907 (slsr_process_mul): Modify first_interp field.
11908 (slsr_process_add): Likewise.
11909 (slsr_process_cast): Modify first_interp field for each new
11911 (slsr_process_copy): Likewise.
11912 (dump_candidate): Dump first_interp field.
11913 (replace_mult_candidate): Process all interpretations, not just
11915 (replace_rhs_if_not_dup): Likewise.
11916 (replace_one_candidate): Likewise.
11918 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
11920 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
11922 (aarch64_needs_frame_chain): New function.
11923 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
11925 2018-05-23 Sudakshina Das <sudi.das@arm.com>
11928 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
11929 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
11930 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
11931 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
11932 as true for strict-align.
11933 (aarch64_can_inline_p): Perform checks even when callee has no
11934 attributes to check for strict alignment.
11935 * doc/extend.texi (AArch64 Function Attributes): Document
11937 * doc/invoke.texi: (AArch64 Options): Likewise.
11939 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
11941 PR tree-optimization/85853
11942 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
11943 the handling of the root of the node to...
11944 (vect_slp_analyze_node_operations_1): ...this new function,
11945 and run the whole thing with the child nodes' def types
11946 set according to their SLP node's def type.
11948 2018-05-23 Richard Biener <rguenther@suse.de>
11950 PR middle-end/85874
11951 * tree-data-ref.c (create_runtime_alias_checks): Defer
11952 and ignore overflow warnings.
11954 2018-05-23 Yury Gribov <tetra2005@gmail.com>
11956 PR tree-optimization/85822
11957 * tree-vrp.c (is_masked_range_test): Fix handling of negative
11960 2018-05-23 Richard Biener <rguenther@suse.de>
11962 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
11963 memset constants via native_interpret_expr.
11965 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
11968 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
11970 (cgraph_node::create_alias): Likewise.
11971 (cgraph_node::get_availability): Check ifunc_resolver instead
11972 of looking up ifunc attribute.
11973 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
11974 * varasm.c (do_assemble_alias): Likewise.
11975 (assemble_alias): Likewise.
11976 (default_binds_local_p_3): Likewise.
11977 * cgraph.h (cgraph_node): Add ifunc_resolver.
11978 (cgraph_node::only_called_directly_or_aliased_p): Return false
11979 for IFUNC resolver.
11980 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
11982 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
11983 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
11984 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
11985 instead of looking up ifunc attribute.
11987 2018-05-22 Luis Machado <luis.machado@linaro.org>
11989 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
11991 2018-05-22 Martin Sebor <msebor@redhat.com>
11993 PR middle-end/85359
11994 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
11995 only when expasion succeeds.
11996 (expand_builtin_strcmp): Same.
11997 (expand_builtin_strncmp): Same.
11999 2018-05-22 Martin Sebor <msebor@redhat.com>
12001 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
12003 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12004 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12006 * config/aarch64/aarch64-ldpstp.md: Replace uses of
12007 aarch64_mem_pair_operand with memory_operand and delete operand swapping
12009 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12010 Add check for legitimate_address.
12011 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
12012 (aarch64_swap_ldrstr_operands): New.
12013 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
12016 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
12017 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12019 * config/aarch64/aarch64.md: New patterns to generate stp
12021 (store_pair_sw, store_pair_dw): New patterns to generate stp for
12022 single words and double words.
12023 (load_pair_sw, load_pair_dw): Likewise.
12024 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
12026 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
12028 * config/aarch64/aarch64-ldpstp.md: Modify peephole
12029 for different mode ldpstp and add peephole for merged zero stores.
12030 Likewise for loads.
12031 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12033 (aarch64_gen_store_pair): Rename calls to match new patterns.
12034 (aarch64_gen_load_pair): Rename calls to match new patterns.
12035 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
12036 (load_pair<DREG:mode><DREG2:mode>): ... This.
12037 (store_pair<mode>): Rename to...
12038 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
12039 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
12040 New mode iterators.
12041 (V_INT_EQUIV): Handle SImode.
12042 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
12045 2018-05-22 Martin Sebor <msebor@redhat.com>
12048 * calls.c (maybe_warn_nonstring_arg): Use string length to set
12049 or ajust the presumed bound on an operation to avoid unnecessary
12052 2018-05-22 Martin Sebor <msebor@redhat.com>
12054 PR tree-optimization/85826
12055 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
12056 assuming that a DECL necesarily has a constant size.
12058 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
12060 PR middle-end/85862
12061 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
12063 2018-05-22 Richard Biener <rguenther@suse.de>
12065 PR tree-optimization/85834
12066 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
12067 non-constant and non-zero memset arguments.
12069 2018-05-22 Martin Liska <mliska@suse.cz>
12072 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
12074 2018-05-22 Richard Biener <rguenther@suse.de>
12076 PR tree-optimization/85863
12077 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
12078 comparisons when vectype is specified.
12079 (vectorizable_condition): Do not specify vectype for
12080 vect_is_simple_cond when SLP vectorizing.
12082 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
12085 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
12086 define __ibm128 as long double.
12087 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
12088 as a distinct type when IEEE 128-bit support is enabled.
12089 (init_float128_ieee): Fix up conversions between IFmode and IEEE
12090 128-bit types to use the correct functions.
12091 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
12092 convert between 128-bit floating point types that have different
12093 modes but the same representation, instead of using gen_lowpart to
12095 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
12097 (IFKF_reg): New attributes to give the register constraints for
12099 (extend<mode>tf2_internal): New insns to mark an explicit
12100 conversion between 128-bit floating point types that have a
12101 different mode but share the same representation.
12103 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
12105 PR tree-optimization/85814
12106 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
12107 a null return from get_strinfo when unsharing the next
12108 strinfo in the chain.
12110 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12113 * varasm.c (weak_finish): Clean up weak_decls.
12115 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12117 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
12118 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
12119 UNSPEC_UADALP values.
12120 * config/aarch64/iterators.md (ABAL): New int iterator.
12123 (sur): Add mappings for the above.
12124 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
12126 (aarch64_<sur>abal<mode>_4): Likewise.
12127 (aarch64_<sur>adalp<mode>_3): Likewise.
12128 (<sur>sadv16qi): New define_expand.
12130 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
12132 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
12133 (*movdf_internal): Ditto.
12134 (*rcpsf2_sse): Ditto.
12135 (*rsqrtsf2_sse): Ditto.
12136 (*sqrt<mode>2_sse): Ditto.
12138 2018-05-21 Tamar Christina <tamar.christina@arm.com>
12140 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
12142 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
12143 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
12144 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
12145 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
12147 * config/aarch64/arm_neon.h: Likewise.
12148 * config/aarch64/iterators.md (VQ_I): New.
12150 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
12152 * config.gcc: Add arc/t-multilib-linux to tmake_file for
12154 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
12157 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12159 * config/nds32/constraints.md (S): New constraint.
12160 * config/nds32/nds32.md (call_internal): Use constraint S.
12161 (call_value_internal): Likewise.
12162 (sibcall_internal): Likewise.
12163 (sibcall_value_internal): Likewise.
12165 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12166 Chung-Ju Wu <jasonwucj@gmail.com>
12168 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
12169 into consideration.
12171 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
12172 Chung-Ju Wu <jasonwucj@gmail.com>
12174 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
12175 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
12176 (nds32_rtx_costs_impl): Simplify.
12177 (nds32_address_cost_impl): Simplify.
12178 (nds32_init_rtx_costs): New function.
12179 (nds32_rtx_costs_speed_prefer): Likewise.
12180 (nds32_rtx_costs_size_prefer): Likewise.
12181 (nds32_address_cost_speed_prefer): Likewise.
12182 (nds32_address_cost_speed_fwprop): Likewise.
12183 (nds32_address_cost_size_prefer): Likewise.
12184 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
12185 * config/nds32/nds32.c (nds32_option_override): Use
12186 nds32_init_rtx_costs function.
12188 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
12190 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
12191 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
12192 (TARGET_PIPELINE_N8): Likewise.
12193 (TARGET_PIPELINE_N10): Likewise.
12194 (TARGET_PIPELINE_N13): Likewise.
12195 (TARGET_PIPELINE_GRAYWOLF): Likewise.
12197 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
12199 * config/nds32/nds32-fpu.md: Update copyright year.
12201 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12203 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
12205 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12207 * config/nds32/nds32.c
12208 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
12209 * config/nds32/nds32.opt (minline-asm-r15): New option.
12211 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12213 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
12215 * config/nds32/nds32.md (abssi2): New pattern.
12217 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
12219 * config/i386/i386.md (rex64namesuffix): New mode attribute.
12220 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
12221 Merge insn pattern from sse_cvtsi2ss<round_name> and
12222 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
12223 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
12224 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
12225 using SWI48 mode iterator.
12226 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
12227 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
12228 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
12229 pattern from sse_cvttss2si<round_saeonly_name>
12230 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
12231 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
12232 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
12233 using SWI48 mode iterator.
12234 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
12235 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
12236 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
12237 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
12238 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
12239 using SWI48 mode iterator.
12240 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
12241 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
12242 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
12243 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
12244 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
12245 SWI48 mode iterator.
12246 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
12247 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
12248 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
12249 pattern from sse_cvttsd2si<round_saeonly_name>
12250 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
12252 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12254 * config/nds32/nds32-md-auxiliary.c
12255 (nds32_valid_smw_lwm_base_p): Refine.
12256 (nds32_output_smw_single_word): Refine.
12257 (nds32_output_smw_double_word): New.
12258 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
12260 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12262 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
12263 (nds32_output_stack_pop): Refine.
12264 (nds32_expand_unaligned_load): Refine.
12265 (nds32_expand_unaligned_store): Refine.
12267 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12268 Chung-Ju Wu <jasonwucj@gmail.com>
12270 * config/nds32/constants.md: Add TP_REGNUM constant.
12271 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
12272 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
12274 * config/nds32/nds32-doubleword.md: Consider flag_pic.
12275 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
12276 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
12277 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
12278 and PIC code generation.
12279 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
12281 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
12283 * config/nds32/nds32.md: Support TLS and PIC.
12284 * config/nds32/nds32.c: Support TLS and PIC.
12285 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
12286 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
12289 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
12291 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
12292 mode with E_ prefix.
12294 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12295 Chung-Ju Wu <jasonwucj@gmail.com>
12297 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
12298 * config/nds32/nds32-md-auxiliary.c
12299 (symbolic_reference_mentioned_p): New.
12300 (nds32_legitimize_ict_address): New.
12301 (nds32_expand_ict_move): New.
12302 (nds32_indirect_call_referenced_p): New.
12303 (nds32_symbol_binds_local_p): Delete.
12304 (nds32_long_call_p): Modify.
12305 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
12306 * config/nds32/nds32-protos.h
12307 (symbolic_reference_mentioned_p): Declare.
12308 (nds32_legitimize_ict_address): Declare.
12309 (nds32_expand_ict_move): Declare.
12310 (nds32_indirect_call_referenced_p): Declare.
12311 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
12312 (nds32_relax_group): Use nds32_ict_const_p as condition.
12313 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
12314 (nds32_asm_file_start): Output ict_model directive in asm code.
12315 (nds32_legitimate_address_p): Consider indirect call.
12316 (nds32_print_operand): Consider indirect call.
12317 (nds32_print_operand_address): Consider indirect call.
12318 (nds32_insert_attributes): Handle "indirect_call" attribute.
12319 (TARGET_LEGITIMATE_ADDRESS_P): Define.
12320 (TARGET_LEGITIMATE_CONSTANT_P): Define.
12321 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
12322 (TARGET_DELEGITIMIZE_ADDRESS): Define.
12323 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
12324 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
12325 (TARGET_ICT_MODEL_SMALL): Define.
12326 (TARGET_ICT_MODEL_LARGE): Define.
12327 * config/nds32/nds32.md (movsi): Consider ict model.
12328 (call, call_value): Consider ict model.
12329 (sibcall, sibcall_value): Consider ict model.
12330 * config/nds32/nds32.opt (mict-model): New option.
12331 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
12334 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
12335 Monk Chiang <sh.chiang04@gmail.com>
12336 Jim Wilson <jimw@sifive.com>
12338 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
12339 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
12340 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
12341 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
12342 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
12343 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
12344 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
12345 compute save_libcall_adjustment properly.
12346 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
12347 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
12348 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
12349 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
12350 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
12351 (ABI_SPEC): Handle mabi=ilp32e.
12352 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
12353 (RVE): Add RVE mask.
12354 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
12355 <-march>: Add rv32e as an example.
12357 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
12360 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
12361 (intra_create_variable_infos): Handle C++ constructors.
12363 2018-05-18 Martin Liska <mliska@suse.cz>
12365 * passes.def: Remove a redundant pass.
12367 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
12370 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
12372 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12374 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
12376 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
12378 (armv2, armv2a, armv3, armv3m): Delete architectures.
12379 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
12380 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
12381 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
12383 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
12384 (*mulsidi3adddi): Likewise.
12385 (mulsidi3): Likewise.
12386 (*mulsidi3_nov6): Likewise.
12387 (umulsidi3): Likewise.
12388 (umulsidi3_nov6): Likewise.
12389 (umaddsidi4): Likewise.
12390 (*umulsidi3adddi): Likewise.
12391 (smulsi3_highpart): Likewise.
12392 (*smulsi3_highpart_nov6): Likewise.
12393 (umulsi3_highpart): Likewise.
12394 (*umulsi3_highpart_nov6): Likewise.
12395 * config/arm/arm.h (arm_arch3m): Delete.
12396 * config/arm/arm.c (arm_arch3m): Delete.
12397 (arm_option_override_internal): Update armv3-related comment.
12398 (arm_configure_build_target): Delete use of isa_bit_mode32.
12399 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
12400 (arm_rtx_costs_internal): Delete check of arm_arch3m.
12401 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
12402 (mulsa3): Likewise.
12403 (mulusa3): Likewise.
12404 * config/arm/arm-protos.h (arm_arch3m): Delete.
12405 * config/arm/arm-tables.opt: Regenerate.
12406 * config/arm/arm-tune.md: Likewise.
12407 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
12408 deleted architectures.
12410 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12412 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
12413 (armv5t, armv5te): New features.
12414 (ARMv5, ARMv5e): Delete fgroups.
12415 (ARMv5t, ARMv5te): Adjust for above changes.
12416 (ARMv6m): Likewise.
12417 (armv5, armv5e): Delete arches.
12418 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
12420 (*call_reg_arm): Likewise.
12421 (*call_value_reg_armv5): Likewise.
12422 (*call_value_reg_arm): Likewise.
12423 (*call_symbol): Likewise.
12424 (*call_value_symbol): Likewise.
12425 (*sibcall_insn): Likewise.
12426 (*sibcall_value_insn): Likewise.
12427 (clzsi2): Likewise.
12428 (prefetch): Likewise.
12429 (define_split and define_peephole2 dependent on arm_arch5):
12431 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
12433 (TARGET_ARM_QBIT): Likewise.
12434 (TARGET_DSP_MULTIPLY): Likewise.
12435 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
12436 (arm_arch5, arm_arch5e): Delete.
12437 (arm_arch5t, arm_arch5te): Declare.
12438 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
12439 (arm_arch5t): Declare.
12440 (arm_option_reconfigure_globals): Update for the above.
12441 (arm_options_perform_arch_sanity_checks): Update comment, replace
12442 use of arm_arch5 with arm_arch5t.
12443 (use_return_insn): Likewise.
12444 (arm_emit_call_insn): Likewise.
12445 (output_return_instruction): Likewise.
12446 (arm_final_prescan_insn): Likewise.
12447 (arm_coproc_builtin_available): Likewise.
12448 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
12449 arm_arch5e with arm_arch5t and arm_arch5te.
12450 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
12451 (arm_arch5t, arm_arch5te): Declare.
12452 * config/arm/arm-tables.opt: Regenerate.
12453 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
12454 * config/arm/t-multilib: Likewise.
12455 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
12456 instead of arm_arch5.
12457 (*call_reg_thumb1): Likewise.
12458 (*call_value_reg_thumb1_v5): Likewise.
12459 (*call_value_reg_thumb1): Likewise.
12460 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
12462 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
12464 2018-05-18 Martin Liska <mliska@suse.cz>
12466 PR gcov-profile/84846
12467 * doc/gcov.texi: Document -t option of gcov tool.
12469 2018-05-18 Martin Liska <mliska@suse.cz>
12471 PR gcov-profile/84846
12472 * gcov.c (print_usage): Add new -t option.
12473 (process_args): Handle the option.
12474 (generate_results): Use stdout as output when requested by
12477 2018-05-18 Martin Liska <mliska@suse.cz>
12479 PR gcov-profile/84846
12480 * coverage.c (coverage_init): Write PWD to .gcno file.
12481 * doc/gcov.texi: Document how working directory is printed.
12482 * gcov-dump.c (dump_gcov_file): Print PWD.
12483 * gcov.c (output_intermediate_file): Likewise.
12484 (read_graph_file): Read PWD string.
12485 (output_lines): Print PWD.
12487 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12489 PR middle-end/85817
12490 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
12491 for retval and return false if all args to phi are zero.
12493 2018-05-18 Richard Biener <rguenther@suse.de>
12495 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
12497 (evrp_dom_walker::before_dom_children): Call it.
12499 2018-05-18 Richard Biener <rguenther@suse.de>
12501 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
12502 results when processing array refs with variable index.
12504 2018-05-18 Toon Moene <toon@moene.org>
12506 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
12507 directly after that of -floop-interchange. Indicate that both
12508 options are enabled by default when specifying -O3.
12510 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12512 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
12513 iterator. Delete separate integer-mode vec_set<mode> expander.
12514 (aarch64_simd_vec_setv2di): Delete.
12515 (vec_setv2di): Delete.
12516 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
12517 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
12518 the "w, r" alternative.
12520 2018-05-18 Martin Liska <mliska@suse.cz>
12522 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
12523 * tree-pass.h (make_pass_lower_switch_O0): New function.
12524 * tree-switch-conversion.c (node_has_low_bound): Remove.
12525 (node_has_high_bound): Likewise.
12526 (node_is_bounded): Likewise.
12527 (class pass_lower_switch): Make it a template type and create
12529 (pass_lower_switch::execute): Add template argument.
12530 (make_pass_lower_switch): New function.
12531 (make_pass_lower_switch_O0): New function.
12532 (do_jump_if_equal): Remove.
12533 (emit_case_nodes): Simplify to just handle all 3 cases and leave
12534 all the hard work to tree optimization passes.
12536 2018-05-18 Martin Liska <mliska@suse.cz>
12538 * dbgcnt.c (limit_low): Renamed from limit.
12539 (limit_high): New variable.
12540 (dbg_cnt_is_enabled): Check for upper limit.
12541 (dbg_cnt): Adjust dumping.
12542 (dbg_cnt_set_limit_by_index): Add new argument for high
12544 (dbg_cnt_set_limit_by_name): Likewise.
12545 (dbg_cnt_process_single_pair): Parse new format.
12546 (dbg_cnt_process_opt): Use strtok.
12547 (dbg_cnt_list_all_counters): Remove 'value' and add
12549 * doc/invoke.texi: Document changes.
12551 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
12553 * doc/sourcebuild.texi (scalar_all_fma): Document.
12554 * tree.def (FMA_EXPR): Delete.
12555 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
12556 * internal-fn.c (ternary_direct): New macro.
12557 (expand_ternary_optab_fn): Likewise.
12558 (direct_ternary_optab_supported_p): Likewise.
12559 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
12560 * builtins.c (fold_builtin_fma): Delete.
12561 (fold_builtin_3): Don't call it.
12562 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
12563 * expr.c (expand_expr_real_2): Likewise.
12564 * fold-const.c (operand_equal_p): Likewise.
12565 (fold_ternary_loc): Likewise.
12566 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
12567 * gimple.c (DEFTREECODE): Likewise.
12568 * gimplify.c (gimplify_expr): Likewise.
12569 * optabs-tree.c (optab_for_tree_code): Likewise.
12570 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
12571 * tree-eh.c (operation_could_trap_p): Likewise.
12572 (stmt_could_throw_1_p): Likewise.
12573 * tree-inline.c (estimate_operator_cost): Likewise.
12574 * tree-pretty-print.c (dump_generic_node): Likewise.
12575 (op_code_prio): Likewise.
12576 * tree-ssa-loop-im.c (stmt_cost): Likewise.
12577 * tree-ssa-operands.c (get_expr_operands): Likewise.
12578 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
12579 * fold-const-call.h (fold_fma): Delete.
12580 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
12581 CFN_FNMA and CFN_FNMS.
12582 (fold_fma): Delete.
12583 * genmatch.c (combined_fn): New enum.
12584 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
12585 (commutative_op): New function.
12586 (commutate): Use it. Handle more than 2 operands.
12587 (dt_operand::gen_gimple_expr): Use commutative_op.
12588 (parser::parse_expr): Allow :c to be used with non-binary
12589 operators if the commutative operand is known.
12590 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
12591 CFN_FMS, CFN_FNMA and CFN_FNMS.
12592 (backprop::process_assign_use): Remove FMA_EXPR handling.
12593 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
12594 (gen_hsa_fma): New function.
12595 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
12596 IFN_FNMA and IFN_FNMS.
12597 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
12598 * gimple-fold.h (follow_all_ssa_edges): Declare.
12599 * gimple-fold.c (follow_all_ssa_edges): New function.
12600 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
12601 gimple_build interface and use follow_all_ssa_edges to fold the result.
12602 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
12603 instead of checking for optabs directly.
12604 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
12605 rather than FMA_EXPRs.
12606 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
12607 call to IFN_FMA instead of an FMA_EXPR.
12609 2018-05-17 Jim Wilson <jimw@sifive.com>
12611 * expr.c (do_tablejump): When converting index to Pmode, if we have a
12612 sign extended promoted subreg, and the range does not have the sign bit
12613 set, then do a sign extend.
12615 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
12616 test, check for sign extended subreg and/or constant operands, and
12617 do a sign extend in that case.
12619 2018-05-17 Steve Ellcey <sellcey@cavium.com>
12621 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
12622 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
12624 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
12625 Change logics_shift_reg to logics_shift_imm.
12626 (thunderx2t99_fp_loadpair_basic): Delete.
12627 (thunderx2t99_fp_storepair_basic): Delete.
12628 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
12629 (thunderx2t99_asimd_polynomial): Delete.
12630 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
12631 and neon_fp_mul_d_scalar_q.
12632 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
12633 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
12634 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
12635 (thunderx2t99_asimd_lut): Add missing tbl types.
12636 (thunderx2t99_asimd_ext): Delete.
12637 (thunderx2t99_asimd_load1_1_mult): Delete.
12638 (thunderx2t99_asimd_load1_2_mult): Delete.
12639 (thunderx2t99_asimd_load1_ldp): New.
12640 (thunderx2t99_asimd_load1): New.
12641 (thunderx2t99_asimd_load2): Add missing *load2* types.
12642 (thunderx2t99_asimd_load3): New.
12643 (thunderx2t99_asimd_load4): New.
12644 (thunderx2t99_asimd_store1_1_mult): Delete.
12645 (thunderx2t99_asimd_store1_2_mult): Delete.
12646 (thunderx2t99_asimd_store2_mult): Delete.
12647 (thunderx2t99_asimd_store2_onelane): Delete.
12648 (thunderx2t99_asimd_store_stp): New.
12649 (thunderx2t99_asimd_store1): New.
12650 (thunderx2t99_asimd_store2): New.
12651 (thunderx2t99_asimd_store3): New.
12652 (thunderx2t99_asimd_store4): New.
12654 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
12656 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
12657 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
12659 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
12660 Segher Boessenkool <segher@kernel.crashing.org>
12663 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
12666 2018-05-17 Richard Biener <rguenther@suse.de>
12668 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
12669 for pruning loop and prune defs feeding only already visited PHIs.
12671 2018-05-17 Richard Biener <rguenther@suse.de>
12673 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
12675 2018-05-17 Bin Cheng <bin.cheng@arm.com>
12676 Richard Biener <rguenther@suse.de>
12678 PR tree-optimization/85793
12679 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
12680 for VMAT_ELEMENTWISE.
12682 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
12684 * internal-fn.h (lookup_internal_fn): Declare
12685 * internal-fn.c (lookup_internal_fn): New function.
12686 * gimple.c (gimple_build_call_from_tree): Handle calls to
12687 internal functions.
12688 * gimple-pretty-print.c (dump_gimple_call): Print "." before
12689 internal function names.
12690 * tree-pretty-print.c (dump_generic_node): Likewise.
12691 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
12693 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
12695 * gimple-fold.h (gimple_build): Make the function forms take
12696 combined_fn rather than built_in_function.
12697 (gimple_simplify): Likewise.
12698 * gimple-match-head.c (gimple_simplify): Likewise.
12699 * gimple-fold.c (gimple_build): Likewise.
12700 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
12701 rather than gimple_build_call_internal.
12702 (get_initial_defs_for_reduction): Likewise.
12703 (vect_create_epilog_for_reduction): Likewise.
12704 (vectorizable_live_operation): Likewise.
12706 2018-05-17 Martin Liska <mliska@suse.cz>
12708 * gimple-ssa-sprintf.c (format_directive): Do not use
12709 space in between 'G_' and '('.
12711 2018-05-17 Jakub Jelinek <jakub@redhat.com>
12714 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
12715 even if the mask is not all ones.
12718 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
12720 (ix86_gimple_fold_builtin): Likewise.
12723 * config/i386/i386.c: Include tree-vector-builder.h.
12724 (ix86_vector_shift_count): New function.
12725 (ix86_fold_builtin): Fold shift builtins by scalar count.
12726 (ix86_gimple_fold_builtin): Likewise.
12728 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
12729 _mm512_setzero): New intrinsics.
12731 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
12732 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12734 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
12735 code generation for cases where splatting a value is not useful.
12736 * simplify-rtx.c (simplify_ternary_operation): Simplify
12737 vec_merge across a vec_duplicate and a paradoxical subreg forming
12738 a vector mode to a vec_concat.
12740 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
12742 * config.gcc: Support "goldmont-plus".
12743 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
12745 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12746 PROCESSOR_GOLDMONT_PLUS.
12747 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
12748 (processor_target_table): Add "goldmont-plus".
12749 (PTA_GOLDMONT_PLUS): Define.
12750 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
12751 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
12752 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
12753 (fold_builtin_cpu): Add "goldmont-plus".
12754 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
12755 (ix86_option_override_internal): Add "goldmont-plus".
12756 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
12757 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
12758 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
12759 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
12761 2018-05-17 Richard Biener <rguenther@suse.de>
12763 PR tree-optimization/85757
12764 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
12765 remove defs that only feed that PHI from further processing.
12767 2018-05-16 Jim Wilson <jimw@sifive.com>
12769 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
12771 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
12773 2018-05-16 Mark Wielaard <mark@klomp.org>
12775 * dwarf2out.c (count_index_strings): New function.
12776 (output_indirect_strings): Call count_index_strings and generate
12777 header for dwarf_version >= 5.
12779 2018-05-16 Mark Wielaard <mark@klomp.org>
12781 * dwarf2out.c (dwarf_FORM): New function.
12782 (set_indirect_string): Use dwarf_FORM.
12783 (reset_indirect_string): Likewise.
12784 (size_of_die): Likewise.
12785 (value_format): Likewise.
12786 (output_die): Likewise.
12787 (add_skeleton_AT_string): Likewise.
12788 (output_macinfo_op): Likewise.
12789 (index_string): Likewise.
12790 (output_index_string_offset): Likewise.
12791 (output_index_string): Likewise.
12792 (count_index_strings): Likewise.
12794 2018-05-16 Carl Love <cel@us.ibm.com>
12796 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
12797 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
12799 2018-05-16 Martin Jambor <mjambor@suse.cz>
12801 * ipa-prop.c (ipa_free_all_edge_args): Remove.
12802 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
12804 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
12806 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
12807 (fnma<mode>4): Likewise.
12808 (fms<mode>4): Likewise.
12809 (fnms<mode>4): Likewise.
12810 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
12811 (aarch64_fnma<mode>4): Likewise.
12812 (aarch64_fms<mode>4): Likewise.
12813 (aarch64_fnms<mode>4): Likewise.
12814 (aarch64_fnmadd<mode>4): Likewise.
12816 2018-05-16 Jason Merrill <jason@redhat.com>
12818 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
12820 2018-05-16 Richard Biener <rguenther@suse.de>
12822 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
12823 (dump_stmt_cost): Declare.
12824 (add_stmt_cost): Dump cost we add.
12825 (add_stmt_costs): New function.
12826 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
12827 No longer exported.
12828 (vect_analyze_stmt): Adjust prototype.
12829 (vectorizable_condition): Likewise.
12830 (vectorizable_live_operation): Likewise.
12831 (vectorizable_reduction): Likewise.
12832 (vectorizable_induction): Likewise.
12833 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
12834 cost vector to pass to vectorizable_ and record afterwards.
12835 (vect_model_reduction_cost): Take cost vector argument and adjust.
12836 (vect_model_induction_cost): Likewise.
12837 (vectorizable_reduction): Likewise.
12838 (vectorizable_induction): Likewise.
12839 (vectorizable_live_operation): Likewise.
12840 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
12841 SLP_TREE_NUMBER_OF_VEC_STMTS.
12842 (vect_analyze_slp_cost_1): Remove.
12843 (vect_analyze_slp_cost): Likewise.
12844 (vect_slp_analyze_node_operations): Take visited args and
12845 a target cost vector. Avoid processing already visited stmt sets.
12846 (vect_slp_analyze_operations): Use a local cost vector to gather
12847 costs and register those of non-discarded instances.
12848 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
12849 (vect_schedule_slp_instance): Remove copying of
12850 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
12852 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
12853 adding cost. Record cost entry location.
12854 (vect_prologue_cost_for_slp_op): Function to compute cost of
12855 a constant or invariant generated for SLP vect in the prologue,
12856 split out from vect_analyze_slp_cost_1.
12857 (vect_model_simple_cost): Make static. Adjust for SLP costing.
12858 (vect_model_promotion_demotion_cost): Likewise.
12859 (vect_model_store_cost): Likewise, make static.
12860 (vect_model_load_cost): Likewise.
12861 (vectorizable_bswap): Add cost vector arg and adjust.
12862 (vectorizable_call): Likewise.
12863 (vectorizable_simd_clone_call): Likewise.
12864 (vectorizable_conversion): Likewise.
12865 (vectorizable_assignment): Likewise.
12866 (vectorizable_shift): Likewise.
12867 (vectorizable_operation): Likewise.
12868 (vectorizable_store): Likewise.
12869 (vectorizable_load): Likewise.
12870 (vectorizable_condition): Likewise.
12871 (vectorizable_comparison): Likewise.
12872 (can_vectorize_live_stmts): Likewise.
12873 (vect_analyze_stmt): Likewise.
12874 (vect_transform_stmt): Adjust calls to vectorizable_*.
12875 * tree-vectorizer.c: Include gimple-pretty-print.h.
12876 (dump_stmt_cost): New function.
12878 2018-05-16 Richard Biener <rguenther@suse.de>
12880 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
12881 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
12882 * tree-ssa-dse.c: Include tree-ssa-loop.h.
12883 (check_name): New callback.
12884 (dse_classify_store): Track cycles via a visited bitmap of PHI
12885 defs and simplify handling of in-loop and across loop dead stores
12886 and properly fail for loop-variant refs. Handle byte-tracking with
12887 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
12890 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
12892 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
12893 (vect_get_mask_type_for_stmt): Likewise.
12894 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
12896 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
12897 to determine the statement's vector type and the vector type that
12898 should be used for calculating nunits. Deal with cases in which
12899 the type has to be deferred.
12900 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
12901 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
12902 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
12903 (vect_determine_vf_for_stmt): New functions, split out from...
12904 (vect_determine_vectorization_factor): ...here.
12905 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
12906 (vect_get_mask_type_for_stmt): New functions, split out from
12907 vect_determine_vectorization_factor.
12909 2018-05-16 Richard Biener <rguenther@suse.de>
12911 * tree-cfg.c (verify_gimple_assign_ternary): Properly
12912 verify the [VEC_]COND_EXPR embedded comparison.
12914 2018-05-15 Martin Sebor <msebor@redhat.com>
12916 PR tree-optimization/85753
12917 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
12918 RECORD_TYPE in addition to ARRAY_TYPE.
12920 2018-05-15 Martin Sebor <msebor@redhat.com>
12922 PR middle-end/85643
12923 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
12925 2018-05-15 Richard Biener <rguenther@suse.de>
12927 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
12928 add by_clobber_p one. Change algorithm to collect all defs
12929 representing uses we need to walk and try reducing them to
12930 a single one before failing.
12931 (dse_dom_walker::dse_optimize_stmt): Adjust.
12933 2018-05-13 Mark Wielaard <mark@klomp.org>
12935 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
12936 (size_of_loc_descr): Likewise.
12937 (output_loc_operands): Likewise.
12938 (output_loc_operands_raw): Likewise.
12939 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
12940 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
12941 (hash_loc_operands): Likewise.
12942 (compare_loc_operands): Likewise.
12944 2018-05-14 Mark Wielaard <mark@klomp.org>
12946 * dwarf2out.c (count_index_addrs): New function.
12947 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
12949 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12951 PR tree-optimization/83648
12952 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
12953 return value as malloc candidate.
12955 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12958 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
12959 param as true in call to suggest_attribute.
12961 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
12963 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
12964 -mreadonly-in-sdata.
12966 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12968 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
12970 (aarch64_crypto_aesd_fused): Likewise.
12972 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
12974 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
12975 (movsi_aarch64): Likewise.
12976 (load_pairsi): Likewise.
12977 (load_pairdi): Likewise.
12978 (store_pairsi): Likewise.
12979 (store_pairdi): Likewise.
12980 (load_pairsf): Likewise.
12981 (load_pairdf): Likewise.
12982 (store_pairsf): Likewise.
12983 (store_pairdf): Likewise.
12984 (zero_extend): Likewise.
12985 (trunc): Swap alternatives.
12986 (fcvt_target): Add '?' to prefer w over r.
12988 2018-05-14 Jakub Jelinek <jakub@redhat.com>
12991 * config/i386/i386.md: Disallow non-commutative arithmetics in
12992 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
12993 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
12994 in the peephole2 before it.
12996 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
12998 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
12999 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
13000 (ix86_handle_option): Handle -mcldemote.
13001 * config.gcc: New header.
13002 * config/i386/cldemoteintrin.h: New file.
13003 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
13004 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13006 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13007 OPTION_MASK_ISA_CLDEMOTE.
13008 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
13009 (ix86_valid_target_attribute_inner_p): Ditto.
13010 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
13011 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
13012 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
13013 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
13014 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
13016 * config/i386/i386.opt: Add -mcldemote.
13017 * config/i386/x86intrin.h: New header.
13018 * doc/invoke.texi: Add -mcldemote.
13020 2018-05-14 Richard Biener <rguenther@suse.de>
13022 * doc/match-and-simplify.texi: Adjust :s documentation.
13024 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
13026 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
13027 intended memcpy size.
13028 (REORDER_45): Likewise.
13030 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
13032 * sort.cc: New file.
13033 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
13034 * vec.c (qsort_chk): Use gcc_qsort.
13035 * Makefile.in (OBJS-libcommon): Add sort.o.
13036 (build/sort.o): New target. Use it...
13037 (BUILD_RTL): ... here, and...
13038 (build/gencfn-macros): ... here, and...
13039 (build/genmatch): ... here.
13041 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13042 Chung-Ju Wu <jasonwucj@gmail.com>
13044 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
13045 * config/nds32/nds32-graywolf.md: New file.
13046 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
13047 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
13049 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
13050 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
13051 * config/nds32/nds32.md (pipeline_model): Add graywolf.
13052 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
13053 * config/nds32/pipelines.md: Include n15 settings.
13055 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13056 Chung-Ju Wu <jasonwucj@gmail.com>
13058 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
13059 * config/nds32/nds32-n13.md: New file.
13060 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
13061 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
13063 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
13064 * config/nds32/nds32.md (pipeline_model): Add n13.
13065 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
13066 * config/nds32/pipelines.md: Include n13 settings.
13068 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
13069 Chung-Ju Wu <jasonwucj@gmail.com>
13071 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
13072 * config/nds32/nds32-n10.md: New file.
13073 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
13074 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
13076 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
13077 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
13078 * config/nds32/nds32.md (pipeline_model): Add n10.
13079 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
13080 * config/nds32/pipelines.md: Include n10 settings.
13082 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
13083 Kito Cheng <kito.cheng@gmail.com>
13084 Chung-Ju Wu <jasonwucj@gmail.com>
13086 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
13087 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13088 Add enum values for DSP extension instructions.
13089 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
13091 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
13092 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
13093 New code iterators.
13094 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
13095 * config/nds32/nds32-dspext.md: New file for DSP implementation.
13096 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
13097 * config/nds32/nds32-intrinsic.md: Likewise.
13098 * config/nds32/nds32_intrinsic.h: Likewise.
13099 * config/nds32/nds32-md-auxiliary.c: Likewise.
13100 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
13101 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
13102 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
13103 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
13104 * config/nds32/nds32-protos.h: New declarations for DSP extension.
13105 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
13106 TYPE_DMAC in switch statement.
13107 * config/nds32/nds32.c: New checking and implementation for DSP
13108 extension instructions.
13109 * config/nds32/nds32.h: Likewise.
13110 * config/nds32/nds32.md: Likewise.
13111 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
13112 * config/nds32/predicates.md: Implement new predicates for DSP
13115 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13117 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
13118 Reformat alternatives and attributes so it is easier to identify
13119 which constraints/attributes go with which instruction.
13120 (mov<mode>_hardfloat32, FMOVE64): Likewise.
13121 (mov<mode>_softfloat32, FMOVE64): Likewise.
13122 (mov<mode>_hardfloat64, FMOVE64): Likewise.
13123 (mov<mode>_softfloat64, FMOVE64): Likewise.
13125 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
13127 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13129 (Basic PowerPC Built-in Functions): The new name of the
13130 subsection previously known as "PowerPC Built-in Functions".
13131 (Basic PowerPC Built-in Functions Available on all Configurations):
13133 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
13134 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
13135 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
13136 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
13138 2018-05-11 Martin Jambor <mjambor@suse.cz>
13141 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
13144 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13147 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
13149 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
13151 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
13152 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
13153 (ix86_handle_option): Handle -mwaitpkg.
13154 * config.gcc: New header.
13155 * config/i386/cpuid.h (bit_WAITPKG): New bit.
13156 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
13157 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
13159 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13160 OPTION_MASK_ISA_WAITPKG.
13161 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
13162 (ix86_option_override_internal): Add PTA_WAITPKG.
13163 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
13164 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
13165 IX86_BUILTIN_TPAUSE.
13166 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
13167 __builtin_ia32_umwait and __builtin_ia32_tpause.
13168 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
13169 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
13170 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
13171 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
13172 UNSPECV_TPAUSE): New.
13173 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
13174 * config/i386/i386.opt: Add -mwaitpkg.
13175 * config/i386/waitpkgintrin.h: New file.
13176 * config/i386/x86intrin.h: New header.
13177 * doc/invoke.texi: Add -mwaitpkg.
13179 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
13182 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
13184 (cortex-m0): Use armv6s-m isa.
13185 (cortex-m0plus): Likewise.
13186 (cortex-m1): Likewise.
13187 (cortex-m0.small-multiply): Likewise.
13188 (cortex-m0plus.small-multiply): Likewise.
13189 (cortex-m1.small-multiply): Likewise.
13191 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
13192 Jakub Jelinek <jakub@redhat.com>
13194 PR tree-optimization/85692
13195 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
13196 source permute as well.
13198 2018-05-11 Martin Liska <mliska@suse.cz>
13201 * doc/extend.texi: Document LLVM style format for no_sanitize
13204 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13206 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
13207 mode_supports_vsx_dform_quad to mode_supports_dq_form.
13208 (mode_supports_vsx_dform_quad): Likewise.
13209 (mode_supports_vmx_dform): Move these functions to be next to the
13210 other mode_supports functions.
13211 (mode_supports_dq_form): Likewise.
13212 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
13213 mode_supports_dq_form.
13214 (reg_offset_addressing_ok_p): Likewise.
13215 (offsettable_ok_by_alignment): Likewise.
13216 (rs6000_legitimate_offset_address_p): Likewise.
13217 (legitimate_lo_sum_address_p): Likewise.
13218 (rs6000_legitimize_address): Likewise.
13219 (rs6000_legitimize_reload_address): Likewise.
13220 (rs6000_secondary_reload_inner): Likewise.
13221 (rs6000_preferred_reload_class): Likewise.
13222 (rs6000_output_move_128bit): Likewise.
13224 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13226 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
13227 Generate SImode target register for null target.
13228 <case IX86_BUILTIN_XGETBV>: Ditto.
13229 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
13230 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
13232 2018-05-10 Carl Love <cel@us.ibm.com>
13234 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13235 dcbtt and dcbtstt if operands[2] is 0.
13237 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
13240 * config/i386/sse.md (usadv64qi): New expander.
13242 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
13244 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
13245 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
13246 -maltivec=be support.
13247 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
13248 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
13249 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
13250 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
13251 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
13252 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
13253 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
13254 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
13255 altivec_vsumsws): Adjust.
13256 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
13257 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
13258 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
13260 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
13261 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
13262 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
13263 (altivec_lve<VI_char>x): Delete expand.
13264 (*altivec_lve<VI_char>x_internal): Rename to...
13265 (altivec_lve<VI_char>x): ... this.
13266 (altivec_lvxl_<mode>): Delete expand.
13267 (*altivec_lvxl_<mode>_internal): Rename to ...
13268 (altivec_lvxl_<mode>): ... this.
13269 (altivec_stvxl_<mode>): Delete expand.
13270 (*altivec_stvxl_<mode>_internal): Rename to ...
13271 (altivec_stvxl_<mode>): ... this.
13272 (altivec_stve<VI_char>x): Delete expand.
13273 (*altivec_stve<VI_char>x_internal): Rename to ...
13274 (altivec_stve<VI_char>x): ... this.
13275 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
13276 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
13277 reduc_plus_scal_<mode>): Adjust.
13278 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
13280 (rs6000_cpu_cpp_builtins): Adjust.
13281 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
13282 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
13283 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
13284 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
13285 -maltivec=be support.
13286 (rs6000_split_vec_extract_var): Adjust.
13287 (rs6000_split_v4si_init): Adjust.
13288 (swap_selector_for_mode): Delete.
13289 (altivec_expand_lvx_be, altivec_expand_stvx_be,
13290 altivec_expand_stvex_be): Delete.
13291 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
13292 -maltivec=be support.
13293 (rs6000_gimple_fold_builtin): Ditto.
13294 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
13296 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
13297 (TARGET_DIRECT_MOVE_64BIT): Adjust.
13298 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
13299 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
13300 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
13301 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
13302 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
13303 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
13304 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
13305 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
13306 anonymous split): Adjust.
13307 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
13308 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
13310 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
13312 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
13313 when --with-gxx-include-dir is also specified.
13314 * configure: Regenerate.
13316 2018-05-09 Jim Wilson <jimw@sifive.com>
13319 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
13320 * config/riscv/t-withmultilib: New.
13321 * config/riscv/withmultilib.h: New.
13322 * doc/install.texi: Document RISC-V --with-multilib-list support.
13324 2018-05-09 Richard Biener <rguenther@suse.de>
13326 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
13328 (vect_bb_vectorization_profitable_p): Adjust. Compute
13329 actual scalar cost using the cost vector and the add_stmt_cost
13332 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13334 PR rtl-optimization/85645
13335 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
13336 in the REG_CFA_REGISTER note for LR, don't leave it empty.
13338 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13340 PR rtl-optimization/85645
13341 * shrink-wrap.c (spread_components): Return a boolean saying if
13342 anything was changed.
13343 (try_shrink_wrapping_separate): Iterate spread_components until
13344 nothing changes anymore.
13346 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13348 PR rtl-optimization/85645
13349 * regrename.c (build_def_use): Also kill the chains that include the
13350 destination of a REG_CFA_REGISTER note.
13352 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
13354 PR rtl-optimization/85645
13355 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
13356 insn that has a REG_CFA_REGISTER note.
13358 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
13360 * cfgexpand.c (expand_clobber): New function.
13361 (expand_gimple_stmt_1): Use it.
13362 * tree-vect-stmts.c (vect_clobber_variable): New function,
13364 (vectorizable_simd_clone_call): ...here.
13365 (vectorizable_store): Emit a clobber either side of an
13366 IFN_STORE_LANES sequence.
13367 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
13369 2018-05-09 Tom de Vries <tom@codesourcery.com>
13372 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
13373 (define_insn "trap_if_false"): Add exit after trap.
13375 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
13377 PR rtl-optimization/85638
13378 * bb-reorder.c: Include common/common-target.h.
13379 (create_forwarder_block): New function extracted from...
13380 (fix_up_crossing_landing_pad): ...here. Rename into...
13381 (dw2_fix_up_crossing_landing_pad): ...this.
13382 (sjlj_fix_up_crossing_landing_pad): New function.
13383 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
13384 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
13385 from both partitions and exit the loop after one iteration.
13387 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13390 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13392 (Basic PowerPC Built-in Functions): The new name of the
13393 subsection previously known as "PowerPC Built-in Functions".
13394 (Basic PowerPC Built-in Functions Available on all Configurations):
13396 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13398 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13400 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13402 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13405 2018-05-08 Jim Wilson <jimw@sifive.com>
13407 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
13408 (LD_EMUL_SUFFIX): New.
13409 (LINK_SPEC): Use it.
13411 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
13413 * doc/extend.texi (PowerPC Built-in Functions): Rename this
13415 (Basic PowerPC Built-in Functions): The new name of the
13416 subsection previously known as "PowerPC Built-in Functions".
13417 (Basic PowerPC Built-in Functions Available on all Configurations):
13419 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13421 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13423 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13425 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13428 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13431 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
13432 after cmpelim optimization.
13434 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
13436 * config.gcc: Support "goldmont".
13437 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
13438 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13439 PROCESSOR_GOLDMONT.
13440 * config/i386/i386.c (m_GOLDMONT): Define.
13441 (processor_target_table): Add "goldmont".
13442 (PTA_GOLDMONT): Define.
13443 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
13444 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
13445 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
13446 (fold_builtin_cpu): Add "goldmont".
13447 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
13448 (ix86_option_override_internal): Add "goldmont".
13449 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
13450 (processor_type): Add PROCESSOR_GOLDMONT.
13451 * config/i386/i386.md: Add CPU "glm".
13452 * config/i386/glm.md: New file.
13453 * config/i386/x86-tune.def: Add m_GOLDMONT.
13454 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
13456 2018-05-08 Jakub Jelinek <jakub@redhat.com>
13459 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
13461 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
13462 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
13463 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
13466 * config/i386/i386.c (ix86_fold_builtin): Handle
13467 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
13470 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
13471 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
13473 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
13476 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
13477 (check_arch): Likewise.
13478 (check_fpu): Return the result rather than printing it.
13479 (end arch): Fix operator precedence.
13480 (end cpu): Likewise.
13481 (END): Print the result from check_fpu.
13483 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13484 Alan Hayward <alan.hayward@arm.com>
13485 David Sherwood <david.sherwood@arm.com>
13487 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
13488 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
13489 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
13490 (*fcmuo<mode>_and): New patterns.
13492 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
13494 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
13495 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
13496 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
13497 (cmp_op, sve_imm_con): New code attributes.
13498 (SVE_COND_INT_CMP, imm_con): Delete.
13499 (cmp_op): Remove above unspecs from int attribute.
13500 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
13502 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
13503 comparison-specific unspecs.
13504 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
13505 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
13506 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
13507 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
13508 (*vec_fcm<cmp_op><mode>): Rename to...
13509 (*fcm<cmp_op><mode>): ...this and adjust likewise.
13510 (*vec_fcmuo<mode>): Rename to...
13511 (*fcmuo<mode>): ...this and adjust likewise.
13512 (*pred_fcm<cmp_op><mode>): New pattern.
13513 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
13514 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
13516 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
13518 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
13519 (aarch64_emit_sve_predicated_cond): New function.
13520 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
13521 (aarch64_emit_unspec_cond_or): Replace with...
13522 (aarch64_emit_sve_or_conds): ...this new function. Use
13523 aarch64_emit_sve_ptrue_op for the individual comparisons and
13524 aarch64_emit_binop to OR them together.
13525 (aarch64_emit_inverted_unspec_cond): Replace with...
13526 (aarch64_emit_sve_inverted_cond): ...this new function. Use
13527 aarch64_emit_sve_ptrue_op for the comparison and
13528 aarch64_emit_unop to invert the result.
13529 (aarch64_expand_sve_vec_cmp_float): Update after the above
13530 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
13532 2018-05-07 Nathan Sidwell <nathan@acm.org>
13534 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
13535 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
13536 (Backwards Compatibility): Likewise.
13538 2018-05-07 Luis Machado <luis.machado@linaro.org>
13542 2018-05-07 Luis Machado <luis.machado@linaro.org>
13544 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13545 <prefetch_dynamic_strides>: New const bool field.
13546 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13547 prefetch_dynamic_strides.
13548 (exynosm1_prefetch_tune): Likewise.
13549 (thunderxt88_prefetch_tune): Likewise.
13550 (thunderx_prefetch_tune): Likewise.
13551 (thunderx2t99_prefetch_tune): Likewise.
13552 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13554 (aarch64_override_options_internal): Update to set
13555 PARAM_PREFETCH_DYNAMIC_STRIDES.
13556 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13557 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13558 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13559 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13560 prefetch-dynamic-strides setting.
13562 2018-05-07 Luis Machado <luis.machado@linaro.org>
13564 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13565 <minimum_stride>: New const int field.
13566 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13567 minimum_stride field.
13568 (exynosm1_prefetch_tune): Likewise.
13569 (thunderxt88_prefetch_tune): Likewise.
13570 (thunderx_prefetch_tune): Likewise.
13571 (thunderx2t99_prefetch_tune): Likewise.
13572 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13573 (aarch64_override_options_internal): Update to set
13574 PARAM_PREFETCH_MINIMUM_STRIDE.
13575 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13576 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13577 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13578 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13579 stride is constant and is below the minimum stride threshold.
13581 2018-05-07 Luis Machado <luis.machado@linaro.org>
13583 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
13586 2018-05-07 Luis Machado <luis.machado@linaro.org>
13588 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13589 <prefetch_dynamic_strides>: New const bool field.
13590 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13591 prefetch_dynamic_strides.
13592 (exynosm1_prefetch_tune): Likewise.
13593 (thunderxt88_prefetch_tune): Likewise.
13594 (thunderx_prefetch_tune): Likewise.
13595 (thunderx2t99_prefetch_tune): Likewise.
13596 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13598 (aarch64_override_options_internal): Update to set
13599 PARAM_PREFETCH_DYNAMIC_STRIDES.
13600 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13601 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13602 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13603 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13604 prefetch-dynamic-strides setting.
13606 2018-05-07 Luis Machado <luis.machado@linaro.org>
13608 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13609 <minimum_stride>: New const int field.
13610 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13611 minimum_stride field.
13612 (exynosm1_prefetch_tune): Likewise.
13613 (thunderxt88_prefetch_tune): Likewise.
13614 (thunderx_prefetch_tune): Likewise.
13615 (thunderx2t99_prefetch_tune): Likewise.
13616 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13617 (aarch64_override_options_internal): Update to set
13618 PARAM_PREFETCH_MINIMUM_STRIDE.
13619 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13620 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13621 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13622 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13623 stride is constant and is below the minimum stride threshold.
13625 2018-05-06 Jakub Jelinek <jakub@redhat.com>
13628 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
13629 the type is addressable. Don't force op into register if it has
13632 2018-05-05 Roland McGrath <mcgrathr@google.com>
13635 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
13636 any section for which we don't know a specific type it should have,
13637 regardless of name. Previously this was done only for the exact
13638 names ".init_array", ".fini_array", and ".preinit_array".
13639 (default_elf_asm_named_section): Add comment about
13640 relationship with default_section_type_flags and SECTION_NOTYPE.
13641 (get_section): Don't consider it a type conflict if one side has
13642 SECTION_NOTYPE and the other doesn't, as long as neither has the
13643 SECTION_BSS et al used in the default_section_type_flags logic.
13645 2018-05-05 Tom de Vries <tom@codesourcery.com>
13648 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
13649 (workaround_barsyncs): New function.
13650 (nvptx_reorg): Use workaround_barsyncs.
13651 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13652 (define_expand "nvptx_membar_cta"): New define_expand.
13653 (define_insn "*nvptx_membar_cta"): New insn.
13655 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
13657 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
13658 To improve optimization opportunities.
13659 * builtin-types.def: The new needed builtin types for the above.
13661 2018-05-04 Richard Biener <rguenther@suse.de>
13663 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
13664 * gimple-ssa-store-merging.c
13665 (imm_store_chain_info::output_merged_store): Remove redundant create,
13666 release split_store vector contents on failure.
13667 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
13668 scalar stmt vector on cache hit.
13670 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
13672 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
13674 * config.gcc (powerpc-xilinx-eabi*): Remove.
13675 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
13677 (fusion_addis_mem_combo_load): Ditto.
13678 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
13680 (rs6000_cpu_cpp_builtins): Ditto.
13681 * config/rs6000/rs6000-linux.c
13682 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
13683 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
13684 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
13686 (rs6000_setup_reg_addr_masks): Ditto.
13687 (rs6000_init_hard_regno_mode_ok): Ditto.
13688 (rs6000_option_override_internal): Ditto.
13689 (legitimate_lo_sum_address_p): Ditto.
13690 (rs6000_legitimize_address): Ditto.
13691 (rs6000_legitimize_reload_address): Ditto.
13692 (rs6000_legitimate_address_p): Ditto.
13693 (abi_v4_pass_in_fpr): Ditto.
13694 (setup_incoming_varargs): Ditto.
13695 (rs6000_gimplify_va_arg): Ditto.
13696 (rs6000_split_multireg_move): Ditto.
13697 (rs6000_savres_strategy): Ditto.
13698 (rs6000_emit_prologue_components): Ditto.
13699 (rs6000_emit_epilogue_components): Ditto.
13700 (rs6000_emit_prologue): Ditto.
13701 (rs6000_emit_epilogue): Ditto.
13702 (rs6000_elf_file_end): Ditto.
13703 (rs6000_function_value): Ditto.
13704 (rs6000_libcall_value): Ditto.
13705 * config/rs6000/rs6000.h: Ditto.
13706 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
13707 (TARGET_MINMAX): ... this. New.
13708 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
13709 * config/rs6000/rs6000.md: Remove Xilinx FP support.
13710 (*movsi_internal1_single): Delete.
13711 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
13712 mfpu=, mxilinx-fpu): Delete.
13713 * config/rs6000/singlefp.h: Delete.
13714 * config/rs6000/sysv4.h: Remove Xilinx FP support.
13715 * config/rs6000/t-rs6000: Ditto.
13716 * config/rs6000/t-xilinx: Delete.
13717 * config/rs6000/titan.md: Adjust for fp_type removal.
13718 * config/rs6000/vsx.md: Remove Xilinx FP support.
13719 (VStype_simple): Delete.
13720 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
13721 * config/rs6000/xfpu.h: Delete.
13722 * config/rs6000/xfpu.md: Delete.
13723 * config/rs6000/xilinx.h: Delete.
13724 * config/rs6000/xilinx.opt: Delete.
13725 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
13726 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
13728 2018-05-04 Tom de Vries <tom@codesourcery.com>
13731 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
13734 2018-05-04 Richard Biener <rguenther@suse.de>
13736 PR middle-end/85627
13737 * tree-complex.c (update_complex_assignment): We are always in SSA form.
13738 (expand_complex_div_wide): Likewise.
13739 (expand_complex_operations_1): Likewise.
13740 (expand_complex_libcall): Preserve EH info of the original stmt.
13741 (tree_lower_complex): Handle removed blocks.
13742 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
13743 on complex multiplication and division libcall builtins.
13745 2018-05-04 Richard Biener <rguenther@suse.de>
13747 PR middle-end/85574
13748 * fold-const.c (negate_expr_p): Restrict negation of operand
13749 zero of a division to when we know that can happen without
13751 (fold_negate_expr_1): Likewise.
13753 2018-05-04 Jakub Jelinek <jakub@redhat.com>
13756 * real.h (real_nextafter): Declare.
13757 * real.c (real_nextafter): New function.
13758 * fold-const-call.c (fold_const_nextafter): New function.
13759 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
13760 CASE_CFN_NEXTTOWARD.
13761 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
13762 even when arg1_mode is different from arg0_mode.
13764 2018-05-03 Nathan Sidwell <nathan@acm.org>
13766 * doc/extend.texi (Deprecated Features): Remove
13767 -ffriend-injection.
13768 (Backwards Compatibility): Likewise.
13769 * doc/invoke.texi (C++ Language Options): Likewise.
13770 (C++ Dialect Options): Likewise.
13772 2018-05-03 Jakub Jelinek <jakub@redhat.com>
13775 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
13776 _mm512_mask_mullox_epi64): New intrinsics.
13778 2018-05-03 Tom de Vries <tom@codesourcery.com>
13781 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13782 dump files): Add offload-tree.
13784 2018-05-03 Richard Biener <rguenther@suse.de>
13786 PR tree-optimization/85615
13787 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
13788 to loops not nested in BBs loop father to avoid creating multi-entry
13791 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13793 PR tree-optimization/70291
13794 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
13795 arguments. Change return type to tree. Emit libcall as a new
13796 statement rather than replacing existing one when inplace_p is true.
13797 (expand_complex_multiplication_components): New function.
13798 (expand_complex_multiplication): Expand floating-point complex
13799 multiplication using the above.
13800 (expand_complex_division): Rename inner_type parameter to type.
13801 Update expand_complex_libcall call-site.
13802 (expand_complex_operations_1): Update expand_complex_multiplication
13803 and expand_complex_division call-sites.
13805 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13808 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13809 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13810 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
13811 the highest significant bit of the shift count mask is clear. In
13812 check whether and[sq]i3 is needed verify that all significant bits
13813 of the shift count other than the highest are set.
13815 2018-05-02 Tom de Vries <tom@codesourcery.com>
13818 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
13819 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
13820 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
13821 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
13822 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
13823 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
13824 __builtin_goacc_parlevel_size.
13826 2018-05-02 Richard Biener <rguenther@suse.de>
13828 PR tree-optimization/85597
13829 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
13830 do not use split vect_get_vec_defs call but call vect_get_slp_defs
13833 2018-05-02 Tom de Vries <tom@codesourcery.com>
13836 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13837 dump files): Add ltrans-tree.
13839 2018-05-02 Tom de Vries <tom@codesourcery.com>
13842 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13843 dump files): Add wpa-ipa.
13845 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
13847 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
13848 powerpc*-*-linux*paired* target.
13849 * config/rs6000/750cl.h: Delete.
13850 * config/rs6000/paired.h: Delete.
13851 * config/rs6000/paired.md: Delete.
13852 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
13854 * config/rs6000/rs6000-builtin.def: Remove paired float support.
13855 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
13856 comment. Remove paired float support.
13857 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
13858 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
13860 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
13861 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
13863 * config/rs6000/rs6000.c: Remove paired float support.
13864 (paired_expand_vector_init, paired_expand_vector_move,
13865 paired_emit_vector_compare, paired_emit_vector_cond_expr,
13866 (paired_expand_lv_builtin, paired_expand_stv_builtin,
13867 paired_expand_builtin, paired_expand_predicate_builtin,
13868 paired_init_builtins): Delete.
13869 * config/rs6000/rs6000.h: Remove paired float support.
13870 * config/rs6000/rs6000.md: Remove paired float support.
13871 (move_from_CR_ov_bit): Delete.
13872 * config/rs6000/rs6000.opt (mpaired): Delete.
13873 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
13874 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
13876 2018-05-02 Richard Biener <rguenther@suse.de>
13878 PR middle-end/85567
13879 * gimplify.c (gimplify_save_expr): When in SSA form allow
13880 SAVE_EXPRs to compute to SSA vars.
13882 2018-05-02 Jakub Jelinek <jakub@redhat.com>
13885 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13886 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13887 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
13888 clobber operands[2], instead use a new pseudo. Formatting fixes.
13890 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
13892 PR tree-optimization/85586
13893 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
13894 exit early for statements in the same group if the accesses are
13897 2018-05-02 Tom de Vries <tom@codesourcery.com>
13900 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
13903 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
13905 PR tree-optimization/85143
13906 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
13908 2018-05-01 Tom de Vries <tom@codesourcery.com>
13911 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
13912 not found" error message.
13914 2018-05-01 Tom de Vries <tom@codesourcery.com>
13917 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
13918 * vec.c (test_ordered_remove_if): New function.
13919 (vec_c_tests): Call test_ordered_remove_if.
13920 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
13921 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
13922 * tree-vect-patterns.c (vect_pattern_recog_1): Use
13923 VEC_ORDERED_REMOVE_IF.
13925 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13927 PR tree-optimization/82665
13928 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
13929 pointer subtraction where arguments come from a memchr call.
13931 2018-05-01 Jakub Jelinek <jakub@redhat.com>
13933 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
13934 --push-state --as-needed and --pop-state instead of --as-needed and
13935 --no-as-needed if ld supports it.
13936 * configure: Regenerated.
13939 * doc/install.texi2html: Replace _002d with - and _002a with * in
13940 generated html files using sed.
13942 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13945 * gcc-rich-location.c (blank_line_before_p): New function.
13946 (use_new_line): New function.
13947 (gcc_rich_location::add_fixit_insert_formatted): New function.
13948 * gcc-rich-location.h
13949 (gcc_rich_location::add_fixit_insert_formatted): New function.
13951 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13953 * selftest.c (assert_streq): Rename "expected" and "actual" to
13954 "val1" and "val2". Extend NULL-handling to cover both inputs
13955 symmetrically, while still requiring both to be non-NULL for a pass.
13956 * selftest.h (assert_streq): Rename "expected" and "actual" to
13958 (ASSERT_EQ): Likewise.
13959 (ASSERT_EQ_AT): Likewise.
13960 (ASSERT_KNOWN_EQ): Likewise.
13961 (ASSERT_KNOWN_EQ_AT): Likewise.
13962 (ASSERT_NE): Likewise.
13963 (ASSERT_MAYBE_NE): Likewise.
13964 (ASSERT_MAYBE_NE_AT): Likewise.
13965 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
13966 the assertion to pass.
13967 (ASSERT_STREQ_AT): Likewise.
13969 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
13971 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
13972 interaction with -pie.
13974 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13976 * selftest.h: Fix alphabetization of per-source-file selftest
13979 2018-04-30 Jason Merrill <jason@redhat.com>
13981 PR c++/61982 - dead stores to destroyed objects.
13982 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
13985 2018-04-30 Jason Merrill <jason@redhat.com>
13987 * tree.c (build_clobber): New.
13988 * tree.h: Declare it.
13989 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
13991 2018-04-30 David Malcolm <dmalcolm@redhat.com>
13993 * diagnostic-show-locus.c (layout::layout): Update for
13994 location_get_source_line returning a char_span.
13995 (struct char_span): Move to input.h.
13996 (struct correction): Update for fields in char_span becoming
13998 (struct source_line): Update for location_get_source_line
13999 returning a char_span.
14000 (layout::print_line): Likewise.
14001 * edit-context.c (edited_file::print_content): Likewise.
14002 (edited_file::print_diff_hunk): Likewise.
14003 (edited_file::print_run_of_changed_lines): Likewise.
14004 (edited_file::get_num_lines): Likewise.
14005 (edited_line::edited_line): Likewise.
14006 * final.c (asm_show_source): Likewise.
14007 * input.c (location_get_source_line): Convert return type
14008 from const char * to char_span, losing the final "line_len"
14010 (dump_location_info): Update for the above.
14011 (get_substring_ranges_for_loc): Likewise. Use a char_span
14012 when handling the literal within the line.
14013 (test_reading_source_line): Update for location_get_source_line
14014 returning a char_span.
14015 * input.h (class char_span): Move here from
14016 diagnostic-show-locus.c, converting from a struct to a class.
14017 Make data members private.
14018 (char_span::operator bool): New.
14019 (char_span::length): New.
14020 (char_span::get_buffer): New.
14021 (char_span::operator[]): New.
14022 (char_span::subspan): Make const.
14023 (char_span::xstrdup): New.
14024 (location_get_source_line): Convert return type from const char *
14025 to char_span, losing the final "line_size" param.
14027 2018-04-30 Jan Hubicka <jh@suse.cz>
14029 * lto-wrapper.c (ltrans_priorities): New static var.
14030 (cmp_priority): New.
14031 (run_gcc): Read priorities and if doing parallel build order
14032 the Makefile by them.
14034 2018-04-30 David Malcolm <dmalcolm@redhat.com>
14036 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
14038 2018-04-30 Richard Biener <rguenther@suse.de>
14040 * tree-cfg.c (verify_address): Remove base argument, add
14041 flag whether to check TREE_ADDRESSABLE and do that.
14042 (verify_expr): Remove.
14043 (verify_types_in_gimple_reference): Add pieces from verify_expr.
14044 (verify_gimple_assign_single): Likewise.
14045 (verify_gimple_switch): Likewise.
14046 (verify_expr_location_1): Dereference tp once. Add (disabled)
14047 piece from verify_expr.
14048 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
14050 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14052 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
14054 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14056 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
14057 (small_data_pattern): Likewise.
14058 (arc_rewrite_small_data): Likewise.
14059 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
14060 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
14061 (get_symbol_alignment): New function.
14062 (legitimate_small_data_address_p): Likewise.
14063 (legitimate_scaled_address): Update, call
14064 legitimate_small_data_address_p.
14065 (output_sdata): New static variable.
14066 (arc_print_operand): Update how we handle small data operands.
14067 (arc_print_operand_address): Likewise.
14068 (arc_legitimate_address_p): Update, use
14069 legitimate_small_data_address_p.
14070 (arc_rewrite_small_data_p): Remove.
14071 (arc_rewrite_small_data_1): Likewise.
14072 (arc_rewrite_small_data): Likewise.
14073 (small_data_pattern): Likewise.
14074 (compact_sda_memory_operand): Update to use
14075 legitimate_small_data_address_p and get_symbol_alignment.
14076 (prepare_move_operands): Don't rewite sdata pattern.
14077 (prepare_extend_operands): Remove.
14078 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
14080 (zero_extendqisi2): Likewise.
14081 (zero_extendhisi2): Likewise.
14082 (extendqihi2): Likewise.
14083 (extendqisi2): Likewise.
14084 (extendhisi2): Likewise.
14085 (addsi3): Likewise.
14086 (subsi3): Likewise.
14087 (andsi3): Likewise.
14088 * config/arc/constraints.md (Usd): Change it to memory constraint.
14090 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
14092 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
14093 as source of std instructions.
14094 * config/arc/arc.md (movsi_insn): Update pattern predicate to
14095 allow 6-bit constants as source for store instructions.
14096 (movdi_insn): Update instruction pattern to allow 6-bit constants
14097 as source for store instructions.
14099 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
14101 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
14103 2018-04-30 Nathan Sidwell <nathan@acm.org>
14104 Sandra Loosemore <sandra@codesourcery.com>
14106 * dumpfile.c (dump_open): Allow '-' for stdout.
14107 * doc/invoke.texi (Developer Options): Document dump filename
14108 determination early. Document stdin/stdout selection.
14110 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
14112 Microblaze Target: PIC data text relative
14114 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
14115 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
14117 * config/microblaze/microblaze.h (microblaze_constant_address_p):
14118 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
14119 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
14120 New addressing mode for data-text relative position indepenedent code.
14121 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
14122 'ADDRESS_SYMBOLIC_TXT_REL'.
14123 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
14124 (microblaze_legitimate_pic_operand): Exclude function calls from
14125 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
14126 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
14128 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14129 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14130 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
14131 for 'address + offset'.
14132 (microblaze_expand_prologue): Add new function prologue call for
14134 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
14135 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
14136 table in case of TARGET_PIC_DATA_TEXT_REL.
14137 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
14138 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
14139 Add new macros 'UNSPEC_TEXT',
14140 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
14141 + exclude function calls from 'UNSPEC_PLT' in case of data text
14143 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14144 new target hook for generating address diff vector tables in case of
14146 * doc/tm.texi : Regenerate.
14147 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
14148 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
14149 of addr diff vector generation.
14150 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14151 target hook definition.
14152 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
14153 Add default function for generate_pic_addr_diff_vec -> flag_pic.
14154 * doc/invoke.texi (Add new pic option): Add new microblaze pic
14155 option for data text relative.
14157 2018-04-30 Richard Biener <rguenther@suse.de>
14159 * tree-chrec.h (evolution_function_is_constant_p): Remove
14161 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
14163 2018-04-30 Richard Biener <rguenther@suse.de>
14166 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
14168 2018-04-30 Richard Biener <rguenther@suse.de>
14170 PR tree-optimization/28364
14171 PR tree-optimization/85275
14172 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
14173 copying first exit test.
14175 2018-04-28 Mark Wielaard <mark@klomp.org>
14177 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
14178 dwarf_version >= 5.
14179 (dwarf_AT): Handle DW_AT_addr_base.
14180 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
14182 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
14185 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
14186 (*ashl<dwi>3_doubleword_mask_1): Ditto.
14187 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
14188 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
14190 2018-04-28 Richard Biener <rguenther@suse.de>
14192 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
14193 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
14194 to reflect use. Only add interesting stmts.
14196 2018-04-27 Martin Jambor <mjambor@suse.cz>
14199 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
14200 the jump function allows for passing through aggregate values.
14202 2018-04-27 David Malcolm <dmalcolm@redhat.com>
14204 * input.h (in_system_header_at): Convert from macro to inline
14206 (from_macro_expansion_at): Likewise.
14207 (from_macro_definition_at): Likewise.
14209 2018-04-27 Jeff Law <law@redhat.com>
14211 * config.gcc: Mark tile* targets as deprecated/obsolete.
14213 2018-04-27 Richard Biener <rguenther@suse.de>
14215 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
14218 2018-04-27 Richard Biener <rguenther@suse.de>
14220 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
14222 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
14224 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
14225 with Yd constraint. Set "preferred_for_speed" attribute from
14226 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
14227 with Yd constraint.
14228 (*movdi_internal): Ditto.
14229 (movti_interunit splitters): Remove
14230 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
14231 (movdi_interunit splitters): Ditto.
14232 * config/i386/constraints.md (Ye): Remove.
14233 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
14235 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14238 * config/aarch64/constraints.md (Usg): Limit to 31.
14239 (Usj): Limit to 63.
14241 2018-04-27 Jakub Jelinek <jakub@redhat.com>
14243 PR tree-optimization/85529
14244 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
14245 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
14246 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
14247 zero extension or masking of the MSB bit.
14248 (optimize_range_tests): Add FIRST_BB argument, pass it through
14249 to optimize_range_tests_var_bound.
14250 (maybe_optimize_range_tests, reassociate_bb): Adjust
14251 optimize_range_tests callers.
14253 2018-04-26 Richard Biener <rguenther@suse.de>
14254 Jakub Jelinek <jakub@redhat.com>
14256 * cgraph.h (symbol_table): Just declare debug method here.
14257 * symtab.c (symbol_table::debug): Define.
14259 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
14261 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
14263 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
14265 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
14266 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
14267 (*movdi_internal): Substitute Yi and Yj constraint with x
14268 and Ym and Yn constraint with y constraint. Update "isa"
14269 attribute and set "preferred_for_speed" attribute from
14270 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
14271 (*movsi_internal): Ditto.
14272 (*movdf_internal): Ditto.
14273 (*movsf_internal): Ditto.
14274 (*zero_extendsidi2): Ditto.
14275 * config/i386/sse.md (vec_set<mode>_0): Ditto.
14276 (sse2_loadld): Ditto.
14277 (*vec_extract<ssevecmodelower>_0): Ditto.
14278 (*vec_extractv4si_0_zext_sse4): Ditto.
14279 (vec_concatv2di): Ditto.
14280 (*vec_dup<mode>): Ditto.
14281 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
14282 * config/i386/constraints.md (Yi): Remove.
14287 2018-04-26 Nathan Sidwell <nathan@acm.org>
14289 * dumpfile.c (dump_open): New.
14290 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
14291 (dump_finish): Detect stdio/stderr by value not name.
14293 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
14295 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
14297 2018-04-26 Tom de Vries <tom@codesourcery.com>
14300 * config/nvptx/nvptx.c (verify_neutering_jumps)
14301 (verify_neutering_labels): New function
14302 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
14304 2018-04-26 Tom de Vries <tom@codesourcery.com>
14307 * config/nvptx/nvptx.c (needs_neutering_p): New function.
14308 (nvptx_single): Use needs_neutering_p to skip over insns that do not
14311 2018-04-26 Richard Biener <rguenther@suse.de>
14312 Tom de Vries <tom@codesourcery.com>
14315 * lto-streamer-out.c (output_function): Fixup loops if required to match
14316 discovery done in the reader.
14318 2018-04-26 Richard Biener <rguenther@suse.de>
14320 PR tree-optimization/85116
14321 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
14322 have a loop exit from the single latch predecessor. Remove
14323 case of header with just condition.
14324 (ch_base::copy_headers): Exclude infinite loops from any
14326 (pass_ch::execute): Record exits.
14328 2018-04-26 Richard Biener <rguenther@suse.de>
14330 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
14331 prologue cost vector and pass it to vect_get_load_cost.
14332 (vect_get_peeling_costs_all_drs): Likewise.
14333 (vect_peeling_hash_get_lowest_cost): Likewise.
14334 (vect_enhance_data_refs_alignment): Likewise.
14336 2018-04-26 Richard Biener <rguenther@suse.de>
14338 PR middle-end/85450
14339 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
14340 checking of integer<->pointer conversions.
14341 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
14342 sign-/zero-extending pointer types.
14343 (expand_omp_for_static_chunk): Likewise.
14345 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
14346 Jean Lee <xiaoyur347@gmail.com>
14348 * config/mips/mips.c (mips_asan_shadow_offset): New function.
14349 (TARGET_ASAN_SHADOW_OFFSET): Define.
14350 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
14351 true for -fsanitize=address.
14353 2018-04-25 Mark Wielaard <mark@klomp.org>
14355 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
14358 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14360 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
14361 than "alu", remove explicit "memory" and "imm_disp" attributes.
14362 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
14364 PR middle-end/85414
14365 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
14366 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
14367 gen_lowpart_no_emit.
14369 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
14372 * config/i386/i386.c (ix86_expand_builtin): Change memory
14373 operand to XI, extend p0 to Pmode.
14374 * config/i386/i386.md: Change unspec volatile and operand
14375 1 mode to XI, change operand 0 mode to P.
14377 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14379 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
14380 GET_MODE_MASK before any checking.
14381 (nds32_can_use_bset_p): Likewise.
14382 (nds32_can_use_btgl_p): Likewise.
14384 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14386 * config/nds32/nds32-doubleword.md: New define_split pattern for
14387 illegal register number.
14389 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14391 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
14393 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
14395 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
14397 2018-04-25 Richard Biener <rguenther@suse.de>
14399 * lto-streamer.h (LTO_major_version): Bump to 8.
14401 2018-04-25 Jakub Jelinek <jakub@redhat.com>
14403 * BASE-VER: Set to 9.0.0.
14405 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
14407 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
14408 in __abskf2 and __powikf2.
14410 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14413 * config/aarch64/constraints.md (Usg, Usj): New constraints.
14414 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
14415 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
14416 Use the above on operand 2. Reindent.
14417 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
14419 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14422 * common/config/i386/i386-common.c (ix86_handle_option): Don't
14424 * config/i386/i386.opt (mcet): Removed.
14425 * doc/install.texi: Remove -mcet documentation.
14426 * doc/invoke.texi: Likewise.
14428 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
14431 * doc/install.texi: Remove -mcet from bootstrap-cet.
14433 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14436 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
14437 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
14441 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
14442 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
14443 containing a CONST_VECTOR.
14445 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
14447 * doc/install.texi: Update newlib dependency for nvptx.
14449 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14452 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
14453 instead of INTVAL when shifting x left.
14455 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
14457 PR tree-optimization/85478
14458 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
14459 vect_grouped_store_supported for single element vectors.
14461 2018-04-24 Richard Biener <rguenther@suse.de>
14464 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
14465 load cost increase to the case of non-constant step.
14467 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14470 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
14471 destination if any_malformed_asm.
14473 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
14475 PR middle-end/85496
14476 * expr.c (store_field): In the bitfield case, if the value comes from
14477 a function call and is returned in registers by means of a PARALLEL,
14478 do not change the mode of the temporary unless BLKmode and VOIDmode.
14480 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
14482 PR rtl-optimization/85423
14483 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
14484 dependencies to debug insns when the previous insn is non-debug.
14486 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14488 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
14489 enums into a single definition.
14490 (fls): Fix predicates and printing.
14493 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
14495 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
14496 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
14497 and short u6 immediate.
14498 (check_if_valid_sleep_operand): Remove.
14499 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
14501 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14503 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
14504 flag_always_save_lp condition.
14505 * config/nds32/nds32.opt (malways-save-lp): New option.
14507 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14509 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
14510 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
14511 * config/nds32/nds32.h
14512 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
14513 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
14515 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14517 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
14518 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
14520 2018-04-22 Shiva Chen <shiva0217@gmail.com>
14521 Chung-Ju Wu <jasonwucj@gmail.com>
14523 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
14525 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
14526 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
14528 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14530 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
14532 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14534 * config/nds32/nds32-protos.h (nds32_data_alignment,
14535 nds32_local_alignment): Declare.
14536 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
14537 nds32_local_alignment): New functions.
14538 (TARGET_CONSTANT_ALIGNMENT): Define.
14539 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
14541 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14543 * config/nds32/nds32.c
14544 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
14545 (TARGET_MODES_TIEABLE_P): Likewise.
14547 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
14549 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
14550 level Ofast and Og.
14552 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
14553 Chung-Ju Wu <jasonwucj@gmail.com>
14555 * config/nds32/constants.md (unspec_volatile_element): Add enum values
14556 for unaligned access.
14557 * config/nds32/nds32-intrinsic.c: Implementation of expanding
14559 * config/nds32/nds32-intrinsic.md: Likewise.
14560 * config/nds32/nds32_intrinsic.h: Likewise.
14561 * config/nds32/nds32.h (nds32_builtins): Likewise.
14562 * config/nds32/nds32.opt (munaligned-access): New option.
14563 * config/nds32/nds32.c (nds32_asm_file_start): Display
14564 flag_unaligned_access status.
14566 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14568 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
14569 -mno-relax is present.
14570 * config/riscv/linux.h (LINK_SPEC): Ditto.
14572 2018-04-20 Martin Sebor <msebor@redhat.com>
14575 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
14577 (gimple_fold_builtin_stxcpy_chk): Same.
14578 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
14580 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
14583 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
14584 __powikf2 when long double is IEEE 128-bit.
14586 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
14588 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
14589 step to make sure stack always aligned.
14591 2018-04-20 Carl Love <cel@us.ibm.com>
14594 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
14595 size check for arg0.
14597 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
14598 Tom de Vries <tom@codesourcery.com>
14601 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
14602 Emit insns for calls too.
14603 (nvptx_find_par): Always look for worker-level predecessor insn.
14604 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
14606 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
14607 (nvptx_process_pars): Propagate frames for calls.
14609 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
14612 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
14614 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
14615 (ix86_handle_option): Don't handle OPT_mibt.
14616 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
14618 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
14620 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
14621 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
14622 (ix86_target_macros): Define __CET__ with flag_cf_protection
14623 for -fcf-protection.
14624 * config/i386/i386.c (isa2_opts): Remove -mibt.
14625 * config/i386/i386.h (TARGET_IBT): Removed.
14626 (TARGET_IBT_P): Likewise.
14627 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
14628 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
14629 * config/i386/i386.opt (mcet): Update help message.
14630 (mshstk): Likewise.
14632 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
14633 -mcet as an alias for -mshstk.
14635 2018-04-20 Richard Biener <rguenther@suse.de>
14637 PR middle-end/85475
14638 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
14639 complexity by forcing a single use of the multiply operand.
14641 2018-04-20 Martin Jambor <mjambor@suse.cz>
14644 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
14645 recursion dependency to only apply to non-clones.
14647 2018-04-20 Martin Jambor <mjambor@suse.cz>
14650 * ipa-cp.c (create_specialized_node): Check that clones of
14651 self-recursive edges exist during IPA-CP.
14653 2018-04-19 Toon Moene <toon@moene.org>
14655 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
14658 2018-04-19 Jakub Jelinek <jakub@redhat.com>
14660 PR tree-optimization/85467
14661 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
14662 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
14663 VECTOR_CST element to type.
14665 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14668 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
14669 * config/i386/i386.md (builtin_setjmp_setup): Removed.
14670 (builtin_longjmp): Likewise.
14671 (save_stack_nonlocal): New pattern.
14672 (restore_stack_nonlocal): Likewise.
14674 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14677 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
14678 Replace ASM_OUTPUT_LABEL with fprintf.
14680 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
14683 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
14684 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
14685 * config/i386/i386-c.c (ix86_target_macros_internal): Also
14686 define __IBT__ and __SHSTK__ for -fcf-protection.
14687 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
14689 (ix86_trampoline_init): Likewise.
14690 (x86_output_mi_thunk): Likewise.
14691 (ix86_notrack_prefixed_insn_p): Likewise.
14692 (ix86_option_override_internal): Don't disallow -fcf-protection.
14693 * config/i386/i386.md (rdssp<mode>): Also enable for
14695 (incssp<mode>): Likewise.
14696 (nop_endbr): Likewise.
14697 * config/i386/i386.opt (mcet): Change help message to built-in
14700 (mshstk): Likewise.
14701 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
14702 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
14703 enable CET built-in functions.
14705 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
14707 * common/config/i386/i386-common.c
14708 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
14709 OPTION_MASK_ISA_MOVDIRI_UNSET,
14710 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
14711 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
14712 * config.gcc (movdirintrin.h): New header.
14713 * config/i386/cpuid.h (bit_MOVDIRI,
14714 bit_MOVDIR64B): New bits.
14715 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
14717 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
14718 (VOID, PVOID, PCVOID)): New function types.
14719 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
14720 __builtin_ia32_directstoreu_u64,
14721 __builtin_ia32_movdir64b): New builtins.
14722 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
14723 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
14725 (ix86_valid_target_attribute_inner_p): Ditto.
14726 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
14727 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
14728 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
14729 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
14730 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
14731 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
14732 (movdiri<mode>, movdir64b_<mode>): New.
14733 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
14734 * config/i386/immintrin.h: Include movdirintrin.h.
14735 * config/i386/movdirintrin.h: New file.
14736 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
14738 2018-04-19 Richard Biener <rguenther@suse.de>
14740 PR middle-end/85455
14741 * cfg.c (clear_bb_flags): When loop state says we have
14742 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
14744 2018-04-19 Richard Biener <rguenther@suse.de>
14746 PR tree-optimization/84737
14747 * tree-vect-data-refs.c (vect_copy_ref_info): New function
14748 copying restrict info.
14749 (vect_setup_realignment): Use it.
14750 * tree-vectorizer.h (vect_copy_ref_info): Declare.
14751 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
14752 the first DR to all generated stores.
14753 (vectorizable_load): Likewise for loads.
14755 2018-04-19 Jakub Jelinek <jakub@redhat.com>
14757 PR tree-optimization/85446
14758 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
14759 the integral and pointer types to have the same precision.
14761 * doc/install.texi: Document --disable-cet being the default and
14764 2018-04-18 Martin Liska <mliska@suse.cz>
14766 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
14769 2018-04-18 Martin Liska <mliska@suse.cz>
14772 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
14775 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
14776 arguments if they are comparable.
14778 2018-04-18 Martin Liska <mliska@suse.cz>
14781 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14784 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14787 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
14790 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
14791 ENDBR after calling __morestack.
14793 2018-04-18 David Malcolm <dmalcolm@redhat.com>
14796 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
14797 by using gcc_base_ver to generate a gcc_driver_version, and use
14798 it when generating GCC_DRIVER_NAME.
14799 * configure: Regenerate.
14801 2018-04-18 Jakub Jelinek <jakub@redhat.com>
14804 * config.gcc: Obsolete powerpc*-*-*spe*.
14806 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14809 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
14810 (stabstr_D): Change type of unum from unsigned int to
14811 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
14814 2018-04-17 Jim Wilson <jimw@sifive.com>
14817 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
14818 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
14819 Set arg_pointer_offset after using pretend_args_size.
14821 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14823 PR rtl-optimization/85431
14824 * dse.c (record_store): Ignore zero width stores.
14827 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
14828 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
14829 __builtin_stack_restore rather than after it.
14830 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
14831 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
14832 argument instead of virtual_dynamic_stack_rtx.
14834 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
14836 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
14838 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14839 Add note to error message to explain internal mapping of overloaded
14840 built-in function name to non-overloaded built-in function name.
14841 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
14844 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
14847 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
14848 where the inputs overlap with the output.
14850 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14852 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
14853 (=v, v) alternative and explicit "memory" attribute.
14854 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
14855 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14857 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
14858 "sselog1" type instead of "sselog".
14859 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
14860 "sselog". Remove explicit "memory" attribute.
14861 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
14862 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14864 (vec_extract_hi_v32hi): Merge all alternatives into one, use
14865 "sselog1" type instead of "sselog". Remove explicit "memory"
14867 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
14868 use "sselog1" type instead of "sselog". Remove explicit "memory"
14870 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
14871 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14873 (vec_extract_hi_v64qi): Merge all alternatives into one, use
14874 "sselog1" type instead of "sselog". Remove explicit "memory"
14876 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
14877 use "sselog1" type instead of "sselog". Remove explicit "memory"
14881 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
14883 PR middle-end/85414
14884 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
14887 2018-04-17 Martin Jambor <mjambor@suse.cz>
14890 * ipa-cp.c (create_specialized_node): Call
14891 expand_all_artificial_thunks if necessary.
14893 2018-04-17 Martin Liska <mliska@suse.cz>
14896 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
14897 in message, remote space in between '_G' and '('.
14899 2018-04-17 Jakub Jelinek <jakub@redhat.com>
14902 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
14903 avx512f_vmcmp<mode>3<round_saeonly_name>,
14904 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
14905 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
14906 avx512f_rndscale<mode><round_saeonly_name>,
14907 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
14908 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
14909 Use %<iptr>2 instead of %2 for -masm=intel.
14910 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
14911 avx512f_vcvttss2usi<round_saeonly_name>,
14912 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
14914 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
14915 avx512f_vcvttsd2usi<round_saeonly_name>,
14916 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
14917 Use %q1 instead of %1 for -masm=intel.
14918 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
14919 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
14920 of %3 for -masm=intel.
14921 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
14923 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
14925 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
14927 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
14928 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
14930 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
14932 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
14933 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
14934 %g1 and one with %0 and %1.
14935 (avx512er_vmrcp28<mode><round_saeonly_name>,
14936 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
14937 %1 for -masm=intel.
14938 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
14939 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
14940 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
14941 of %0 and %{%4%} for -masm=intel.
14942 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
14943 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
14944 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
14945 order of %0 and %{%5%}%{z%} for -masm=intel.
14947 2018-04-17 Jan Hubicka <jh@suse.cz>
14950 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
14952 2018-04-17 Martin Liska <mliska@suse.cz>
14955 * multiple_target.c (create_dispatcher_calls): Set apostrophes
14956 for target_clone error message. Make default implementation
14957 clone to be a local declaration.
14958 (separate_attrs): Add new argument and check for an empty
14960 (expand_target_clones): Handle it.
14961 (ipa_target_clone): Make redirection just for target_clones
14964 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
14965 Tom de Vries <tom@codesourcery.com>
14967 PR middle-end/84955
14968 * omp-expand.c (expand_oacc_for): Add dummy false branch for
14969 tiled basic blocks without omp continue statements.
14971 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
14974 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
14975 vec_extract expression as having side effects to make sure it gets
14978 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
14981 * config/i386/i386.c (get_builtin_code_for_version): Check
14984 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
14987 * config.gcc: Support "skylake".
14988 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14990 * config/i386/i386.c (m_SKYLAKE): Define.
14991 (processor_target_table): Add "skylake".
14992 (ix86_option_override_internal): Add "skylake".
14993 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
14994 PROCESSOR_CANNONLAKE.
14995 (get_builtin_code_for_version): Fix priority for
14996 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
14997 PROCESSOR_SKYLAKE-AVX512.
14998 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
14999 (processor_type): Add PROCESSOR_SKYLAKE.
15001 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
15002 Jason Merrill <jason@redhat.com>
15005 * convert.c (convert_to_integer_1): Use direct recursion for
15006 enumeral types and types with a precision less than the number
15007 of bits in their mode.
15009 2018-04-16 Julia Koval <julia.koval@intel.com>
15012 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
15013 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
15015 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
15018 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
15019 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
15020 and -mno-direct-move.
15022 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
15025 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
15026 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
15027 Ensure negative shifts result in {0}.
15029 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
15031 PR rtl-optimization/79916
15032 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
15033 regs (if any) to define how to gnerate SD moves when LRA is in
15036 2018-04-13 Jakub Jelinek <jakub@redhat.com>
15038 PR rtl-optimization/85393
15039 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
15040 * except.c (expand_dw2_landing_pad_for_region): Make static.
15041 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
15042 a label and unconditional jump to old_bb, rather than
15043 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
15046 PR rtl-optimization/85376
15047 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
15048 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
15049 instead of a specific value.
15051 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15052 Bin Cheng <bin.cheng@arm.com>
15054 PR tree-optimization/82965
15055 PR tree-optimization/83991
15056 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
15057 by_profile_only parameter.
15058 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
15059 information if the loop was predicted to iterate too many times.
15060 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
15062 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
15065 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
15068 2018-04-13 Martin Liska <mliska@suse.cz>
15069 Jakub Jelinek <jakub@redhat.com>
15071 PR middle-end/81657
15072 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
15073 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
15074 * builtins.c (expand_builtin_memory_copy_args): Use
15075 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
15076 handle dest_addr == pc_rtx.
15078 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
15081 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
15082 asked to not generate direct moves.
15083 (fix_trunc<mode>si2_stfiwx): Similar.
15084 (fix_trunc<mode>si2_internal): Similar.
15086 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15089 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
15090 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
15091 lookup if dest in some wider mode is known to be const0_rtx and
15092 if so, record permanent equivalence for it to be ZERO_EXTEND of
15093 the narrower mode destination.
15095 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15097 * lto-streamer-out.c (output_function): Revert 259346.
15098 * omp-expand.c (expand_oacc_for): Likewise.
15100 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
15102 PR rtl-optimization/85354
15103 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
15104 * sel-sched.c (sel_global_init): ... here.
15106 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
15109 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
15110 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
15111 mode for PE-COFF targets.
15112 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
15113 (i386_pe_asm_lto_end): Likewise.
15114 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
15115 (TARGET_ASM_LTO_END): Likewise.
15116 * config/i386/winnt.c (saved_debug_info_level): New static variable.
15117 (i386_pe_asm_lto_start): New function.
15118 (i386_pe_asm_lto_end): Likewise.
15120 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
15121 Richard Biener <rguenther@suse.de>
15123 PR middle-end/84955
15124 * lto-streamer-out.c (output_function): Fix CFG loop state before
15126 * omp-expand.c (expand_oacc_for): Handle calls to internal
15127 functions like regular functions.
15129 2018-04-12 Richard Biener <rguenther@suse.de>
15132 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
15133 for the early LTO debug to properly generate references to it
15134 during DIE emission. Do not re-use that for the skeleton for
15136 (dwarf2out_early_finish): Likewise.
15138 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15141 * config/i386/sse.md
15142 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
15143 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
15144 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
15145 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
15146 and output is a reg, avoid creating invalid lowpart subreg, but
15147 instead split into a 512-bit move. Don't split if not AVX512VL,
15148 input is xmm16+ reg and output is a mem.
15149 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
15150 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
15151 xmm16+ reg and output is a mem.
15153 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15155 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
15156 also for flag_dwarf2_cfi_asm.
15158 2018-04-12 Jakub Jelinek <jakub@redhat.com>
15160 PR rtl-optimization/85342
15161 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
15162 a bool scalar var inside of the loop instead. Don't try to update
15163 recog_data.operand after failed apply_change_group.
15165 2018-04-12 Tom de Vries <tom@codesourcery.com>
15168 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
15169 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
15170 array with flexible array member as array without given dimension.
15171 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
15172 argument for undefined param to true.
15174 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
15177 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15178 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
15179 from PowerPC section.
15180 * config/rs6000/sysv4.opt (mcall-): Improve help text.
15181 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
15182 help text that is too long.
15183 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
15184 help text that is too long.
15185 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
15186 help text that is too long.
15188 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
15190 * config/alpha/alpha.md (stack_probe_internal): Rename
15191 from "probe_stack". Update all callers.
15193 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15195 PR rtl-optimization/84566
15196 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
15197 sched_macro_fuse_insns.
15199 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15202 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
15203 (compute_block_dependences): ... from here.
15205 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15207 PR tree-optimization/85331
15208 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
15209 from int to HOST_WIDE_INT.
15211 2018-04-11 Martin Jambor <mjambor@suse.cz>
15214 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
15215 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
15216 not the same as the source val.
15217 (cgraph_edge_brings_value_p): New parameter.
15218 (gather_edges_for_value): Pass destination value to
15219 cgraph_edge_brings_value_p.
15220 (perhaps_add_new_callers): Likewise.
15221 (get_info_about_necessary_edges): Likewise and exclude values brought
15222 only by self-recursive edges.
15223 (create_specialized_node): Redirect only clones of self-calling edges.
15224 (+self_recursive_pass_through_p): New function.
15225 (find_more_scalar_values_for_callers_subset): Use it.
15226 (find_aggregate_values_for_callers_subset): Likewise.
15227 (known_aggs_to_agg_replacement_list): Removed.
15228 (decide_whether_version_node): Re-calculate known constants for all
15229 remaining context clones.
15231 2018-04-11 Richard Biener <rguenther@suse.de>
15234 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
15235 from early DWARF output.
15236 (dwarf2out_early_finish): Output line info unconditionally into
15237 early DWARF and add reference to it.
15239 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15242 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
15243 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
15244 other than V2DFmode using iptr mode attribute.
15245 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
15247 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
15249 PR rtl-optimization/84659
15250 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
15252 2018-04-11 Jakub Jelinek <jakub@redhat.com>
15255 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
15257 (output_loc_list): Pass address of a dummy size variable even in the
15258 locview handling loop.
15259 (index_location_lists): Add comment on why skip_loc_list_entry can't
15262 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
15265 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
15268 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
15271 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15272 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
15273 and -mstring-compare-inline-limit.
15275 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15278 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
15279 for stack clash protection in a register whenever we need it to be in
15282 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15284 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
15285 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
15287 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
15290 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
15292 (mlong-double-): Ditto.
15293 * config/rs6000/sysv4.opt (msdata=): Ditto.
15294 (mtls-size=): Ditto.
15296 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15298 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15299 erroneous entries for
15300 "vector int vec_ldl (int, long int *)", and
15301 "vector unsigned int vec_ldl (int, unsigned long int *)".
15302 Add comments and entries for
15303 "vector bool char vec_ldl (int, bool char *)",
15304 "vector bool short vec_ldl (int, bool short *)",
15305 "vector bool int vec_ldl (int, bool int *)",
15306 "vector bool long long vec_ldl (int, bool long long *)",
15307 "vector pixel vec_ldl (int, pixel *)",
15308 "vector long long vec_ldl (int, long long *)",
15309 "vector unsigned long long vec_ldl (int, unsigned long long *)".
15310 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
15311 type tree bool_long_long_type_node and correct definition of
15312 bool_V2DI_type_node to make reference to this new type tree.
15313 (rs6000_mangle_type): Replace erroneous reference to
15314 bool_long_type_node with bool_long_long_type_node.
15315 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
15316 comments to emphasize sign distinctions for char and int types and
15317 replace RS6000_BTI_bool_long constant with
15318 RS6000_BTI_bool_long_long constant. Also add comment to restrict
15319 use of RS6000_BTI_pixel.
15320 (bool_long_type_node): Remove this macro definition.
15321 (bool_long_long_type_node): New macro definition
15323 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15325 PR rtl-optimization/85300
15326 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
15327 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
15328 simplify_unary_operation fails.
15330 2018-04-10 Martin Liska <mliska@suse.cz>
15332 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
15333 cgraph_edge and ipa_ref.
15335 2018-04-10 Jakub Jelinek <jakub@redhat.com>
15339 * config/i386/sse.md
15340 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
15341 computation of the VEC_MERGE selector from mask.
15342 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
15343 Fix decoding of the VEC_MERGE selector into mask.
15345 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15347 PR tree-optimization/85286
15348 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
15350 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
15352 * final.c (final_1): Set insn_last_address as well as
15353 insn_current_address.
15355 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15358 * explow.c (emit_stack_probe): Call validize_mem on memory location
15359 before passing it to gen_probe_stack. Create address operand and
15360 legitimize it for the probe_stack_address case.
15362 2018-04-09 Jan Hubicka <jh@suse.cz>
15365 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
15366 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
15367 * tree.c (free_lang_data_in_type): Fix handling of binfos;
15369 (free_lang_data): Rebuild type inheritance graph.
15371 2018-04-09 Martin Sebor <msebor@redhat.com>
15373 * invoke.texi (-finline-small-functions): Mention other optimization
15375 (-findirect-inlining, -fpartial-inlining): Same.
15376 (-finline-functions-called-once): Same.
15377 (-freorder-blocks-and-partition): Same.
15379 2018-04-09 Jan Hubicka <jh@suse.cz>
15382 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
15383 jumps; choose last target that matches the criteria (i.e.
15384 no partition changes for non-crossing jumps).
15385 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
15386 support for redirecting crossing jumps to non-crossing.
15388 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
15390 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
15391 also for naked functions.
15393 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
15395 * config/arc/arc.md (add_shift): New pattern.
15396 (add_shift2): Likewise.
15397 (sub_shift): Likewise.
15398 (sub_shift_cmp0_noout): Likewise.
15399 (compare_si_ashiftsi): Likewise.
15400 (xbfu_cmp0_noout): New combine pattern.
15401 (xbfu_cmp0"): Likewise.
15402 (movsi_set_cc_insn): Place the predicable variant first.
15403 (commutative_binary_cmp0_noout): Remove clobber.
15404 (commutative_binary_cmp0): New pattern.
15405 (noncommutative_binary_cmp0): Likewise.
15406 (noncommutative_binary_cmp0_noout): Likewise.
15407 (noncommutative_binary_comparison_result_used): Removed.
15408 (rsub_cmp0): New pattern.
15409 (rsub_cmp0_noout): Likewise.
15410 (extzvsi): Changed, keep only meaningful variants.
15411 (SQH, SEZ): New iterators.
15412 (SQH_postfix): New mode attribute.
15413 (SEZ_prefix): New code attribute.
15414 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
15415 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
15416 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
15417 of numerical value.
15418 (noncommutative_operator): Check the availability of barrel
15421 2018-04-09 Richard Biener <rguenther@suse.de>
15423 PR tree-optimization/85284
15424 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
15425 Only use the niter constraining form of simple_iv when the exit
15426 is always executed.
15428 2018-04-09 Tom de Vries <tom@codesourcery.com>
15431 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15432 (define_expand "*memory_barrier"): New define_expand.
15433 (define_insn "memory_barrier"): New insn.
15435 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15437 PR rtl-optimization/80463
15438 PR rtl-optimization/83972
15439 PR rtl-optimization/83480
15441 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
15442 correct producer for the insn.
15443 (tidy_control_flow): Fixup seqnos in case of debug insns.
15445 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15447 PR rtl-optimization/83913
15449 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
15450 different sched-times when merging exprs.
15452 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15454 PR rtl-optimization/83962
15456 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
15457 tidy_fallthru_edge and tidy_control_flow.
15459 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
15461 PR rtl-optimization/83530
15463 * sel-sched.c (force_next_insn): New global variable.
15464 (remove_insn_for_debug): When force_next_insn is true, also leave only
15465 next insn in the ready list.
15466 (sel_sched_region): When the region wasn't scheduled, make another pass
15467 over it with force_next_insn set to 1.
15469 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15471 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
15473 * config/nds32/constants.md (unspec_volatile_element): Add enum values
15474 for interrupt control.
15475 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15476 functions for interrupt control.
15477 * config/nds32/nds32-intrinsic.md: Likewise.
15478 * config/nds32/nds32_intrinsic.h: Likewise.
15479 * config/nds32/nds32.h (nds32_builtins): Likewise.
15481 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15483 * config/nds32/nds32.c (nds32_init_machine_status,
15484 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
15485 strict_aligned_p field.
15486 (nds32_expand_to_rtl_hook): New function.
15487 (TARGET_EXPAND_TO_RTL_HOOK): Define.
15488 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
15490 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15491 Chung-Ju Wu <jasonwucj@gmail.com>
15493 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
15494 * config/nds32/nds32-n7.md: New file.
15495 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
15496 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
15498 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
15499 * config/nds32/nds32.md (pipeline_model): Add n7.
15500 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
15501 * config/nds32/pipelines.md: Include n7 settings.
15503 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15504 Chung-Ju Wu <jasonwucj@gmail.com>
15506 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
15507 * config/nds32/nds32-e8.md: New file.
15508 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
15509 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
15511 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
15512 * config/nds32/nds32.md (pipeline_model): Add e8.
15513 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
15514 * config/nds32/pipelines.md: Include e8 settings.
15516 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15517 Chung-Ju Wu <jasonwucj@gmail.com>
15519 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
15520 * config/nds32/nds32-n8.md: New file.
15521 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
15522 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
15524 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
15525 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
15526 * config/nds32/nds32.md (pipeline_model): Add n8.
15527 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
15528 * config/nds32/pipelines.md: Include n8 settings.
15530 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
15531 Chung-Ju Wu <jasonwucj@gmail.com>
15533 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
15534 * config/nds32/nds32-n9-2r1w.md: New file.
15535 * config/nds32/nds32-n9-3r2w.md: New file.
15536 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
15537 nds32_register_ports): New or modify for cpu n9.
15538 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
15540 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
15541 * config/nds32/nds32-utils.c: New file.
15542 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
15543 TARGET_MUL_SLOW): Define.
15544 * config/nds32/nds32.md (pipeline_model): New attribute.
15545 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
15546 New options that support cpu n9.
15547 * config/nds32/pipelines.md: Include n9 settings.
15548 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
15550 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
15552 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
15553 information if necessary.
15554 (output_cond_branch_compare_zero): Likewise.
15555 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
15556 (nds32_target_alignment): Refine for alignment.
15557 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
15558 (FUNCTION_BOUNDARY): Modify.
15559 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
15561 * config/nds32/nds32.opt (malways-align, malign-functions): New.
15563 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
15565 * config/nds32/constants.md (unspec_volatile_element): Add values for
15566 TLB operation and data prefetch.
15567 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15568 functions for TLB operation and data prefetch.
15569 * config/nds32/nds32-intrinsic.md: Likewise.
15570 * config/nds32/nds32_intrinsic.h: Likewise.
15571 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
15572 (nds32_print_operand): Likewise.
15573 * config/nds32/nds32.h (nds32_builtins): Likewise.
15575 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
15576 Andrew Pinski <pinsika@gcc.gnu.org>
15578 PR middle-end/82976
15579 * match.pd: Use constant_boolean_node of correct type instead of
15580 boolean_true_node or boolean_false_node for simplifying
15581 pointer comparisons to zero.
15583 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15585 PR tree-optimization/80021
15586 * tree.c (verify_type_variant): Make error call in verify_variant_match
15587 translatable and remove final full stop.
15589 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15591 * config/nds32/constants.md (unspec_volatile_element): Add
15592 UNSPEC_VOLATILE_EH_RETURN.
15593 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
15594 nds32_output_stack_pop): Support dwarf exception handling process.
15595 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
15596 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
15597 exception handling process.
15598 (nds32_compute_stack_frame): Likewise.
15599 (nds32_return_addr_rtx): Likewise.
15600 (nds32_initial_elimination_offset): Likewise.
15601 (nds32_expand_prologue): Likewise.
15602 (nds32_expand_epilogue): Likewise.
15603 (nds32_dynamic_chain_address): New function.
15604 * config/nds32/nds32.h (machine_function): Add fields for dwarf
15605 exception handling.
15606 (DYNAMIC_CHAIN_ADDRESS): Define.
15607 (EH_RETURN_DATA_REGNO): Define.
15608 (EH_RETURN_STACKADJ_RTX): Define.
15609 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
15610 patterns for dwarf exception handling.
15612 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15614 * config/nds32/nds32.h: Clean up obsolete macros.
15616 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15618 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15619 Add enum values for particular instructions.
15620 * config/nds32/nds32-intrinsic.c: Implementation of expanding
15621 particular intrinsic functions.
15622 * config/nds32/nds32-intrinsic.md: Likewise.
15623 * config/nds32/nds32_intrinsic.h: Likewise.
15624 * config/nds32/nds32.h (nds32_builtins): Likewise.
15625 * config/nds32/nds32.md (type): Add pbsad and pbsada.
15626 (btst, ave): New patterns for particular instructions.
15628 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15630 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15631 Add enum values for atomic load/store and memory sync.
15632 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
15634 * config/nds32/nds32-intrinsic.md: Likewise.
15635 * config/nds32/nds32_intrinsic.h: Likewise.
15636 * config/nds32/nds32.h (nds32_builtins): Likewise.
15638 2018-04-07 Jakub Jelinek <jakub@redhat.com>
15640 PR tree-optimization/85257
15641 * fold-const.c (native_encode_vector): If not all elts could fit
15642 and off is -1, return 0 rather than offset.
15643 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
15644 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
15645 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
15646 adjust buffer in native_interpret_expr call.
15648 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15650 * config/nds32/constants.md (unspec_volatile_element): Add cache
15651 control enum values.
15652 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
15653 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
15654 * config/nds32/nds32.c (nds32_cctl_names): New.
15655 (nds32_print_operand): Handle cache control register names.
15656 * config/nds32/nds32.h (nds32_builtins): New enum values.
15657 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
15659 * config/nds32/nds32.md (type): Add mmu.
15660 * config/nds32/pipelines.md (simple_insn): Add mmu.
15662 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15664 * config/nds32/nds32.md (type): Remove call.
15665 * config/nds32/pipelines.md (simple_insn): Likewise.
15667 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15669 * config/nds32/constants.md (unspec_volatile_element): Add
15670 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
15671 UNSPEC_VOLATILE_FMFCFG.
15672 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
15673 description for fmfcfg and fmfcsr.
15674 (bdesc_1arg): Add fmtcsr.
15675 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
15676 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
15677 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
15678 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
15679 unspec_fmfcfg): New patterns.
15680 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
15681 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
15682 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
15683 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
15684 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
15685 __nds32__fmfcfg): Define.
15687 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
15689 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
15690 intrinsic register names.
15691 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
15692 intrinsic register enum values and macros.
15694 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
15696 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
15697 for load/store addressing form.
15698 (nds32_print_operand_address): Likewise.
15700 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
15703 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
15704 based on LABEL_REF. Remove useless assertion.
15705 (pic_address_needs_scratch): Fix formatting.
15706 (sparc_legitimize_pic_address): Minor tweaks.
15707 (sparc_delegitimize_address): Adjust assertion accordingly.
15708 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
15709 into symbolic_operand.
15710 (movsi_high_pic_label_ref): Likewise.
15711 (movsi_lo_sum_pic_label_ref): Likewise.
15712 (movdi_pic_label_ref): Likewise.
15713 (movdi_high_pic_label_ref): Likewise.
15714 (movdi_lo_sum_pic_label_ref): Likewise.
15716 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
15718 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
15719 custom LIB_SPEC setup.
15721 2018-04-06 Ruslan Bukin <br@bsdpad.com>
15722 Kito Cheng <kito.cheng@gmail.com>
15724 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
15725 * config/riscv/freebsd.h: New.
15727 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15729 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
15730 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
15733 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15734 Kito Cheng <kito.cheng@gmail.com>
15736 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
15737 nds32_output_call, nds32_symbol_binds_local_p): New functions.
15738 * config/nds32/nds32-protos.h (nds32_output_call,
15739 nds32_output_return): Declare.
15740 * config/nds32/nds32.md: Refine all the call and return patterns.
15742 2018-04-06 Jakub Jelinek <jakub@redhat.com>
15745 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
15746 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
15748 PR rtl-optimization/84872
15749 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
15750 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
15751 EDGE_CROSSING edge.
15753 2018-04-06 Tamar Christina <tamar.christina@arm.com>
15755 * expr.c (copy_blkmode_to_reg): Revert 254862.
15756 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
15758 2018-04-06 Richard Biener <rguenther@suse.de>
15760 PR middle-end/85244
15761 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
15762 after seeing a component reference with an adjacent field. Treat
15763 refs to arrays at struct end of external decls similar to
15764 refs to unconstrained commons.
15766 2018-04-06 Jakub Jelinek <jakub@redhat.com>
15769 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
15770 look through SAVE_EXPRs with non-side-effects argument. Adjust
15772 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
15775 2018-04-06 Richard Biener <rguenther@suse.de>
15777 PR middle-end/85180
15778 * alias.c (find_base_term): New wrapper around find_base_term
15779 unwinding CSELIB_VAL_PTR changes.
15780 (find_base_term): Do not restore CSELIB_VAL_PTR during the
15783 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15785 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
15787 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
15788 constant definitions.
15789 ("nop"): lr 0,0 -> nopr r0
15790 ("nop_lr0", "nop_lr1"): New insn definitions.
15792 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
15794 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
15795 NDS32_V3PUSH_AVAILABLE_P macro.
15797 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
15798 Chung-Ju Wu <jasonwucj@gmail.com>
15800 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
15801 (nds32*-*-*): Add float and fpu_config into supported_defaults.
15802 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
15803 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
15804 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
15805 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
15806 * config/nds32/constraints.md: New constraints and checking for hard
15807 float configuration.
15808 * config/nds32/iterators.md: New mode iterator and attribute for hard
15809 float configuration.
15810 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
15812 * config/nds32/nds32-fpu.md: New file.
15813 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
15814 deal with hard float code generation.
15815 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
15817 (abi_type, float_reg_number): New enum type.
15818 * config/nds32/nds32-predicates.c: New predicates for hard float.
15819 * config/nds32/nds32-protos.h: Declare functions for hard float.
15820 * config/nds32/nds32.c: Implementation for hard float configuration.
15821 * config/nds32/nds32.h: Definitions for hard float configuration.
15822 * config/nds32/nds32.md: Include hard float machine description and
15823 modify patterns for hard float configuration.
15824 * config/nds32/nds32.opt: New options for hard float configuration.
15825 * config/nds32/predicates.md: New predicates for hard float
15828 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
15830 * common/config/nds32/nds32-common.c
15831 (nds32_option_optimization_table): Enable -mreleax-hint by default.
15833 2018-04-05 Jakub Jelinek <jakub@redhat.com>
15835 PR middle-end/85195
15836 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
15837 CONSTRUCTOR_ELT (ctor, ...)->value.
15839 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
15842 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
15844 2018-04-05 Tom de Vries <tom@codesourcery.com>
15847 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
15850 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15851 Kito Cheng <kito.cheng@gmail.com>
15853 * config/nds32/constraints.md (U33): Fine-tune checking condition.
15854 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
15855 * config/nds32/nds32.h (nds32_16bit_address_type): Add
15856 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
15858 2018-04-05 Shiva Chen <shiva0217@gmail.com>
15859 Kito Cheng <kito.cheng@gmail.com>
15861 * config/nds32/constraints.md (Ufe): New memory constraint.
15862 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
15863 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
15864 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
15866 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
15867 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
15869 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15871 * config/nds32/nds32.md: Use optimize_size in the condition for
15872 alu-shift instructions.
15874 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15876 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
15878 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15880 * config/nds32/nds32.md (negsi2): Refine pattern.
15882 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15883 Chung-Ju Wu <jasonwucj@gmail.com>
15885 * config/nds32/iterators.md (shift_rotate): New code iterator.
15886 (shift): New code attribute.
15887 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
15888 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
15889 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
15890 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
15891 bit-wise operations.
15892 (andsi3, *andsi3): Ditto.
15893 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
15894 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
15895 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
15896 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
15897 nds32_ior_operand, nds32_xor_operand): New predicates.
15899 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15901 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
15902 (addsi3, subsi3): ... this.
15904 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15906 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
15908 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15910 * config/nds32/nds32.md: Adjust indention.
15912 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
15914 * config/nds32/nds32.md (feature): New attribute.
15916 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
15918 * config/nds32/nds32.md (subtype): New attribute.
15920 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
15923 * config/arm/arm-builtins.c (arm_expand_builtin): Change
15924 expansion to perform a bitwise AND of the argument followed by a
15925 boolean negation of the result.
15927 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
15929 PR rtl-optimization/84878
15930 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
15931 the basic block. Assert the use reference is not artificial and that
15932 it has an associated insn.
15934 2018-04-04 Michael Matz <matz@suse.de>
15936 * builtins.c (compute_objsize): Pass correct operand
15937 to array_at_struct_end_p.
15939 2018-04-04 Richard Biener <rguenther@suse.de>
15942 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
15943 from contexts for DINFO_LEVEL_TERSE and below.
15945 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15947 * config/nds32/nds32-doubleword.md (move_<mode>): Require
15948 resiter_operand condition.
15949 * config/nds32/nds32.md (*move<mode>): Ditto.
15951 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15952 Monk Chiang <sh.chiang04@gmail.com>
15954 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
15956 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15958 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
15960 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15961 Kito Cheng <kito.cheng@gmail.com>
15963 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
15964 nds32_cond_code_str, output_cond_branch,
15965 output_cond_branch_compare_zero, nds32_expand_cbranch,
15966 nds32_expand_cstore, nds32_expand_movcc,
15967 nds32_output_cbranchsi4_equality_zero,
15968 nds32_output_cbranchsi4_equality_reg,
15969 nds32_output_cbranchsi4_equality_reg_or_const_int,
15970 nds32_output_cbranchsi4_greater_less_zero: New functions.
15971 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
15972 nds32_expand_cstore, nds32_expand_movcc,
15973 nds32_output_cbranchsi4_equality_zero,
15974 nds32_output_cbranchsi4_equality_reg,
15975 nds32_output_cbranchsi4_equality_reg_or_const_int,
15976 nds32_output_cbranchsi4_greater_less_zero): Declare.
15977 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
15978 nds32_rimm11s_operand): New predicates.
15979 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
15980 * config/nds32/nds32.md: Rewrite all the branch and conditional move
15983 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
15985 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
15986 * config/nds32/nds32.md: Ditto.
15987 * config/nds32/pipelines.md: Ditto.
15989 2018-04-04 Richard Biener <rguenther@suse.de>
15991 PR tree-optimization/85168
15992 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
15993 propagating abnormals.
15995 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
15997 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
15999 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
16000 Kito Cheng <kito.cheng@gmail.com>
16002 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
16003 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
16004 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
16005 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
16006 * config/nds32/nds32.md (sibcall_internal): New.
16007 (sibcall_register): Remove.
16008 (sibcall_immediate): Remove.
16009 (sibcall_value_internal): New.
16010 (sibcall_value_register): Remove.
16011 (sibcall_value_immediate): Remove.
16012 * config/nds32/predicates.md (nds32_general_register_operand): New.
16013 (nds32_call_address_operand): New.
16015 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16017 PR rtl-optimization/85167
16018 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
16019 bb_defs if *split_p, instead preinitialize it to NULL.
16021 PR tree-optimization/85156
16022 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
16023 evaluating the argument multiple times.
16025 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
16027 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
16029 (_mm_cvtpd_ps): Likewise.
16030 (_mm_cvttpd_epi32): Likewise.
16031 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
16032 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
16033 vector, pixel, and bool following altivec.h include.
16035 2018-04-03 Martin Sebor <msebor@redhat.com>
16037 * doc/extend.texi (Common Function Attributes): Clarify.
16038 (const attribute): Likewise.
16039 (pure attribute): Likewise.
16041 2018-04-03 Jakub Jelinek <jakub@redhat.com>
16044 * config/i386/i386.c (ix86_expand_vector_set): Use
16045 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
16047 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
16049 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
16050 instructions when changing rounding bits to preserve precision bits
16051 in the x87 control word.
16053 2018-04-03 Martin Liska <mliska@suse.cz>
16055 PR tree-optimization/82491
16056 * rtl.h (strip_offset_and_add): Replace += suboffset with
16057 poly_uint64 () + suboffset.
16059 2018-03-29 Martin Liska <mliska@suse.cz>
16060 Martin Jambor <mjambor@suse.cz>
16063 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
16064 param_type is not an integral or pointer type.
16066 2018-04-03 Richard Biener <rguenther@suse.de>
16068 * sese.h (recompute_all_dominators): Remove.
16070 2018-04-02 Martin Sebor <msebor@redhat.com>
16072 * doc/invoke.texi (-Wrestrict): Fix typos.
16074 2018-04-02 Jim Wilson <jimw@sifive.com>
16076 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
16077 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
16078 (<optab>di3, <optab>si3_extend): Likewise.
16079 (<optab>si3_mask, <optab>si3_mask_1): New.
16080 (<optab>di3_mask, <optab>di3_mask_1): New.
16081 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
16082 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
16083 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
16085 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
16087 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
16090 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
16092 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
16093 (nds32_canonicalize_comparison): New function.
16095 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16096 Kito Cheng <kito.cheng@gmail.com>
16097 Kuan-Lin Chen <kuanlinchentw@gmail.com>
16099 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
16100 * config/nds32/constants.md (unspec_volatile_element): Add
16101 UNSPEC_VOLATILE_RELAX_GROUP.
16102 * config/nds32/nds32-relax-opt.c: New file.
16103 * config/nds32/nds32-predicates.c
16104 (nds32_symbol_load_store_p): New function.
16105 * config/nds32/nds32-protos.h
16106 (nds32_symbol_load_store_p): Declare function.
16107 (make_pass_nds32_relax_opt): Declare new rtl pass function.
16108 * config/nds32/nds32.c
16109 (nds32_register_pass): New function to register pass.
16110 (nds32_register_passes): New function to register passes.
16111 * config/nds32/nds32.md (relax_group): New pattern.
16112 * config/nds32/nds32.opt (mrelax-hint): New option.
16113 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
16115 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
16117 * config/nds32/t-nds32: Modify files dependency.
16119 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16121 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
16122 (PROFILE_HOOK): Define its implementation.
16124 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
16126 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
16127 type and 32-bit size.
16129 2018-04-01 Jakub Jelinek <jakub@redhat.com>
16131 PR middle-end/85090
16132 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
16133 (V_128_256): New mode iterator.
16134 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
16135 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
16136 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
16138 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
16139 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
16141 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
16144 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
16145 NaN inputs correctly.
16147 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
16150 * config/rs6000/vsx.md (??r): New mode attribute.
16151 (*vsx_mov<mode>_64bit): Use it.
16152 (*vsx_mov<mode>_32bit): Likewise.
16154 2018-03-30 Martin Sebor <msebor@redhat.com>
16156 PR tree-optimization/84818
16157 * builtins.c (check_access): Use warning_n.
16159 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16162 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
16164 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
16167 2018-03-30 Julia Koval <julia.koval@intel.com>
16170 * x86-tune.def (movx, partial_reg_dependency): Enable for
16173 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
16175 PR inline-asm/84985
16176 * lra-constraints.c (process_alt_operands): Move setting
16177 this_alternative_matches below.
16179 2018-03-29 Martin Liska <mliska@suse.cz>
16182 * doc/invoke.texi: Document how LTO works with debug info.
16183 Describe auto-load support of binutils. Mention 'x86-64'
16184 as valid option value of -march option.
16186 2018-03-29 Jakub Jelinek <jakub@redhat.com>
16188 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
16191 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
16192 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
16193 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
16196 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16199 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
16200 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
16201 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
16202 for RS6000_BTM_POWERPC64.
16203 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
16204 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
16205 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
16208 (DIVDEU): Likewise.
16210 2018-03-28 Carl Love <cel@us.ibm.com>
16213 2017-09-27 Carl Love <cel@us.ibm.com>
16215 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
16216 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
16217 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
16220 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16222 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
16223 instead of __vector bool.
16224 (_mm_max_pu8): Likewise.
16225 (_mm_min_pi16): Likewise.
16227 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
16230 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
16231 (DIVWEUO): Likewise.
16232 (DIVDEO): Likewise.
16233 (DIVDEUO): Likewise.
16234 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
16235 DIVWEUO and DIVDEUO.
16236 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
16237 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
16238 (div_extend): Likewise.
16239 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
16241 (__builtin_divweuo): Likewise.
16242 (__builtin_divdeo): Likewise.
16243 (__builtin_divdeuo): Likewise.
16245 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16248 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
16249 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
16251 PR tree-optimization/82004
16252 * gimple-match-head.c (optimize_pow_to_exp): New function.
16253 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
16254 Don't fold to exp if optimize_pow_to_exp is false.
16256 2018-03-28 Martin Liska <mliska@suse.cz>
16259 * calls.c (initialize_argument_information): Fix trailing space.
16260 * common.opt: Fix typo and provide better explanation for
16261 -fsanitize-coverage option.
16262 * config/i386/i386.opt: Fix typo.
16264 2018-03-28 Jakub Jelinek <jakub@redhat.com>
16265 Martin Liska <mliska@suse.cz>
16268 * gimplify.c (asan_poison_variable): Don't do the check for
16269 gimplify_omp_ctxp here.
16270 (gimplify_decl_expr): Do it here.
16271 (gimplify_target_expr): Likewise.
16273 2018-03-28 Martin Liska <mliska@suse.cz>
16276 * config/i386/i386.c (ix86_function_arg_advance): Do not call
16277 chkp_type_bounds_count if MPX is not enabled.
16279 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
16281 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
16283 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
16286 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
16287 function to create the function decl for complex long double
16288 multiply and divide for -mabi=ieeelongdouble.
16289 (init_float128_ieee): Call it.
16291 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
16294 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
16295 -fcf-protection=branch -mibt.
16296 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
16298 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16301 * config/arm/arm.c (arm_valid_symbolic_address): Handle
16302 arm_word_relocations.
16304 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
16307 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
16308 extern array declarations.
16310 2018-03-27 Richard Biener <rguenther@suse.de>
16312 PR middle-end/84067
16313 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
16314 explicit single_use checks.
16316 2018-03-27 Richard Biener <rguenther@suse.de>
16318 PR tree-optimization/85082
16319 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
16322 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16324 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
16325 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
16326 Turn on fasynchronous-unwind-tables and funwind-tables.
16328 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
16331 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
16332 (*bmi_blsr_<mode>_ccz): Ditto.
16334 2018-03-26 Tom de Vries <tom@codesourcery.com>
16336 PR tree-optimization/85063
16337 * omp-general.c (offloading_function_p): New function. Factor out
16339 * omp-offload.c (pass_omp_target_link::gate): ... here.
16340 * omp-general.h (offloading_function_p): Declare.
16341 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
16342 with attribute omp declare target for offloading functions.
16344 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
16346 PR tree-optimization/84005
16347 * tree-data-ref.h (get_base_for_alignment): Declare.
16348 * tree-data-ref.c (get_base_for_alignment_1): New function.
16349 (get_base_for_alignment): Likewise.
16350 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
16351 get_base_for_alignment to find a suitable base object, instead
16352 of always using drb->base_address.
16354 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16356 PR inline-asm/85022
16357 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
16358 known size by default.
16360 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
16362 PR inline-asm/85030
16363 * lra-constraints.c (process_alt_operands): Don't match BLKmode
16364 and non BLKmode operands.
16366 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16369 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
16370 Clean up attributes.
16372 2018-03-23 Richard Biener <rguenther@suse.de>
16375 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
16376 we are going to emit early debug for LTO.
16378 2018-03-23 Jakub Jelinek <jakub@redhat.com>
16380 PR inline-asm/85034
16381 * function.c (match_asm_constraints_1): Don't optimize if input
16382 doesn't satisfy general_operand predicate for output's mode.
16384 PR inline-asm/85022
16385 * alias.c (write_dependence_p): Don't require for x_canonicalized
16386 non-VOIDmode if x has VOIDmode.
16389 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
16390 just don't try to optimize it rather than assert it never happens.
16392 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16394 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
16395 macro expansions for definition of ST_INTERNAL_<mode> and
16396 LD_INTERNAL_<mode> builtins.
16397 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
16399 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
16401 (altivec_expand_st_builtin): Likewise.
16402 (altivec_expand_builtin): Remove calls to deleted functions.
16403 (rs6000_address_for_altivec): Delete this function.
16404 * config/rs6000/vector.md: Remove expands for
16405 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
16407 2018-03-22 Sudakshina Das <sudi.das@arm.com>
16410 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
16411 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
16412 re-computing once computed.
16413 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
16414 (arm_init_machine_status): Initialize
16415 machine->static_chain_stack_bytes.
16417 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
16420 * doc/extend.texi: Add four new prototypes for vec_ld.
16421 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
16422 definitions for more logical presentation.
16423 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
16424 entries for V1TI variants of __builtin_altivec_ld builtin.
16425 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
16426 handling of V1TI variant of LVX icode pattern.
16427 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
16428 (rs6000_gimple_fold_builtin): Likewise.
16429 (altivec_init_builtins): Add code to define
16430 __builtin_altivec_lvx_v1ti function.
16432 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16434 PR inline-asm/84941
16435 * function.c (match_asm_constraints_1): Don't do the optimization
16436 if input isn't a REG, SUBREG, MEM or constant.
16438 2018-03-22 Tom de Vries <tom@codesourcery.com>
16440 PR tree-optimization/84956
16441 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
16442 bb_has_abnormal_pred.
16444 2018-03-22 Jakub Jelinek <jakub@redhat.com>
16447 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
16448 DECL_INITIAL (decl) to decl at the end.
16449 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
16450 adjust the comment.
16452 2018-03-21 Joseph Myers <joseph@codesourcery.com>
16454 * doc/extend.texi (__builtin_tgmath): Document when complex
16455 integer types are treated as _Complex _Float64.
16457 2018-03-21 Tom de Vries <tom@codesourcery.com>
16459 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
16461 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16463 PR tree-optimization/84960
16464 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
16465 if it is ENTRY block, move them into single succ of ENTRY in that case.
16467 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
16469 PR tree-optimization/84811
16470 * poly-int.h (poly_span_traits): Remove the T3 parameter and
16471 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
16472 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
16473 (known_subrange_p): Update accordingly. Cast each value involved
16474 in the size comparison, rather than casting the result of the
16477 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16479 PR tree-optimization/84982
16480 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
16481 by flipping the least significant bit rather than all bits from
16482 bitpos to bitpos + bitsize - 1.
16484 2018-03-21 Nathan Sidwell <nathan@acm.org>
16486 * doc/extend.texi (Deprecated Features): Remove mention of
16487 long-deleted deprecations.
16489 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16492 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
16493 * configure: Regenerate.
16495 2018-03-21 Tom de Vries <tom@codesourcery.com>
16497 PR tree-optimization/83126
16498 * tree-parloops.c (num_phis): New function.
16499 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
16501 2018-03-21 Nathan Sidwell <nathan@acm.org>
16503 * doc/extend.texi (Deprecated Features): Update deprecated flags,
16504 mention anon-struct/union members and trailing attributes.
16506 2018-03-21 Bin Cheng <bin.cheng@arm.com>
16508 PR tree-optimization/84969
16509 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
16510 builtin memset partitions if they set different rhs values.
16512 2018-03-21 Jakub Jelinek <jakub@redhat.com>
16514 PR rtl-optimization/84989
16515 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
16516 VEC_DUPLICATE with scalar result mode.
16518 2018-03-21 Martin Liska <mliska@suse.cz>
16521 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
16522 not intended return statement.
16524 2018-03-21 Martin Liska <mliska@suse.cz>
16527 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
16528 (chkp_find_bound_slots_1): Limit number of iterations.
16530 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
16533 * Minor grammar fixes for x86 options.
16535 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16538 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
16539 holding REG_CFA_RESTORE notes, instead turn them into a USE.
16541 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
16544 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
16545 (altivec_lvx_<mode>_1op): Likewise.
16546 (altivec_stvx_<mode>_2op): Likewise.
16547 (altivec_stvx_<mode>_1op): Likewise.
16548 (altivec_lvx_<VM2:mode>): New define_expand.
16549 (altivec_stvx_<VM2:mode>): Likewise.
16550 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
16551 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16552 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
16553 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16554 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
16555 (rs6000_gen_lvx): Likewise.
16556 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
16557 (altivec_expand_stv_builtin): Likewise.
16558 (altivec_expand_builtin): Likewise.
16559 * config/rs6000/vector.md: Likewise.
16561 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16564 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
16567 2018-03-20 Richard Biener <rguenther@suse.de>
16570 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
16571 sign-conversions as zero, fall back to standard scalar_stmt
16574 2018-03-20 Martin Liska <mliska@suse.cz>
16577 * predict.c (rebuild_frequencies): Handle case when we have
16578 PROFILE_ABSENT, but flag_guess_branch_prob is false.
16580 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16583 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
16584 flag_section_anchors.
16585 * varasm.c (use_blocks_for_decl_p): Remove hack for
16586 dw2_force_const_mem.
16589 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
16591 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
16592 be created, use lowpart_subreg of operands[0] rather than operands[0]
16594 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
16595 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
16596 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
16597 and n constraint instead of aarch64_shift_imm_di and Usd.
16598 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
16599 (*aarch64_<optab>_reg_minus<mode>3): ... this.
16601 2018-03-20 Sudakshina Das <sudi.das@arm.com>
16604 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
16605 to favor GPR over NEON registers.
16606 (<shift>di3_neon): Likewise.
16608 2018-03-20 Tom de Vries <tom@codesourcery.com>
16611 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
16612 (nvptx_process_pars): Emit bar.sync asap and alap.
16614 2018-03-20 Tom de Vries <tom@codesourcery.com>
16617 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
16618 seen_label if seen_label is already set.
16620 2018-03-20 Jakub Jelinek <jakub@redhat.com>
16623 * config/i386/i386.c (fold_builtin_cpu): For features above 31
16624 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
16625 Use 1U instead of 1. Formatting fixes.
16628 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
16629 instead of TREE_TYPE (s1) for the return value.
16631 2018-03-19 Jakub Jelinek <jakub@redhat.com>
16633 PR tree-optimization/84946
16634 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
16635 bitsize + bitsize in poly_uint64 rather than poly_int64.
16638 * dwarf2asm.c: Include fold-const.c.
16639 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
16640 of decl rather than decl itself.
16642 PR rtl-optimization/84643
16643 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
16645 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
16648 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
16649 calling assemble_variable.
16651 2018-03-19 Sudakshina Das <sudi.das@arm.com>
16654 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
16655 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
16657 2018-03-19 Jim Wilson <jimw@sifive.com>
16660 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
16661 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
16662 (riscv_first_stack_step): Likewise.
16663 (riscv_option_override): Use STACK_BOUNDARY instead of
16664 MIN_STACK_BOUNDARY.
16665 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
16666 MIN_STACK_BOUNDARY.
16667 (BIGGEST_ALIGNMENT): Set to 128.
16668 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
16669 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
16672 2018-03-19 Richard Biener <rguenther@suse.de>
16674 PR tree-optimization/84933
16675 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
16676 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
16678 2018-03-19 Richard Biener <rguenther@suse.de>
16680 PR tree-optimization/84859
16681 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
16682 (cond_if_else_store_replacement): Perform sinking operation on
16683 single-store BBs regardless of MAX_STORES_TO_SINK setting.
16684 Generalize what a BB with a single eligible store is.
16686 2018-03-19 Richard Biener <rguenther@suse.de>
16688 PR tree-optimization/84929
16689 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
16690 chrec_is_positive against non-chrec arg.
16692 2018-03-19 Tamar Christina <tamar.christina@arm.com>
16695 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
16697 2018-03-18 Martin Liska <mliska@suse.cz>
16699 PR rtl-optimization/84635
16700 * regrename.c (build_def_use): Use matches_mode only when
16703 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
16705 PR tree-optimization/84913
16706 * tree-vect-loop.c (vectorizable_reduction): Don't try to
16707 vectorize chains of COND_EXPRs.
16709 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16711 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
16713 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16715 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
16717 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
16719 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16721 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
16722 Kito Cheng <kito.cheng@gmail.com>
16724 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
16725 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
16726 (nds32_adjust_reg_alloc_order): New function.
16727 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
16729 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
16731 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
16732 nds32_print_operand, nds32_print_operand_address): Use
16733 HOST_WIDE_INT_PRINT_DEC instead.
16735 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
16737 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
16739 2018-03-17 Jakub Jelinek <jakub@redhat.com>
16742 * config/i386/i386.c (initial_ix86_tune_features,
16743 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
16744 unsigned long long.
16745 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
16746 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
16747 rather than 1u << ix86_tune. Formatting fix.
16748 (ix86_option_override_internal): Change ix86_arch_mask from
16749 unsigned int to unsigned HOST_WIDE_INT, initialize to
16750 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
16751 (ix86_function_specific_restore): Likewise.
16753 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16756 * postreload.c (reload_combine_recognize_pattern): Perform
16757 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
16758 truncate_int_for_mode the result for the destination's mode.
16761 * hsa-gen.c (mem_type_for_type): Fix comment typo.
16762 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
16764 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
16767 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
16770 * lra-assigns.c (lra_split_hard_reg_for): Don't use
16771 regno_allocno_class_array and sorted_pseudos.
16772 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
16773 insns where regno is used.
16775 2018-03-16 Martin Liska <mliska@suse.cz>
16778 * multiple_target.c (create_dispatcher_calls): Redirect
16779 reference in the symbol table.
16781 2018-03-16 Martin Liska <mliska@suse.cz>
16784 * multiple_target.c (create_dispatcher_calls): Redirect also
16787 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16791 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
16792 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
16793 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
16795 2018-03-16 Julia Koval <julia.koval@intel.com>
16797 * doc/invoke.texi (Skylake Server): Add CLWB.
16798 Cannonlake): Remove CLWB.
16800 2018-03-16 Jakub Jelinek <jakub@redhat.com>
16802 PR tree-optimization/84841
16803 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
16805 (FLOAT_ONE_CONST_TYPE): Define.
16806 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
16807 (sort_by_operand_rank): Put entries with higher constant_type last
16808 rather than first to match comments.
16810 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
16812 * config/nios2/nios2.md (movsi_internal): Fix thinko in
16815 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16818 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
16819 check and use address of target always.
16821 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
16824 * config/i386/i386.c (indirect_thunk_needed): Update comments.
16825 (indirect_thunk_bnd_needed): Likewise.
16826 (indirect_thunks_used): Likewise.
16827 (indirect_thunks_bnd_used): Likewise.
16828 (indirect_return_needed): New.
16829 (indirect_return_bnd_needed): Likewise.
16830 (output_indirect_thunk_function): Add a bool argument for
16832 (output_indirect_thunk_function): Don't generate alias for
16833 function return thunk.
16834 (ix86_code_end): Call output_indirect_thunk_function to generate
16835 function return thunks.
16836 (ix86_output_function_return): Set indirect_return_bnd_needed
16837 and indirect_return_needed instead of indirect_thunk_bnd_needed
16838 and indirect_thunk_needed.
16840 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
16842 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
16843 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
16844 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
16846 2018-03-15 David Malcolm <dmalcolm@redhat.com>
16847 Paul Hua <paul.hua.gm@gmail.com>
16850 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
16852 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
16854 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
16855 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
16856 resp. SFmode cases.
16858 2018-03-15 Tamar Christina <tamar.christina@arm.com>
16861 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
16862 instead of GET_MODE_SIZE when comparing Units.
16864 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
16867 * varasm.c (hash_section): Return an unchangeble hash value
16868 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
16869 Return !aarch64_can_use_per_function_literal_pools_p ().
16871 2018-03-15 Jakub Jelinek <jakub@redhat.com>
16874 * optabs.c (emit_conditional_move): Pass address of cmode's copy
16875 rather than address of cmode as last argument to prepare_cmp_insn.
16877 2018-03-15 Julia Koval <julia.koval@intel.com>
16879 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
16880 F_AVX512VNNI, F_AVX512BITALG): New.
16882 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
16885 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
16886 insn for floating-point loads and stores.
16888 2018-03-14 Carl Love <cel@us.ibm.com>
16890 * config/rs6000/rs6000-c.c: Add macro definitions for
16891 ALTIVEC_BUILTIN_VEC_PERMXOR.
16892 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
16893 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
16894 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
16895 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
16897 * config/doc/extend.texi: Add prototypes for vec_permxor.
16899 2018-03-14 David Malcolm <dmalcolm@redhat.com>
16902 * diagnostic-show-locus.c (class layout_point): Convert m_line
16903 from int to linenum_type.
16904 (line_span::comparator): Use linenum "compare" function when
16905 comparing line numbers.
16906 (test_line_span): New function.
16907 (layout_range::contains_point): Convert param "row" from int to
16909 (layout_range::intersects_line_p): Likewise.
16910 (layout::will_show_line_p): Likewise.
16911 (layout::print_source_line): Likewise.
16912 (layout::should_print_annotation_line_p): Likewise.
16913 (layout::print_annotation_line): Likewise.
16914 (layout::print_leading_fixits): Likewise.
16915 (layout::annotation_line_showed_range_p): Likewise.
16916 (struct line_corrections): Likewise for field m_row.
16917 (line_corrections::line_corrections): Likewise for param "row".
16918 (layout::print_trailing_fixits): Likewise.
16919 (layout::get_state_at_point): Likewise.
16920 (layout::get_x_bound_for_row): Likewise.
16921 (layout::print_line): Likewise.
16922 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
16923 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
16924 * input.c (selftest::test_linenum_comparisons): New function.
16925 (selftest::input_c_tests): Call it.
16926 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
16927 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
16928 * selftest.h (ASSERT_GT): New macro.
16929 (ASSERT_GT_AT): New macro.
16930 (ASSERT_LT): New macro.
16931 (ASSERT_LT_AT): New macro.
16933 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
16935 PR rtl-optimization/84780
16936 * combine.c (distribute_links): Don't make a link based on pc_rtx.
16938 2018-03-14 Martin Liska <mliska@suse.cz>
16940 * tree.c (record_node_allocation_statistics): Use
16941 get_stats_node_kind.
16942 (get_stats_node_kind): New function extracted from
16943 record_node_allocation_statistics.
16944 (free_node): Use get_stats_node_kind.
16946 2018-03-14 Richard Biener <rguenther@suse.de>
16948 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
16949 that the value-set of ANTIC_IN doesn't grow.
16952 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
16954 (BB_VISITED_WITH_VISITED_SUCCS): New define.
16955 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
16957 2018-03-14 Julia Koval <julia.koval@intel.com>
16959 * config.gcc (icelake-client, icelake-server): New.
16961 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
16962 (initial_ix86_arch_features): Ditto.
16963 (PTA_SKYLAKE): Add SGX.
16964 (PTA_ICELAKE): Remove.
16965 (PTA_ICELAKE_CLIENT): New.
16966 (PTA_ICELAKE_SERVER): New.
16967 (ix86_option_override_internal): Split up icelake on icelake client and
16969 (get_builtin_code_for_version): Ditto.
16970 (fold_builtin_cpu): Ditto.
16971 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
16972 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
16973 * config/i386/i386.h (processor_type): Ditto.
16974 * doc/invoke.texi: Ditto.
16976 2018-03-14 Jakub Jelinek <jakub@redhat.com>
16979 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
16980 INTEGER_CST offset, add it together with bitpos / 8 and
16981 sign extend based on POINTER_SIZE.
16985 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
16988 * config/i386/constraints.md (Yc): New register constraint.
16989 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
16990 Use Yc constraint for alternative 2 of operand 0. Remove
16991 preferred_for_speed attribute.
16993 2018-03-14 Richard Biener <rguenther@suse.de>
16995 PR tree-optimization/84830
16996 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
16997 with the old one to avoid oscillations.
16999 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
17002 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
17004 (assign_by_spills): Return a flag of reload assignment failure.
17005 Do not process the reload assignment failures. Do not spill other
17006 reload pseudos if they has the same reg class. Update n if
17008 (lra_assign): Add a return arg. Set up from the result of
17009 assign_by_spills call.
17010 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17011 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17012 usage_insns if it is not NULL.
17013 (spill_hard_reg_in_range): New function.
17014 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17015 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17016 function prototypes.
17017 (lra_assign): Change prototype.
17018 * lra.c (lra): Add code to deal with fails by splitting hard reg
17021 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
17023 * config/riscv/riscv.opt (mrelax): New option.
17024 * config/riscv/riscv.c (riscv_file_start): Emit ".option
17025 "norelax" when riscv_mrelax is disabled.
17026 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
17028 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17031 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
17032 reassociation for int modes.
17034 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17036 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
17037 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
17039 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
17040 * config/aarch64/aarch64-sve.md
17041 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
17042 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
17043 (*extend<mode><Vwide>2): Rename to...
17044 (aarch64_sve_extend<mode><Vwide>2): ...this.
17045 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
17046 renaming the old pattern to...
17047 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
17049 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
17050 define_expand, renaming the old pattern to...
17051 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
17052 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
17053 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
17054 account when deciding which SVE instruction the optab should use.
17055 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
17057 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17059 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
17060 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
17061 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
17062 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
17063 (tlsdesc_small_<mode>): Turn a define_expand and use
17064 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
17065 (tlsdesc_small_advsimd_<mode>): ...this.
17066 (tlsdesc_small_sve_<mode>): New pattern.
17068 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
17070 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
17071 (UNSPEC_UMUL_HIGHPART): New constants.
17072 (MUL_HIGHPART): New int iteraor.
17073 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
17074 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
17076 (*<su>mul<mode>3_highpart): New define_insn.
17078 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
17081 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17084 2018-03-13 Martin Liska <mliska@suse.cz>
17087 * (sem_item_optimizer::sem_item_optimizer): Initialize new
17089 (sem_item_optimizer::~sem_item_optimizer): Release it.
17090 (sem_item_optimizer::merge_classes): Register variable aliases.
17091 (sem_item_optimizer::fixup_pt_set): New function.
17092 (sem_item_optimizer::fixup_points_to_sets): Likewise.
17093 * ipa-icf.h: Declare new variables and functions.
17095 2018-03-13 Jakub Jelinek <jakub@redhat.com>
17097 PR middle-end/84834
17098 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
17099 integer_pow2p@2 and test integer_pow2p in condition.
17100 (A < 0 ? C : 0): Similarly for @1.
17102 PR middle-end/84831
17103 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
17104 characters starting at p contain '\0' character, don't look beyond
17108 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
17109 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
17112 * reg-stack.c (change_stack): Change update_end var from int to
17113 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
17114 also call set_block_for_insn on the newly added insns and rescan.
17117 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
17118 on the last operand.
17121 * tree.c (stabilize_reference_1): Return save_expr (e) for
17122 STATEMENT_LIST even if it doesn't have side-effects.
17124 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
17126 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
17128 2018-03-12 Renlin Li <renlin.li@arm.com>
17130 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
17131 aarch64_output_scalar_simd_mov_immediate.
17133 2018-03-12 Martin Sebor <msebor@redhat.com>
17135 PR tree-optimization/83456
17136 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
17137 for perfectly overlapping calls to memcpy.
17138 (gimple_fold_builtin_memory_chk): Same.
17139 (gimple_fold_builtin_strcpy): Handle no-warning.
17140 (gimple_fold_builtin_stxcpy_chk): Same.
17141 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
17143 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17145 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
17146 parameter. Use it for SFmode.
17147 (rs6000_function_arg_advance_1): Adjust.
17148 (rs6000_function_arg): Adjust.
17149 (rs6000_gimplify_va_arg): Pass false for that new parameter.
17151 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
17153 PR rtl-optimization/84169
17154 PR rtl-optimization/84780
17155 * combine.c (can_combine_p): Check for a 2-insn combination whether
17156 the destination register is used between the two insns, too.
17158 2018-03-12 Richard Biener <rguenther@suse.de>
17160 PR tree-optimization/84803
17161 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
17162 for refs DR analysis didn't process.
17164 2018-03-12 Richard Biener <rguenther@suse.de>
17166 PR tree-optimization/84777
17167 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
17168 force-vectorize loops ignore whether we are optimizing for size.
17170 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
17172 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
17173 (TARGET_MD_ASM_ADJUST): Define.
17175 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
17176 Kito Cheng <kito.cheng@gmail.com>
17177 Chung-Ju Wu <jasonwucj@gmail.com>
17179 * config/nds32/nds32.c (nds32_compute_stack_frame,
17180 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
17181 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
17182 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
17183 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
17184 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
17185 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
17186 * config/nds32/nds32.md (prologue, epilogue): Use macro
17187 NDS32_V3PUSH_AVAILABLE_P to do checking.
17189 2018-03-11 Jakub Jelinek <jakub@redhat.com>
17192 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
17193 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
17194 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
17195 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
17196 addition of most attributes on !orig_type_die or the attribute not
17197 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
17199 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17200 Chung-Ju Wu <jasonwucj@gmail.com>
17202 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
17203 __NDS32_VH__ macro.
17204 * config/nds32/nds32.opt (mvh): New option.
17206 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17207 Chung-Ju Wu <jasonwucj@gmail.com>
17209 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
17211 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
17212 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
17215 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
17216 Chung-Ju Wu <jasonwucj@gmail.com>
17218 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
17220 * config/nds32/nds32-multiple.md (strlensi): New pattern.
17221 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
17223 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
17224 Kito Cheng <kito.cheng@gmail.com>
17225 Chung-Ju Wu <jasonwucj@gmail.com>
17227 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
17228 UNSPEC_FFMISM and UNSPEC_FLMISM.
17229 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
17230 for ffb, ffmism and flmism.
17231 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
17232 (unspec_ffmism): Ditto.
17233 (unspec_flmism): Ditto.
17234 (nds32_expand_builtin_impl): Check if string extension is available.
17235 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
17236 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
17238 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
17241 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17244 * lra-assigns.c (assign_by_spills): Return a flag of reload
17245 assignment failure. Do not process the reload assignment
17246 failures. Do not spill other reload pseudos if they has the same
17248 (lra_assign): Add a return arg. Set up from the result of
17249 assign_by_spills call.
17250 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17251 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17252 usage_insns if it is not NULL.
17253 (spill_hard_reg_in_range): New function.
17254 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17255 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17256 function prototypes.
17257 (lra_assign): Change prototype.
17258 * lra.c (lra): Add code to deal with fails by splitting hard reg
17261 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
17264 * config/i386/i386.opt: Replace Enforcment with Enforcement.
17266 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
17269 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
17270 (dw_val_node): Add val_symbolic_view.
17271 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
17272 (symview_upper_bound): New.
17273 (new_line_info_table): Initialize symviews_since_reset.
17274 (dwarf2out_source_line): Count symviews_since_reset and set
17275 symview_upper_bound.
17276 (dw_val_equal_p): Handle symview.
17277 (add_AT_symview): New.
17278 (print_dw_val): Handle symview.
17279 (attr_checksum, attr_checksum_ordered): Likewise.
17280 (same_dw_val_p, size_of_die): Likewise.
17281 (value_format, output_die): Likewise.
17282 (add_high_low_attributes): Use add_AT_symview for entry_view.
17283 (dwarf2out_finish): Reset symview_upper_bound, clear
17286 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
17289 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
17290 Add strict argument and use it.
17291 (rs6000_split_multireg_move): Update for new strict argument.
17292 (mem_operand_gpr): Disallow all non-offsettable addresses.
17293 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
17295 2018-03-09 Jakub Jelinek <jakub@redhat.com>
17298 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
17299 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
17300 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
17303 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
17304 decl, use remap_type if we want to use the type.
17306 2018-03-09 Martin Sebor <msebor@redhat.com>
17308 PR tree-optimization/84526
17309 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17311 (builtin_access::generic_overlap): Be prepared to handle non-array
17314 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
17316 PR rtl-optimization/84682
17317 * lra-constraints.c (process_address_1): Check is_address flag
17318 for address constraints.
17319 (process_alt_operands): Likewise.
17320 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
17321 preprocess_constraints.
17322 * recog.h (preprocess_constraints): Add oploc parameter.
17324 * recog.c (preprocess_constraints): Test address_operand for
17325 CT_ADDRESS constraints.
17327 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
17330 * lra-assigns.c (assign_by_spills): Return a flag of reload
17331 assignment failure. Do not process the reload assignment
17332 failures. Do not spill other reload pseudos if they has the same
17334 (lra_assign): Add a return arg. Set up from the result of
17335 assign_by_spills call.
17336 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17337 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
17338 usage_insns if it is not NULL.
17339 (spill_hard_reg_in_range): New function.
17340 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17341 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17342 function prototypes.
17343 (lra_assign): Change prototype.
17344 * lra.c (lra): Add code to deal with fails by splitting hard reg
17347 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17350 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
17351 Accept complain bool parameter. Only emit errors if it is true.
17352 (arm_parse_cpu_option_name): Likewise.
17353 (arm_target_thumb_only): Adjust callers of the above.
17354 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
17355 prototype to take a default true bool parameter.
17356 (arm_parse_arch_option_name): Likewise.
17358 2018-03-09 David Malcolm <dmalcolm@redhat.com>
17359 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
17363 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
17364 * configure: Regenerate.
17365 * configure.ac ("linker --version-script option"): New.
17366 ("linker soname option"): New.
17368 2018-03-09 Richard Biener <rguenther@suse.de>
17370 PR tree-optimization/84775
17371 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
17372 immediate uses of predicate stmts and mark them modified.
17375 PR tree-optimization/84178
17376 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17378 (version_loop_for_if_conversion): Delay update_ssa call.
17379 (tree_if_conversion): Delay update_ssa until after predicate
17382 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
17385 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
17386 when the function accesses prior frames.
17388 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17391 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
17392 gen_llsym, otherwise call maybe_gen_llsym.
17394 PR inline-asm/84742
17395 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
17396 has ',' character inside of it.
17398 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17401 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
17402 as clobbering CC_REGNUM.
17404 2018-03-08 Richard Biener <rguenther@suse.de>
17406 PR middle-end/84552
17407 * tree-scalar-evolution.c: Include tree-into-ssa.h.
17408 (follow_copies_to_constant): Do not follow SSA names registered
17411 2018-03-08 Richard Biener <rguenther@suse.de>
17413 PR tree-optimization/84178
17414 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17416 (version_loop_for_if_conversion): Delay update_ssa call.
17417 (tree_if_conversion): Delay update_ssa until after predicate
17420 2018-03-08 David Malcolm <dmalcolm@redhat.com>
17422 PR tree-optimization/84178
17423 * tree-if-conv.c (release_bb_predicate): Remove the
17424 the assertion that the stmts have NULL use_ops.
17425 Discard the statements, asserting that they haven't
17426 yet been added to a BB.
17428 2018-03-08 Richard Biener <rguenther@suse.de>
17430 PR tree-optimization/84746
17431 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
17432 (phi_translate): Pass in destination ANTIC_OUT set.
17433 (phi_translate_1): Likewise. For a simplified result lookup
17434 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
17435 (phi_translate_set): Adjust.
17436 (do_pre_regular_insertion): Likewise.
17437 (do_pre_partial_partial_insertion): Likewise.
17439 2018-03-08 Martin Liska <mliska@suse.cz>
17441 PR gcov-profile/84735
17442 * doc/gcov.texi: Document usage of profile files.
17443 * gcov-io.h: Document changes in the format.
17445 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
17449 * dwarf2out.c (struct dw_line_info_table): Update comments for
17451 (FORCE_RESET_NEXT_VIEW): New.
17452 (FORCE_RESETTING_VIEW_P): New.
17453 (RESETTING_VIEW_P): Check for -1 too.
17454 (ZERO_VIEW_P): Likewise.
17455 (new_line_info_table): Force-reset next view.
17456 (dwarf2out_begin_function): Likewise.
17457 (dwarf2out_source_line): Simplify zero_view_p initialization.
17458 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
17459 view directly. Omit view when omitting .loc at line 0.
17461 2018-03-08 Jakub Jelinek <jakub@redhat.com>
17463 PR tree-optimization/84740
17464 * tree-switch-conversion.c (process_switch): Call build_constructors
17465 only if info.phi_count is non-zero.
17467 PR tree-optimization/84739
17468 * tree-tailcall.c (find_tail_calls): Check call arguments against
17469 DECL_ARGUMENTS (current_function_decl) rather than
17470 DECL_ARGUMENTS (func) when checking for tail recursion.
17472 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17474 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
17475 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
17476 Volker Reichelt's entry and add entries for people that perform
17477 GCC fuzzy testing and report numerous bugs.
17479 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
17482 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
17483 readonly data in sdata, if that is disabled.
17484 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
17485 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
17486 -mreadonly-in-sdata option.
17488 2018-03-07 Martin Sebor <msebor@redhat.com>
17490 PR tree-optimization/84468
17491 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
17492 basic block when looking for nul assignment.
17494 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
17497 * except.h (output_function_exception_table): Adjust prototype.
17498 * except.c (output_function_exception_table): Remove FNNAME parameter
17499 and add SECTION parameter. Ouput one part of the table at a time.
17500 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
17501 the first part of the exception table and emit unwind directives.
17502 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
17503 (i386_pe_seh_cold_init): Likewise.
17504 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
17505 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
17506 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
17507 (ix86_output_call_insn): Emit a nop in one more case for SEH.
17508 * config/i386/winnt.c: Include except.h.
17509 (struct seh_frame_state): Add reg_offset, after_prologue and
17510 in_cold_section fields.
17511 (i386_pe_seh_end_prologue): Set seh->after_prologue.
17512 (i386_pe_seh_cold_init): New function.
17513 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
17514 to seh->in_cold_section.
17515 (seh_emit_push): Record the offset of the push.
17516 (seh_emit_save): Record the offet of the save.
17517 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
17518 Test seh->after_prologue to disregard the epilogue.
17519 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
17520 (i386_pe_end_cold_function): New function.
17522 2018-03-07 Jakub Jelinek <jakub@redhat.com>
17525 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
17526 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
17529 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
17531 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
17532 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
17534 PR middle-end/84723
17535 * multiple_target.c: Include tree-inline.h and intl.h.
17536 (expand_target_clones): Diagnose and fail if node->definition and
17537 !tree_versionable_function_p (node->decl).
17539 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
17541 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
17543 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
17544 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
17545 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
17547 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17550 * combine.c (try_combine): Use reg_or_subregno instead of handling
17551 just paradoxical SUBREGs and REGs.
17553 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
17555 * config/arc/arc.c (arc_finalize_pic): Remove function.
17556 (arc_must_save_register): We use single base PIC register, remove
17557 checks to save/restore the PIC register.
17558 (arc_expand_prologue): Likewise.
17559 * config/arc/arc-protos.h (arc_set_default_type_attributes):
17561 (arc_verify_short): Likewise.
17562 (arc_attr_type): Likewise.
17563 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
17564 (walk_stores): Likewise.
17565 (arc_address_cost): Make it static.
17566 (arc_verify_short): Likewise.
17567 (branch_dest): Likewise.
17568 (arc_attr_type): Likewise.
17569 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
17570 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
17571 (arc_final_prescan_insn): Remove inserting the nops due to
17572 hardware hazards. It is done in reorg step.
17573 (insn_length_variant_t): Remove.
17574 (insn_length_parameters_t): Likewise.
17575 (arc_insn_length_parameters): Likewise.
17576 (arc_get_insn_variants): Likewise.
17577 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
17579 2018-03-06 Jakub Jelinek <jakub@redhat.com>
17581 PR inline-asm/84683
17582 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
17585 PR tree-optimization/84687
17586 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
17588 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
17590 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17592 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
17593 Rename to ppc_speculation_barrier.
17594 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
17595 __builtin_ppc_speculation_barrier.
17597 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17600 * combine.c (combine_simplify_rtx): Don't try to simplify if
17601 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
17604 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
17606 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
17607 to 32 bytes when compiling for POWER9.
17609 2018-03-05 Jakub Jelinek <jakub@redhat.com>
17612 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
17613 regparm >= 3 with no arg reg available also for calls with
17614 flag_force_indirect_call. Pass decl to ix86_function_regparm.
17617 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
17619 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
17621 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
17624 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
17626 2018-03-05 Richard Biener <rguenther@suse.de>
17628 PR tree-optimization/84486
17629 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
17630 When inserting a __builtin_assume_aligned call set the LHS
17631 SSA name alignment info accordingly.
17633 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
17635 PR tree-optimization/84114
17636 * config/aarch64/aarch64.c (aarch64_reassociation_width)
17637 Avoid reassociation of FLOAT_MODE addition.
17639 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
17641 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
17642 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
17643 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
17644 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
17645 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
17646 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
17647 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
17649 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
17650 __builtin_ia32_wbinvd): New builtins.
17651 (SPECIAL_ARGS2): New.
17652 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
17653 (SPECIAL_ARGS2): New.
17654 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
17655 (ix86_valid_target_attribute_inner_p): Ditto.
17656 (ix86_init_mmx_sse_builtins): Add special_args2.
17657 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
17658 TARGET_WBNOINVD_P): New.
17659 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
17660 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
17661 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
17662 * config/i386/immintrin.h (_wbinvd): New intrinsic.
17663 * config/i386/pconfigintrin.h: New file.
17664 * config/i386/wbnoinvdintrin.h: Ditto.
17665 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
17667 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
17669 2018-03-05 Richard Biener <rguenther@suse.de>
17671 PR tree-optimization/84670
17672 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17674 (BB_VISITED_WITH_VISITED_SUCCS): New define.
17675 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17676 (compute_antic_aux): Only assert the number of values in ANTIC_IN
17677 doesn't grow if all successors (recursively) were visited at least
17680 2018-03-05 Richard Biener <rguenther@suse.de>
17682 PR tree-optimization/84650
17683 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
17684 if executed in the loop pipeline.
17686 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
17688 * doc/configfiles.texi (Configuration Files): Move info about
17689 conditionalizing $target-protos.h to...
17690 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
17691 differs from $target-protos.h.
17693 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
17694 Chung-Ju Wu <jasonwucj@gmail.com>
17696 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
17697 * config/nds32/nds32-multiple.md (setmemsi): Define.
17698 * config/nds32/nds32-memory-manipulation.c
17699 (nds32_gen_dup_4_byte_to_word_value): New.
17700 (emit_setmem_word_loop): New.
17701 (emit_setmem_byte_loop): New.
17702 (nds32_expand_setmem_loop): New.
17703 (nds32_expand_setmem_loop_v3m): New.
17704 (nds32_expand_setmem_unroll): New.
17705 (nds32_expand_setmem): New.
17707 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17708 Chung-Ju Wu <jasonwucj@gmail.com>
17710 * config/nds32/nds32-memory-manipulation.c
17711 (nds32_emit_load_store): New.
17712 (nds32_emit_post_inc_load_store): New.
17713 (nds32_emit_mem_move): New.
17714 (nds32_emit_mem_move_block): New.
17715 (nds32_expand_movmemsi_loop_unknown_size): New.
17716 (nds32_expand_movmemsi_loop_known_size): New.
17717 (nds32_expand_movmemsi_loop): New.
17718 (nds32_expand_movmemsi_unroll): New.
17719 (nds32_expand_movmemqi): Rename ...
17720 (nds32_expand_movmemsi): ... to this.
17721 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
17722 (movmemsi): ... to this.
17723 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
17724 (nds32_expand_movmemsi): ... to this.
17726 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17727 Monk Chiang <sh.chiang04@gmail.com>
17728 Chung-Ju Wu <jasonwucj@gmail.com>
17730 * config/nds32/nds32-protos.h
17731 (nds32_expand_load_multiple): New arguments.
17732 (nds32_expand_store_multiple): Ditto.
17733 (nds32_valid_multiple_load_store): Rename ...
17734 (nds32_valid_multiple_load_store_p): ... to this.
17735 * config/nds32/nds32-memory-manipulation.c
17736 (nds32_expand_load_multiple): Refine implementation.
17737 (nds32_expand_store_multiple): Ditto.
17738 * config/nds32/nds32-multiple.md
17739 (load_multiple): Update nds32_expand_load_multiple interface.
17740 (store_multiple): Update nds32_expand_store_multiple interface.
17741 * config/nds32/nds32-predicates.c
17742 (nds32_valid_multiple_load_store): Rename ...
17743 (nds32_valid_multiple_load_store_p): ... to this and refine
17745 * config/nds32/predicates.md
17746 (nds32_load_multiple_and_update_address_operation): New predicate.
17747 (nds32_store_multiple_and_update_address_operation): New predicate.
17749 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
17750 Chung-Ju Wu <jasonwucj@gmail.com>
17752 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
17753 (combo): New attribute.
17754 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
17756 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
17758 * config/nds32/nds32.opt: Change -mcmodel= default value.
17760 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
17761 Monk Chiang <sh.chiang04@gmail.com>
17762 Chung-Ju Wu <jasonwucj@gmail.com>
17764 * config/nds32/constants.md (unspec_element): New enum.
17765 * config/nds32/constraints.md (Umw): New constraint.
17766 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
17767 * config/nds32/nds32-intrinsic.md: Likewise.
17768 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
17769 (nds32_valid_smw_lwm_base_p): New.
17770 (nds32_output_smw_single_word): New.
17771 (nds32_output_lmw_single_word): New.
17772 (nds32_expand_unaligned_load): New.
17773 (nds32_expand_unaligned_store): New.
17774 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
17775 (nds32_output_smw_single_word): Declare.
17776 (nds32_output_lmw_single_word): Declare.
17777 (nds32_expand_unaligned_load): Declare.
17778 (nds32_expand_unaligned_store): Declare.
17779 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
17780 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
17781 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
17782 NDS32_BUILTIN_UASTORE_DW.
17783 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
17786 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
17787 Kito Cheng <kito.cheng@gmail.com>
17788 Chung-Ju Wu <jasonwucj@gmail.com>
17790 * config/nds32/nds32-intrinsic.c
17791 (nds32_expand_builtin_null_ftype_reg): Delete.
17792 (nds32_expand_builtin_reg_ftype_imm): Ditto.
17793 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
17794 (nds32_read_argument): New.
17795 (nds32_legitimize_target): Ditto.
17796 (nds32_legitimize_argument): Ditto.
17797 (nds32_check_constant_argument): Ditto.
17798 (nds32_expand_unop_builtin): Ditto.
17799 (nds32_expand_unopimm_builtin): Ditto.
17800 (nds32_expand_binop_builtin): Ditto.
17801 (nds32_builtin_decl_impl): Ditto.
17802 (builtin_description): Ditto.
17803 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
17804 (nds32_init_builtins_impl): Ditto.
17805 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
17806 (nds32_builtin_decl): New.
17807 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
17808 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
17810 2018-03-02 Jeff Law <law@redhat.com>
17812 * reorg.c (stop_search_p): Handle DEBUG_INSN.
17813 (redundant_insn, fill_simple_delay_slots): Likewise.
17814 (fill_slots_from_thread): Likewise.
17815 * resource.c (mark_referenced_resources): Likewise.
17816 (mark_set_resources, find_dead_or_set_registers): Likewise.
17818 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17820 * substring-locations.h (format_warning_va): Formatting fix for
17821 ATTRIBUTE_GCC_DIAG.
17822 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
17824 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
17825 * substring-locations.c: Include intl.h.
17826 (format_warning_va): Turned into small wrapper around
17827 format_warning_n_va, renamed to ...
17828 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
17829 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
17831 (format_warning_at_substring_n): New function.
17832 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
17833 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
17834 format_warning_at_substring with just a shorter name instead of
17835 const function pointer.
17836 (fmtwarn_n): New function.
17837 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
17838 appropriate, get rid of all the fmtstr temporaries, move conditionals
17839 with G_() wrapped string literals directly into fmtwarn arguments,
17840 cast dir.len to (int), formatting fixes.
17842 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
17844 * doc/invoke.texi: Remove "Cilk Plus" references.
17846 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17847 Richard Biener <rguenther@suse.de>
17850 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
17851 for error or warning attributes if CALL_FROM_THUNK_P is set.
17854 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17857 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
17858 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
17861 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
17862 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
17864 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
17865 instead of -1U in last predictors element's probability member.
17867 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
17870 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17871 arguments if they are comparable.
17873 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17875 PR tree-optimization/84634
17876 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
17877 masks and masked_loop_p with a single loop_masks, making sure it's
17878 null for bb vectorization.
17880 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
17882 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
17883 (vect_analyze_data_ref_access): Use loop->safe_len rather than
17884 loop->force_vectorize to check whether there is no alias.
17886 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17889 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
17891 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
17893 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
17894 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
17895 instead of a loop around prev_real_insn.
17896 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
17899 PR inline-asm/84625
17900 * config/i386/i386.c (ix86_print_operand): Use conditional
17901 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
17904 2018-03-02 Richard Biener <rguenther@suse.de>
17906 PR tree-optimization/84427
17907 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
17908 (bitmap_set_subtract_values): Rewrite to handle multiple
17911 (prune_clobbered_mems): Likewise.
17912 (phi_translate): Take edge instead of pred/phiblock.
17913 (phi_translate_1): Likewise.
17914 (phi_translate_set): Likewise. Insert all translated
17915 exprs for a value into the set, keeping possibly multiple
17916 expressions per value.
17917 (compute_antic_aux): Adjust for phi_translate changes.
17918 When intersecting union the expressions and prune those
17919 not in the final value set, keeping possibly multiple
17920 expressions per value. Do not use value-insertion
17921 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
17922 all expressions. Add verification that the value-sets
17923 only shrink during iteration.
17924 (compute_partial_antic_aux): Adjust for the phi_translate changes.
17925 (do_pre_regular_insertion): Likewise.
17926 (do_pre_partial_partial_insertion): Likewise.
17928 2018-03-02 Richard Biener <rguenther@suse.de>
17931 * config/darwin.c (saved_debug_info_level): New static global.
17932 (darwin_asm_lto_start): Disable debug info generation for LTO out.
17933 (darwin_asm_lto_end): Restore debug info generation settings.
17935 2018-03-01 Martin Liska <mliska@suse.cz>
17938 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
17939 volatile arguments.
17941 2018-03-01 Richard Biener <rguenther@suse.de>
17944 * dwarf2out.c (gen_variable_die): Properly handle late VLA
17945 type annotation with LTO when debug was disabled at compile-time.
17947 2018-03-01 Matthew Fortune <mfortune@gmail.com>
17949 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
17951 (mips_final_postscan_insn): Likewise.
17953 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
17955 PR rtl-optimization/84528
17956 * alias.c (init_alias_target): Add commentary.
17957 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
17958 a unique base value if the frame pointer is not eliminated
17959 to the stack pointer.
17961 2018-03-01 Tom de Vries <tom@codesourcery.com>
17963 PR rtl-optimization/83327
17964 * lra-int.h (hard_regs_spilled_into): Declare.
17965 * lra.c (hard_regs_spilled_into): Define.
17966 (init_reg_info): Init hard_regs_spilled_into.
17967 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
17968 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
17969 (process_bb_lives): Handle hard_regs_spilled_into.
17970 (lra_create_live_ranges_1): Before doing liveness propagation, clear
17971 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
17973 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
17975 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
17976 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
17977 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
17978 * config/rs6000/aix72.h: New file.
17980 2018-02-28 Jakub Jelinek <jakub@redhat.com>
17982 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
17983 instead of warning_at with conditional singular and plural messages
17987 * stor-layout.c (update_alignment_for_field): For
17988 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
17989 && !DECL_PACKED (field), do the alignment update, just use
17990 only desired_align instead of MAX (type_align, desired_align)
17992 (place_field): Don't do known_align < desired_align handling
17993 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
17994 is non-NULL, instead do it after rli->prev_field handling and
17995 only if not within a bitfield word. For DECL_PACKED (field)
17996 use type_align of BITS_PER_UNIT.
17998 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
18000 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
18001 superfluous parentheses and trailing spaces.
18003 2018-02-28 Richard Biener <rguenther@suse.de>
18005 PR tree-optimization/84584
18006 * graphite-scop-detection.c (scop_detection::add_scop): Discard
18007 SCoPs with fake exit edge.
18009 2018-02-28 Martin Liska <mliska@suse.cz>
18012 * timevar.c (timer::print): Fix format to properly print 100%
18015 2018-02-28 Richard Biener <rguenther@suse.de>
18017 PR middle-end/84607
18018 * genmatch.c (capture_info::walk_match): Do not mark
18019 captured expressions without operands as expr_p given
18020 they act more like predicates and should be subject to
18021 "lost tail" side-effect preserving.
18023 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
18025 PR rtl-optimization/81611
18026 * auto-inc-dec.c (attempt_change): Move dead note from
18027 mem_insn if it's the next use of regno
18028 (find_address): Take address use of reg holding
18029 non-incremented value. Add parm to limit search to the named
18031 (merge_in_block): Attempt to use a mem insn that is the next
18032 use of the original regno.
18034 2018-02-27 Martin Sebor <msebor@redhat.com>
18037 * doc/invoke.texi (-Wmissing-attributes): New option.
18038 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
18040 2018-02-27 Martin Sebor <msebor@redhat.com>
18042 PR translation/84207
18043 * diagnostic-core.h (warning_n, error_n, inform_n): Change
18044 n argument to unsigned HOST_WIDE_INT.
18045 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
18046 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
18047 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
18048 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
18050 2018-02-27 Richard Biener <rguenther@suse.de>
18052 PR tree-optimization/84512
18053 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
18054 Do not use the estimate returned from record_stmt_cost for
18055 the scalar iteration cost but sum properly using add_stmt_cost.
18057 2018-02-27 Richard Biener <rguenther@suse.de>
18059 PR tree-optimization/84466
18060 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18061 Adjust last change to less strictly validate use operands.
18063 2018-02-27 Martin Liska <mliska@suse.cz>
18065 PR gcov-profile/84548
18066 * gcov.c (process_file): Allow partial overlap and consider it
18067 also as group functions.
18068 (output_lines): Properly calculate range of lines for a group.
18070 2018-02-27 Martin Liska <mliska@suse.cz>
18072 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
18073 'ggc' suffixes. Change first column width.
18074 (timer::print): Fix formatting of the column.
18076 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
18078 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
18079 preserve inline entry blocks for the sake of debug inline
18080 entry point markers alone.
18081 (remove_unused_locals): Suggest in comments a better place to
18082 force the preservation of inline entry blocks that are
18083 otherwise unused, but do not preserve them.
18085 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18087 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
18089 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18092 * config/i386/constraints.md (Bs): Replace
18093 ix86_indirect_branch_register with
18094 TARGET_INDIRECT_BRANCH_REGISTER.
18096 * config/i386/i386.md (indirect_jump): Likewise.
18097 (tablejump): Likewise.
18098 (*sibcall_memory): Likewise.
18099 (*sibcall_value_memory): Likewise.
18100 Peepholes of indirect call and jump via memory: Likewise.
18101 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
18102 (*sibcall_value_GOT_32): Likewise.
18103 * config/i386/predicates.md (indirect_branch_operand): Likewise.
18104 (GOT_memory_operand): Likewise.
18105 (call_insn_operand): Likewise.
18106 (sibcall_insn_operand): Likewise.
18107 (GOT32_symbol_operand): Likewise.
18108 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
18110 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18112 PR rtl-optimization/83496
18113 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
18114 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
18115 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
18116 redundant insn, if any.
18117 (relax_delay_slots): Likewise.
18118 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
18120 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
18122 PR tree-optimization/83965
18123 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
18124 that grouped statements are part of a reduction chain. Return
18125 true if the statement is not marked as a reduction itself but
18126 is part of a group.
18127 (vect_recog_dot_prod_pattern): Don't check whether the statement
18128 is part of a group here.
18129 (vect_recog_sad_pattern): Likewise.
18130 (vect_recog_widen_sum_pattern): Likewise.
18132 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
18135 * final.c (rest_of_clean_state): Also look for calls inside sequences.
18137 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18140 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
18142 (ix86_output_indirect_function_return): New prototype.
18143 (ix86_split_simple_return_pop_internal): Likewise.
18144 * config/i386/i386.c (indirect_return_via_cx): New.
18145 (indirect_return_via_cx_bnd): Likewise.
18146 (indirect_thunk_name): Handle return va CX_REG.
18147 (output_indirect_thunk_function): Create alias for
18148 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
18149 (ix86_output_indirect_jmp): Remove the bool argument.
18150 (ix86_output_indirect_function_return): New function.
18151 (ix86_split_simple_return_pop_internal): Likewise.
18152 * config/i386/i386.md (*indirect_jump): Don't pass false
18153 to ix86_output_indirect_jmp.
18154 (*tablejump_1): Likewise.
18155 (simple_return_pop_internal): Change it to define_insn_and_split.
18156 Call ix86_split_simple_return_pop_internal to split it for
18157 -mfunction-return=.
18158 (simple_return_indirect_internal): Call
18159 ix86_output_indirect_function_return instead of
18160 ix86_output_indirect_jmp.
18162 2018-02-26 Jakub Jelinek <jakub@redhat.com>
18165 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
18166 memset and value initialization afterwards.
18168 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
18170 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
18172 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18175 * common/config/aarch64/aarch64-common.c
18176 (aarch_option_optimization_table[]): Switch
18177 off fomit-frame-pointer
18179 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18180 Chung-Ju Wu <jasonwucj@gmail.com>
18182 * config/nds32/nds32-multiple.md (load_multiple): Disallow
18184 (store_multiple): Ditto.
18186 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
18188 * config.gcc: Add --with-cpu support for nds32 target.
18189 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
18190 * config/nds32/nds32.opt: Add -mcpu= option.
18192 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
18194 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
18195 isel=yes): Warn for these deprecated options.
18197 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
18199 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
18200 ISA_2_5_MASKS_EMBEDDED.
18202 2018-02-23 Jakub Jelinek <jakub@redhat.com>
18204 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
18205 p->max as pointers rather than using iterative_hash_expr.
18207 2018-02-23 Carl Love <cel@us.ibm.com>
18209 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
18210 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
18212 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
18213 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
18214 P8V_BUILTIN_VEC_VUNSIGNED2.
18216 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
18219 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
18220 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
18222 * lra-constraints.c (curr_insn_transform): Set up
18223 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
18225 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
18226 * lra-eliminations.c (spill_pseudos): Ditto.
18227 (process_insn_for_elimination): Ditto.
18228 * lra-lives.c (reg_early_clobber_p): Use the new macros.
18229 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
18230 LRA_NON_CLOBBERED_ALT.
18232 2018-02-22 Martin Sebor <msebor@redhat.com>
18234 PR tree-optimization/84480
18235 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
18236 to maybe_diag_stxncpy_trunc. Call it.
18237 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
18238 from gimple_fold_builtin_strcpy. Print inlining stack.
18239 (handle_builtin_stxncpy): Print inlining stack.
18240 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
18242 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
18245 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
18246 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
18247 and -fcheck-pointer-bounds are used together.
18248 (indirect_thunk_prefix): New enum.
18249 (indirect_thunk_need_prefix): New function.
18250 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
18251 "_nt" instead of "_bnd" for NOTRACK prefix.
18252 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
18253 (output_indirect_thunk_function): Likewise.
18255 (ix86_code_end): Update output_indirect_thunk_function calls.
18256 (ix86_output_indirect_branch_via_reg): Replace
18257 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
18258 (ix86_output_indirect_branch_via_push): Likewise.
18259 (ix86_output_function_return): Likewise.
18260 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
18261 incompatible with -fcf-protection=branch and
18262 -fcheck-pointer-bounds.
18264 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18267 * config/aarch64/aarch64.c (aarch64_print_address_internal):
18268 Change gcc_assert call to output_operand_lossage.
18270 2018-02-22 Steve Ellcey <sellcey@cavium.com>
18272 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
18274 2018-02-22 DJ Delorie <dj@redhat.com>
18275 Sebastian Perta <sebastian.perta@renesas.com>
18276 Oleg Endo <olegendo@gcc.gnu.org>
18278 * config/rx/rx.c (rx_rtx_costs): New function.
18279 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
18281 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
18283 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
18285 2018-02-22 Martin Liska <mliska@suse.cz>
18288 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
18289 Add "native" as a possible value.
18291 2018-02-22 Martin Liska <mliska@suse.cz>
18294 * config/i386/i386.c (ix86_option_override_internal):
18295 Add "native" as a possible value for -march and -mtune.
18297 2018-02-22 Jakub Jelinek <jakub@redhat.com>
18300 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
18301 to all type variants.
18303 PR tree-optimization/84503
18304 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
18305 width as info->bitpos + info->bitsize - start.
18306 (merged_store_group::merge_overlapping): Simplify width computation.
18307 (check_no_overlap): New function.
18308 (imm_store_chain_info::try_coalesce_bswap): Compute expected
18309 start + width and last_order of the group, fail if check_no_overlap
18311 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
18312 to group if check_no_overlap fails.
18314 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18316 * config/rs6000/altivec.md: Delete contraint arguments to
18317 define_expand, define_split, and define_peephole2, and in
18318 define_insn_and_split if always unused.
18319 * config/rs6000/darwin.md: Ditto.
18320 * config/rs6000/dfp.md: Ditto.
18321 * config/rs6000/rs6000.md: Ditto.
18322 * config/rs6000/sync.md: Ditto.
18323 * config/rs6000/vector.md: Ditto.
18324 * config/rs6000/vsx.md: Ditto.
18326 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
18328 * config/rs6000/altivec.md: Write output control strings as braced
18329 blocks instead of double-quoted strings.
18330 * config/rs6000/darwin.md: Ditto.
18331 * config/rs6000/rs6000.md: Ditto.
18332 * config/rs6000/vector.md: Ditto.
18333 * config/rs6000/vsx.md: Ditto.
18335 2018-02-21 Jason Merrill <jason@redhat.com>
18337 PR c++/84314 - ICE with templates and fastcall attribute.
18338 * attribs.c (build_type_attribute_qual_variant): Remove assert.
18340 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18342 * ipa-cp.c (determine_versionability): Fix comment typos.
18344 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
18347 * ipa-cp.c (determine_versionability): Do not version functions caling
18350 2018-02-21 Martin Liska <mliska@suse.cz>
18353 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
18354 Add "native" as a possible value.
18355 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
18356 the macro when native cpu detection is available.
18358 2018-02-21 Martin Liska <mliska@suse.cz>
18361 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
18362 Add "native" as a possible value.
18363 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
18364 when native cpu detection is available.
18366 2018-02-21 Jakub Jelinek <jakub@redhat.com>
18367 Martin Sebor <msebor@redhat.com>
18369 PR tree-optimization/84478
18370 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
18372 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
18373 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
18374 support which is conservatively correct, for 2 only stay conservative
18375 for maxlen. Formatting and comment capitalization fixes. Add STRICT
18376 argument to the 2 argument get_range_strlen, adjust 6 arg
18377 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
18379 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
18380 (gimple_fold_builtin_strlen): Pass true as last argument to
18383 2018-02-20 Martin Sebor <msebor@redhat.com>
18385 PR middle-end/84095
18386 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
18387 (builtin_memref::set_base_and_offset): Same. Handle inner references.
18388 (builtin_memref::builtin_memref): Factor out parts into
18389 set_base_and_offset and call it.
18391 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18393 PR middle-end/84406
18394 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
18395 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
18396 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
18397 search at the associated MODE_INT.
18399 2018-02-20 Jeff Law <law@redhat.com>
18401 PR middle-end/82123
18402 PR tree-optimization/81592
18403 PR middle-end/79257
18404 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
18405 for range data rather than using global data.
18406 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
18407 range data rather than using global data.
18408 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
18409 pass it to children as needed.
18410 (struct directive::fmtresult): Similarly.
18411 (struct directive::set_width): Similarly.
18412 (struct directive::set_precision): Similarly.
18413 (format_integer, format_directive, parse_directive): Similarly.
18414 (format_none): Accept unnamed vr_values parameter.
18415 (format_percent, format_floating, format_character): Similarly.
18416 (format_string, format_plain): Similarly.
18417 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
18418 the EVRP range analyzer for range data rather than using global data.
18419 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
18420 gimple-ssa-evrp-analyze.h
18421 (class sprintf_dom_walker): Add after_dom_children member function.
18422 Add evrp_range_analyzer member.
18423 (sprintf_dom_walker::before_dom_children): Call into the EVRP
18424 range analyzer as needed.
18425 (sprintf_dom_walker::after_dom_children): New member function.
18426 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
18428 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
18429 (evrp_range_analyzer::pop_to_marker): Likewise.
18431 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
18433 PR tree-optimization/84419
18434 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
18435 with the required type if its current type is compatible but
18438 2018-02-20 Jakub Jelinek <jakub@redhat.com>
18440 PR middle-end/82004
18441 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
18442 after vectorization.
18444 2018-02-20 Martin Liska <mliska@suse.cz>
18447 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
18448 possible values if we don't have a hint.
18450 2018-02-20 Martin Liska <mliska@suse.cz>
18454 * final.c (shorten_branches): Build align_tab array with one
18456 * opts.c (finish_options): Add alignment option limit check.
18457 (MAX_CODE_ALIGN): Likewise.
18458 (MAX_CODE_ALIGN_VALUE): Likewise.
18459 * doc/invoke.texi: Document maximum allowed option value for
18460 all -falign-* options.
18462 2018-02-19 Jakub Jelinek <jakub@redhat.com>
18465 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
18466 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
18467 * var-tracking.c (emit_note_insn_var_location): Remove all references
18468 to NOTE_INSN_CALL_ARG_LOCATION.
18469 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
18470 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
18471 Use copy_rtx_if_shared.
18472 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
18473 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
18474 (dwarf2out_var_location): Remove handling of
18475 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
18477 * final.c (final_scan_insn): Remove all references to
18478 NOTE_INSN_CALL_ARG_LOCATION.
18479 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
18480 before dumping final insns.
18481 * except.c (emit_note_eh_region_end): Remove all references to
18482 NOTE_INSN_CALL_ARG_LOCATION.
18483 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
18484 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
18485 * config/arc/arc.c (hwloop_optimize): Likewise.
18486 * config/arm/arm.c (create_fix_barrier): Likewise.
18487 * config/s390/s390.c (s390_chunkify_start): Likewise.
18488 * config/sh/sh.c (find_barrier): Likewise.
18489 * config/i386/i386.c (rest_of_insert_endbranch,
18490 ix86_seh_fixup_eh_fallthru): Likewise.
18491 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
18492 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
18493 * config/frv/frv.c (frv_function_prologue): Likewise.
18494 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
18496 (note_outside_basic_block_p): Remove all references to
18497 NOTE_INSN_CALL_ARG_LOCATION.
18498 * gengtype.c (adjust_field_rtx_def): Likewise.
18499 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
18501 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
18502 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
18505 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
18508 PR tree-optimization/84452
18509 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
18510 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
18513 2018-02-19 Martin Liska <mliska@suse.cz>
18516 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
18518 2018-02-19 Martin Liska <mliska@suse.cz>
18519 Richard Sandiford <richard.sandiford@linaro.org>
18521 PR tree-optimization/82491
18522 * gimple-fold.c (get_base_constructor): Make earlier bail out
18525 2018-02-19 Carl Love <cel@us.ibm.com>
18527 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
18528 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
18530 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
18531 P8V_BUILTIN_VEC_NEG.
18533 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
18535 * config/rl78/rl78.md (movdf): New define expand.
18537 2018-02-19 Martin Liska <mliska@suse.cz>
18540 * doc/invoke.texi: Fix typo.
18541 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
18543 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
18545 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
18546 handle rs6000_single_float and rs6000_double_float specially for
18549 2018-02-16 Jeff Law <law@redhat.com>
18551 * config/rx/rx.c (add_pop_cfi_notes): New function.;
18552 (pop_regs): Use it.
18554 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18557 * ipa-inline.c (inline_small_functions): Fix a typo.
18559 2018-02-16 Nathan Sidwell <nathan@acm.org>
18561 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
18563 2018-02-16 Carl Love <cel@us.ibm.com>
18565 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
18566 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
18567 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
18568 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
18569 expansion to P8V_BUILTIN_VEC_FLOAT2.
18571 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
18573 PR rtl-optimization/70023
18574 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
18575 src_regno into account.
18577 2018-02-16 Carl Love <cel@us.ibm.com>
18579 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
18580 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
18581 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
18582 * config/rs6000/rs6000.c: Remove case statements for
18583 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
18584 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
18585 and P9V_BUILTIN_VEC_VINSERT4B.
18586 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
18587 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
18588 * config/rs6000/vsx.md:
18589 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
18592 2018-02-16 Carl Love <cel@us.ibm.com>
18594 * config/rs6000/altivec.h: Add builtin names vec_extract4b
18596 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
18598 * config/rs6000/rs6000-c.c: Add the definitions for
18599 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
18600 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
18601 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
18602 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
18603 definition for insert4b and define insn *insert3b_internal.
18604 * doc/extend.texi: Add documentation for vec_extract4b.
18606 2018-02-16 Nathan Sidwell <nathan@acm.org>
18608 * doc/extend.texi (Backwards Compatibility): Mention friend
18609 injection. Note for-scope is deprecated.
18610 * doc/invoke.texi (-ffriend-injection): Deprecate.
18612 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
18614 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
18615 that moved to I2, also allow destinations that are a paradoxical
18616 subreg (instead of a normal reg).
18618 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
18621 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
18624 2018-02-16 Richard Biener <rguenther@suse.de>
18626 PR tree-optimization/84037
18627 PR tree-optimization/84016
18629 * config/i386/i386.c (ix86_builtin_vectorization_cost):
18630 Adjust vec_construct for the fact we need additional higher latency
18631 128bit inserts for AVX256 and AVX512 vector builds.
18632 (ix86_add_stmt_cost): Scale vector construction cost for
18635 2018-02-16 Richard Biener <rguenther@suse.de>
18637 PR tree-optimization/84417
18638 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
18639 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
18640 (non_rewritable_lvalue_p): Likewise, use poly-ints.
18642 2018-02-16 Martin Liska <mliska@suse.cz>
18645 * internal-fn.def (ASAN_CHECK): Set proper flags.
18646 (ASAN_MARK): Likewise.
18648 2018-02-16 Julia Koval <julia.koval@intel.com>
18650 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
18651 from PTA_CANNONLAKE.
18653 2018-02-16 Jakub Jelinek <jakub@redhat.com>
18656 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
18657 Use ++iter rather than iter++ for std::list iterators.
18658 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
18659 defer deleting them until all nodes in the forest are processed. Do
18660 free even leaf nodes. Change to_process into auto_vec.
18663 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
18664 * vec.h (vec_default_construct): Use memset instead of placement new
18665 if BROKEN_VALUE_INITIALIZATION is defined.
18666 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
18667 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
18670 PR rtl-optimization/83723
18671 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
18672 * lra.c (lra_substitute_pseudo): Likewise. If true, use
18673 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
18675 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
18677 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
18679 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
18681 PR rtl-optimization/81443
18682 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
18683 from inner REGs to paradoxical SUBREGs.
18685 2018-02-16 Richard Biener <rguenther@suse.de>
18687 PR tree-optimization/84399
18688 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18689 For operands we can analyze at their definition make sure we can
18690 analyze them at each use as well.
18692 2018-02-16 Richard Biener <rguenther@suse.de>
18694 PR tree-optimization/84190
18695 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
18696 volatile accesses if the decl isn't volatile.
18698 2018-02-15 Jason Merrill <jason@redhat.com>
18700 PR c++/84314 - ICE with templates and fastcall attribute.
18701 * attribs.c (build_type_attribute_qual_variant): Don't clobber
18702 TYPE_CANONICAL on an existing type.
18704 2018-02-15 Jakub Jelinek <jakub@redhat.com>
18706 PR tree-optimization/84383
18707 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
18708 dstoff nor call operand_equal_p if dstbase is NULL.
18710 PR tree-optimization/84334
18711 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
18712 also a CONSTANT_CLASS_P, punt.
18714 2018-02-14 Jim Wilson <jimw@sifive.com>
18716 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
18717 first SMALL_OPERAND check. New local min_second_step. Move assert
18718 to where locals are set. Add TARGET_RVC support.
18719 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
18721 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
18723 * doc/invoke.texi: Correct -Wformat-overflow code sample.
18725 2018-02-14 Martin Sebor <msebor@redhat.com>
18727 PR tree-optimization/83698
18728 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
18729 arrays constrain the offset range to their bounds.
18730 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
18731 (builtin_access::overlap): Avoid setting the size of overlap if it's
18733 (maybe_diag_overlap): Also consider arrays when deciding what values
18734 of offsets to include in diagnostics.
18736 2018-02-14 Martin Sebor <msebor@redhat.com>
18739 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
18740 that correspond to the kind of a declaration.
18742 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
18745 * config/pa/pa.md: Load address of PIC label using the linkage table
18746 if the label is nonlocal.
18748 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
18750 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
18751 warning message if user requests -maltivec=be.
18752 * doc/invoke.texi: Document deprecation of -maltivec=be.
18754 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
18757 * config/rs6000/rs6000-c.c: Update definitions for
18758 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
18759 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
18761 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
18764 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
18765 add _get_ssp intrinsics. Remove argument from
18766 __builtin_ia32_rdssp[d|q].
18767 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
18768 * config/i386/i386-builtin.def: Remove argument from
18769 __builtin_ia32_rdssp[d|q].
18770 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
18771 ix86_expand_special_args_builtin for _rdssp[d|q].
18772 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
18773 Clear register before usage.
18774 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
18775 Add documentation for new _get_ssp and _inc_ssp intrinsics.
18777 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
18779 PR tree-optimization/84357
18780 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
18781 operand 1 of an ARRAY_REF too.
18783 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
18786 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
18787 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18789 (set_of_reg): New struct.
18790 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
18791 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
18792 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18794 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
18795 Split into bitclr, bitset, bitinvert patterns if appropriate.
18796 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
18797 use rx_fuse_in_memory_bitop.
18798 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
18799 to named insn, correct maximum insn length.
18801 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
18804 * machmode.def: Define a complex mode for PARTIAL_INT.
18805 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
18807 * doc/rtl.texi: Document CSPImode.
18808 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
18810 (msp430_hard_regno_nregs_with_padding): Likewise.
18812 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
18815 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
18817 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
18819 PR rtl-optimization/84169
18820 * combine.c (try_combine): New variable split_i2i3. Set it to true if
18821 we generated a parallel as new i3 and we split that to new i2 and i3
18822 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
18823 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
18824 those to i2, not i1. Partially rewrite this scan code.
18826 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18829 * stor-layout.c (place_field): For variable length fields, adjust
18830 offset_align afterwards not just based on the field's alignment,
18831 but also on the size.
18833 PR middle-end/84309
18834 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
18835 of exps and logs in the use_exp2 case.
18837 2018-02-13 Jeff Law <law@redhat.com>
18839 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
18840 entry for "vector".
18842 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
18845 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
18849 * common.opt (gas-loc-support, gas-locview-support): New.
18850 (ginline-points, ginternal-reset-location-views): New.
18851 * doc/invoke.texi: Document them. Use @itemx where intended.
18852 (gvariable-location-views): Adjust.
18853 * target.def (reset_location_view): New.
18854 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
18855 (TARGET_RESET_LOCATION_VIEW): New.
18856 * doc/tm.texi: Rebuilt.
18857 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
18858 (dwarf2out_default_as_locview_support): New.
18859 (output_asm_line_debug_info): Use option variables.
18860 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
18861 (output_loc_list): Likewise.
18862 (add_high_low_attributes): Check option variables.
18863 Don't output entry view attribute in strict mode.
18864 (gen_inlined_subroutine_die): Check option variables.
18865 (dwarf2out_inline_entry): Likewise.
18866 (init_sections_and_labels): Likewise.
18867 (dwarf2out_early_finish): Likewise.
18868 (maybe_reset_location_view): New, from...
18869 (dwarf2out_var_location): ... here. Call it.
18870 * debug.h (dwarf2out_default_as_loc_support): Declare.
18871 (dwarf2out_default_as_locview_support): Declare.
18872 * hooks.c (hook_int_rtx_insn_0): New.
18873 * hooks.h (hook_int_rtx_insn_0): Declare.
18874 * toplev.c (process_options): Take -gas-loc-support and
18875 -gas-locview-support from dwarf2out. Enable
18876 -gvariable-location-views by default only with locview
18877 assembler support. Enable -ginternal-reset-location-views by
18878 default only if the target defines the corresponding hook.
18879 Enable -ginline-points by default if location views are
18880 enabled; force it disabled if statement frontiers are
18882 * tree-inline.c (expand_call_inline): Check option variables.
18883 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
18885 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
18887 PR tree-optimization/84321
18888 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
18889 handling. Also check whether the anti-range contains any values
18890 that satisfy the mask; switch to a VR_RANGE if not.
18892 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
18895 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
18897 2018-02-13 Martin Jambor <mjambor@suse.cz>
18900 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
18901 of call statements, also set location of a load to a temporary.
18903 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
18905 * config/rl78/rl78.c (add_vector_labels): New function.
18906 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
18907 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
18908 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
18909 which checks that no arguments are passed.
18910 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
18911 * doc/extend.texi: Documentation for the new attribute.
18913 2018-02-13 Andreas Schwab <schwab@suse.de>
18915 * config/riscv/linux.h (CPP_SPEC): Define.
18917 2018-02-13 Jakub Jelinek <jakub@redhat.com>
18920 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
18921 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
18922 OPTION_MASK_ISA_AES as first argument to def_builtin_const
18923 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
18924 instead of OPTION_MASK_ISA_PCLMUL as first argument to
18925 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
18926 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
18927 temporarily for AES and PCLMUL builtins.
18929 PR tree-optimization/84339
18930 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
18931 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
18934 PR middle-end/84309
18935 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
18936 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
18937 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
18939 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
18941 * omp-simd-clone.h: New file.
18942 * omp-simd-clone.c: Include omp-simd-clone.h.
18943 (expand_simd_clones): No longer static.
18944 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
18945 cgraph.h and omp-simd-clone.h.
18946 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
18947 (vect_recog_widen_shift_pattern): Formatting fix.
18948 (vect_pattern_recog_1): Don't check optab for calls.
18951 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
18952 operands[2] into a REG before using gen_lowpart on it.
18954 2018-02-12 Jeff Law <law@redhat.com>
18957 * config/sh/sh.c (find_barrier): Consider a sibling call
18960 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
18961 successfully back substituting a reg.
18963 2018-02-12 Richard Biener <rguenther@suse.de>
18965 PR tree-optimization/84037
18966 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
18967 parameter, move visited init to caller.
18968 (vect_slp_analyze_operations): Separate cost from validity
18969 check, initialize visited once for all instances.
18970 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
18972 * tree-vect-stmts.c (vect_model_simple_cost): Make early
18974 (vect_model_promotion_demotion_cost): Likewise.
18975 (vectorizable_bswap): Guard cost modeling with !slp_node
18976 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
18978 (vectorizable_call): Likewise.
18979 (vectorizable_conversion): Likewise.
18980 (vectorizable_assignment): Likewise.
18981 (vectorizable_shift): Likewise.
18982 (vectorizable_operation): Likewise.
18983 (vectorizable_store): Likewise.
18984 (vectorizable_load): Likewise.
18985 (vectorizable_condition): Likewise.
18986 (vectorizable_comparison): Likewise.
18988 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
18991 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
18992 (ASAN_MARK): Fix fnspec to account for return value, change pointer
18993 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
18995 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
18997 PR middle-end/83665
18998 * params.def (inline-min-speedup): Increase from 8 to 15.
18999 (max-inline-insns-auto): Decrease from 40 to 30.
19000 * ipa-split.c (consider_split): Add some buffer for function to
19001 be considered inlining candidate.
19002 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
19005 2018-02-12 Richard Biener <rguenther@suse.de>
19007 PR tree-optimization/84037
19008 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
19009 matched stmts if we cannot swap the non-matched ones.
19011 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19013 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
19014 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
19015 _mm_maskz_scalef_round_ss): New intrinsics.
19016 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
19017 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
19018 __builtin_ia32_scalefss_round): Remove.
19019 (__builtin_ia32_scalefsd_mask_round,
19020 __builtin_ia32_scalefss_mask_round): New intrinsics.
19021 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
19022 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
19023 ((match_operand:VF_128 2 "<round_nimm_predicate>"
19024 "<round_constraint>")): Changed to ...
19025 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
19026 "<round_scalar_constraint>")): ... this.
19027 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
19028 %0, %1, %2<round_op3>}"): Changed to ...
19029 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
19030 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
19031 %2<round_scalar_mask_op3>}"): ... this.
19032 * config/i386/subst.md (round_scalar_nimm_predicate): New.
19034 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
19036 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
19037 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
19038 (_mm_maskz_sqrt_round_ss): New intrinsics.
19039 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
19040 (__builtin_ia32_sqrtsd_mask_round)
19041 (__builtin_ia32_sqrtss_mask_round): New builtins.
19042 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
19043 (__builtin_ia32_sqrtss_round): Remove.
19044 (__builtin_ia32_sqrtsd_mask_round)
19045 (__builtin_ia32_sqrtss_mask_round): New builtins.
19046 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
19047 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
19048 ((match_operand:VF_128 1 "vector_operand"
19049 "xBm,<round_constraint>")): Changed to ...
19050 ((match_operand:VF_128 1 "vector_operand"
19051 "xBm,<round_scalar_constraint>")): ... this.
19052 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
19053 %0, %2, %<iptr>1<round_op3>}): Changed to ...
19054 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
19055 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
19056 %<iptr>1<round_scalar_mask_op3>}): ... this.
19057 ((set_attr "prefix" "<round_prefix>")): Changed to ...
19058 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
19060 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
19063 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
19064 Cast vec_cmpeq result to correct type.
19065 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
19066 Cast vec_cmpgt result to correct type.
19068 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
19070 * final.c (final_scan_insn_1): Renamed from...
19071 (final_scan_insn): ... this. New wrapper, to recover
19072 seen from the outermost call in recursive ones.
19073 * config/sparc/sparc.c (output_return): Drop seen from call.
19074 (output_sibcall): Likewise.
19075 * config/visium/visium.c (output_branch): Likewise.
19077 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
19079 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
19082 2018-02-10 Alan Modra <amodra@gmail.com>
19085 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
19086 Specify LR as an input.
19088 2018-02-10 Jakub Jelinek <jakub@redhat.com>
19091 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
19092 remove_member_access_dummy_vars): New functions.
19093 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
19094 lower_omp_1, execute_lower_omp): Use them.
19096 PR rtl-optimization/84308
19097 * shrink-wrap.c (spread_components): Release todo vector.
19099 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
19101 PR rtl-optimization/57193
19102 * ira-color.c (struct allocno_color_data): Add member
19103 conflict_allocno_hard_prefs.
19104 (update_conflict_allocno_hard_prefs): New.
19105 (bucket_allocno_compare_func): Add a preference based on
19106 conflict_allocno_hard_prefs.
19107 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
19108 (color_allocnos): Remove a dead code. Initiate
19109 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
19111 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19114 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
19115 constraint from =wa to wa. Avoid a subreg on the output operand,
19116 instead use a pseudo and subreg it in a move.
19117 (p9_xxbrd_<mode>): Changed to ...
19118 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
19119 (p9_xxbrd_v2df): New expander.
19120 (p9_xxbrw_<mode>): Changed to ...
19121 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
19122 (p9_xxbrw_v4sf): New expander.
19124 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19126 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
19128 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
19131 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
19132 multiply in 32-bit mode.
19133 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
19134 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
19137 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
19139 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
19140 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
19141 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
19142 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
19144 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
19147 * dwarf2out.c (is_trivial_indirect_ref): New function.
19148 (dwarf2out_late_global_decl): Do not generate a location
19149 attribute for variables that have a non-trivial DECL_VALUE_EXPR
19150 and that are not defined in the current unit.
19152 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19154 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
19155 instead of a libcall for UNORDERED.
19157 2018-02-09 Tamar Christina <tamar.christina@arm.com>
19160 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
19161 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
19163 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19166 * config/s390/s390.c (s390_set_current_function): Invoke
19167 s390_indirect_branch_settings also if fndecl didn't change.
19169 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19171 * config/rs6000/rs6000.md (blockage): Set length to zero.
19173 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
19175 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
19177 2018-02-09 Jakub Jelinek <jakub@redhat.com>
19180 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
19181 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
19185 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
19186 PARALLEL incoming that failed vt_get_decl_and_offset check.
19188 PR middle-end/84237
19189 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
19190 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
19192 (get_variable_section): For decls in named .bss* sections pass true as
19193 second argument to bss_initializer_p.
19195 2018-02-09 Marek Polacek <polacek@redhat.com>
19196 Jakub Jelinek <jakub@redhat.com>
19199 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
19200 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
19201 Sync some changes from cxx_fold_indirect_ref.
19203 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
19205 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
19207 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
19208 (BLOCK_INLINE_ENTRY_LABEL): New.
19209 (dwarf2out_var_location): Disregard inline entry markers.
19210 (inline_entry_data): New struct.
19211 (inline_entry_data_hasher): New hashtable type.
19212 (inline_entry_data_hasher::hash): New.
19213 (inline_entry_data_hasher::equal): New.
19214 (inline_entry_data_table): New variable.
19215 (add_high_low_attributes): Add DW_AT_entry_pc and
19216 DW_AT_GNU_entry_view attributes if a pending entry is found
19217 in inline_entry_data_table. Add old entry_pc attribute only
19218 if debug nonbinding markers are disabled.
19219 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
19220 markers are enabled.
19221 (block_within_block_p, dwarf2out_inline_entry): New.
19222 (dwarf2out_finish): Check that no entries remained in
19223 inline_entry_data_table.
19224 * final.c (reemit_insn_block_notes): Handle inline entry notes.
19225 (final_scan_insn, notice_source_line): Likewise.
19226 (rest_of_clean_state): Skip inline entry markers.
19227 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
19229 * gimple.c (gimple_build_debug_inline_entry): New.
19230 * gimple.h (enum gimple_debug_subcode): Add
19231 GIMPLE_DEBUG_INLINE_ENTRY.
19232 (gimple_build_debug_inline_entry): Declare.
19233 (gimple_debug_inline_entry_p): New.
19234 (gimple_debug_nonbind_marker_p): Adjust.
19235 * insn-notes.def (INLINE_ENTRY): New.
19236 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
19237 inline entry marker notes.
19238 (print_insn): Likewise.
19239 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
19240 (INSN_DEBUG_MARKER_KIND): Likewise.
19241 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
19242 * tree-inline.c (expand_call_inline): Build and insert
19243 debug_inline_entry stmt.
19244 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
19245 inline entry blocks early, if nonbind markers are enabled.
19246 (dump_scope_block): Dump fragment info.
19247 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
19248 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
19249 (gimple_build_debug_inline_entry): New.
19250 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
19251 Enable/disable inline entry points too.
19252 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
19253 (DEBUG_INSN): Describe inline entry markers.
19255 * common.opt (gvariable-location-views): New.
19256 (gvariable-location-views=incompat5): New.
19257 * config.in: Rebuilt.
19258 * configure: Rebuilt.
19259 * configure.ac: Test assembler for view support.
19260 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
19261 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
19262 * dwarf2out.c (var_loc_view): New typedef.
19263 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
19264 (dwarf2out_locviews_in_attribute): New.
19265 (dwarf2out_locviews_in_loclist): New.
19266 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
19267 (enum dw_line_info_opcode): Add LI_adv_address.
19268 (struct dw_line_info_table): Add view.
19269 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
19270 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
19271 (zero_view_p): New variable.
19272 (ZERO_VIEW_P): New macro.
19273 (output_asm_line_debug_info): New.
19274 (struct var_loc_node): Add view.
19275 (add_AT_view_list, AT_loc_list): New.
19276 (add_var_loc_to_decl): Add view param. Test it against last.
19277 (new_loc_list): Add view params. Record them.
19278 (AT_loc_list_ptr): Handle loc and view lists.
19279 (view_list_to_loc_list_val_node): New.
19280 (print_dw_val): Handle dw_val_class_view_list.
19281 (size_of_die): Likewise.
19282 (value_format): Likewise.
19283 (loc_list_has_views): New.
19284 (gen_llsym): Set vl_symbol too.
19285 (maybe_gen_llsym, skip_loc_list_entry): New.
19286 (dwarf2out_maybe_output_loclist_view_pair): New.
19287 (output_loc_list): Output view list or entries too.
19288 (output_view_list_offset): New.
19289 (output_die): Handle dw_val_class_view_list.
19290 (output_dwarf_version): New.
19291 (output_compilation_unit_header): Use it.
19292 (output_skeleton_debug_sections): Likewise.
19293 (output_rnglists, output_line_info): Likewise.
19294 (output_pubnames, output_aranges): Update version comments.
19295 (output_one_line_info_table): Output view numbers in asm comments.
19296 (dw_loc_list): Determine current endview, pass it to new_loc_list.
19297 Call maybe_gen_llsym.
19298 (loc_list_from_tree_1): Adjust.
19299 (add_AT_location_description): Create view list attribute if
19300 needed, check it's absent otherwise.
19301 (convert_cfa_to_fb_loc_list): Adjust.
19302 (maybe_emit_file): Call output_asm_line_debug_info for test.
19303 (dwarf2out_var_location): Reset views as needed. Precompute
19304 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
19305 attribute. Set view.
19306 (new_line_info_table): Reset next view.
19307 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
19308 (dwarf2out_source_line): Likewise. Output view resets and labels to
19309 the assembler, or select appropriate line info opcodes.
19310 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
19311 (optimize_string_length): Catch it. Adjust.
19312 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
19313 dw_val_class_view_list, and remove it if no longer needed.
19314 (hash_loc_list): Hash view numbers.
19315 (loc_list_hasher::equal): Compare them.
19316 (optimize_location_lists): Check whether a view list symbol is
19317 needed, and whether the locview attribute is present, and
19318 whether they match. Remove the locview attribute if no longer
19320 (index_location_lists): Call skip_loc_list_entry for test.
19321 (dwarf2out_finish): Call output_asm_line_debug_info for test.
19322 Use output_dwarf_version.
19323 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
19324 (struct dw_val_node): Add val_view_list.
19325 * final.c (SEEN_NEXT_VIEW): New.
19326 (set_next_view_needed): New.
19327 (clear_next_view_needed): New.
19328 (maybe_output_next_view): New.
19329 (final_start_function): Rename to...
19330 (final_start_function_1): ... this. Take pointer to FIRST,
19331 add SEEN parameter. Emit param bindings in the initial view.
19332 (final_start_function): Reintroduce SEEN-less interface.
19333 (final): Rename to...
19334 (final_1): ... this. Take SEEN parameter. Output final pending
19335 next view at the end.
19336 (final): Reintroduce seen-less interface.
19337 (final_scan_insn): Output pending next view before switching
19338 sections or ending a block. Mark the next view as needed when
19339 outputting variable locations. Notify debug backend of section
19340 changes, and of location view changes.
19341 (rest_of_handle_final): Adjust.
19342 * toplev.c (process_options): Autodetect value for debug variable
19343 location views option. Warn on incompat5 without -gdwarf-5.
19344 * doc/invoke.texi (gvariable-location-views): New.
19345 (gvariable-location-views=incompat5): New.
19346 (gno-variable-location-views): New.
19348 2018-02-08 David Malcolm <dmalcolm@redhat.com>
19350 PR tree-optimization/84136
19351 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
19352 that the result of find_edge is non-NULL.
19354 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19357 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
19358 storing integer register in SImode. Fix cost of 256 and 512
19359 byte aligned SSE register store.
19361 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
19363 * config/i386/i386.c (ix86_multiplication_cost): Fix
19364 multiplication cost for TARGET_AVX512DQ.
19366 2018-02-08 Marek Polacek <polacek@redhat.com>
19368 PR tree-optimization/84238
19369 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
19372 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19374 PR tree-optimization/84265
19375 * tree-vect-stmts.c (vectorizable_store): Don't treat
19376 VMAT_CONTIGUOUS accesses as grouped.
19377 (vectorizable_load): Likewise.
19379 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19381 PR tree-optimization/81635
19382 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
19383 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
19384 (test_round_for_mask): New functions.
19385 (wide_int_cc_tests): Call test_round_for_mask.
19386 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
19387 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
19388 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
19389 range returned by get_range_info.
19391 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
19394 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
19395 * symtab.c: Include builtins.h
19396 (symtab_node::output_to_lto_symbol_table_p): Move here
19397 from lto-streamer-out.c:output_symbol_p.
19398 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
19399 (output_symbol_p): Move all logic to symtab.c
19400 (produce_symtab): Update.
19402 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
19404 * config/s390/s390-opts.h (enum indirect_branch): Define.
19405 * config/s390/s390-protos.h (s390_return_addr_from_memory)
19406 (s390_indirect_branch_via_thunk)
19407 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
19408 (enum s390_indirect_branch_type): Define.
19409 * config/s390/s390.c (struct s390_frame_layout, struct
19410 machine_function): Remove.
19411 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
19412 (indirect_branch_table_label_no, indirect_branch_table_name):
19414 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
19415 (enum s390_indirect_branch_option): Define.
19416 (s390_return_addr_from_memory): New function.
19417 (s390_handle_string_attribute): New function.
19418 (s390_attribute_table): Add new attribute handler.
19419 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
19420 (s390_indirect_branch_via_thunk): New function.
19421 (s390_indirect_branch_via_inline_thunk): New function.
19422 (s390_function_ok_for_sibcall): When jumping via thunk disallow
19423 sibling call optimization for non z10 compiles.
19424 (s390_emit_call): Force indirect branch target to be a single
19425 register. Add r1 clobber for non-z10 compiles.
19426 (s390_emit_epilogue): Emit return jump via return_use expander.
19427 (s390_reorg): Handle JUMP_INSNs as execute targets.
19428 (s390_option_override_internal): Perform validity checks for the
19429 new command line options.
19430 (s390_indirect_branch_attrvalue): New function.
19431 (s390_indirect_branch_settings): New function.
19432 (s390_set_current_function): Invoke s390_indirect_branch_settings.
19433 (s390_output_indirect_thunk_function): New function.
19434 (s390_code_end): Implement target hook.
19435 (s390_case_values_threshold): Implement target hook.
19436 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
19438 * config/s390/s390.h (struct s390_frame_layout)
19439 (struct machine_function): Move here from s390.c.
19440 (TARGET_INDIRECT_BRANCH_NOBP_RET)
19441 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
19442 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
19443 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
19444 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
19445 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
19446 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
19447 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
19448 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
19449 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
19450 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
19451 (mnemonic attribute): Add values which aren't recognized
19453 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
19454 pattern for branch conversion. Fix mnemonic attribute.
19455 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
19456 indirect branch via thunk if requested.
19457 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
19458 ("*indirect_jump"): Disable for branch conversion using out of
19460 ("indirect_jump_via_thunk<mode>_z10")
19461 ("indirect_jump_via_thunk<mode>")
19462 ("indirect_jump_via_inlinethunk<mode>_z10")
19463 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
19464 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
19465 ("casesi_jump_via_inlinethunk<mode>_z10")
19466 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
19467 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
19468 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
19469 ("*indirect2_jump"): Disable for branch conversion.
19470 ("casesi_jump"): Turn into expander and expand patterns for branch
19472 ("return_use"): New expander.
19473 ("*return"): Emit return via thunk and rename it to ...
19474 ("*return<mode>"): ... this one.
19475 * config/s390/s390.opt: Add new options and and enum for the
19478 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
19480 * lra-constraints.c (match_reload): Unconditionally use
19481 gen_lowpart_SUBREG, rather than selecting between that
19482 and equivalent gen_rtx_SUBREG code.
19484 2018-02-08 Richard Biener <rguenther@suse.de>
19486 PR tree-optimization/84233
19487 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
19488 changed flag instead of boguously re-using phi_inserted.
19490 2018-02-08 Martin Jambor <mjambor@suse.cz>
19492 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
19493 static local variables.
19495 2018-02-08 Richard Biener <rguenther@suse.de>
19497 PR tree-optimization/84278
19498 * tree-vect-stmts.c (vectorizable_store): When looking for
19499 smaller vector types to perform grouped strided loads/stores
19500 make sure the mode is supported by the target.
19501 (vectorizable_load): Likewise.
19503 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19505 * config/aarch64/aarch64.c (aarch64_components_for_bb):
19506 Increase LDP/STP opportunities by adding adjacent callee-saves.
19508 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
19510 PR rtl-optimization/84068
19511 PR rtl-optimization/83459
19512 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
19514 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
19516 PR tree-optimization/84224
19517 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
19518 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
19519 non-zero arguments.
19521 2018-02-07 Iain Sandoe <iain@codesourcery.com>
19524 * config/rs6000/altivec.md (*restore_world): Remove LR use.
19525 * config/rs6000/predicates.md (restore_world_operation): Adjust op
19526 count, remove one USE.
19528 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
19530 * doc/install.texi (Configuration): Document the
19531 --with-long-double-format={ibm,ieee} PowerPC configuration
19535 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
19536 Convert from define_expand to be define_insn_and_split. Rework
19537 float/double/_Float128 conversions to QI/HI/SImode to work with
19538 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
19539 conversions to QI/HImode types did a store and then a load to
19540 truncate the value. For conversions to VSX registers, don't split
19541 the insn, instead emit the code directly. Use the code iterator
19542 any_fix to combine signed and unsigned conversions.
19543 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
19544 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
19545 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19546 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
19547 (fix_<mode>di2_hw): Likewise.
19548 (fixuns_<mode>di2_hw): Likewise.
19549 (fix_<mode>si2_hw): Likewise.
19550 (fixuns_<mode>si2_hw): Likewise.
19551 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
19552 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19553 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
19554 fix<uns>_trunc<SFDF:mode>si2_p8.
19555 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
19556 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
19557 (fix<uns>_<mode>_mem): Likewise.
19558 (fctiw<u>z_<mode>_mem): Likewise.
19559 (fix<uns>_<mode>_mem): Likewise.
19560 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
19561 the register allocator from doing a direct move to the GPRs to do
19562 a store, and instead use the ISA 3.0 store byte/half-word from
19563 vector register instruction. For IEEE 128-bit floating point,
19564 also optimize stores of 32-bit ints.
19565 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
19567 2018-02-07 Alan Hayward <alan.hayward@arm.com>
19569 * genextract.c (push_pathstr_operand): New function to support
19571 (walk_rtx): Call push_pathstr_operand.
19572 (print_path): Support [a-zA-Z].
19574 2018-02-07 Richard Biener <rguenther@suse.de>
19576 PR tree-optimization/84037
19577 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
19578 (cse_and_gimplify_to_preheader): Declare.
19579 (vect_get_place_in_interleaving_chain): Likewise.
19580 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19582 (_loop_vec_info::~_loop_vec_info): Delete it.
19583 (cse_and_gimplify_to_preheader): New function.
19584 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
19585 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
19586 (vectorizable_load): Likewise. For grouped stores always base
19587 the IV on the first element.
19588 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
19589 condition before gimplifying.
19591 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19593 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
19594 *DIV_EXPR and *MOD_EXPR.
19596 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
19599 * config/i386/i386.c (ix86_option_override_internal): Mask out
19600 the CF_SET bit when checking -fcf-protection.
19602 2018-02-07 Tom de Vries <tom@codesourcery.com>
19605 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
19608 2018-02-07 Richard Biener <rguenther@suse.de>
19610 PR tree-optimization/84204
19611 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
19614 PR tree-optimization/84205
19615 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
19616 special-case isl_ast_op_zdiv_r.
19618 PR tree-optimization/84223
19619 * graphite-scop-detection.c (gather_bbs::before_dom_children):
19620 Only add conditions from within the region.
19621 (gather_bbs::after_dom_children): Adjust.
19623 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
19626 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
19627 * config/avr/avr.md: Only post-reload split REG-REG moves if
19628 either register is GENERAL_REG_P.
19630 2018-02-07 Jakub Jelinek <jakub@redhat.com>
19632 PR tree-optimization/84235
19633 * tree-ssa-scopedtables.c
19634 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
19635 if the subtraction is performed in floating point type where NaNs are
19636 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
19637 build 1. Formatting fix.
19639 2018-02-06 Jakub Jelinek <jakub@redhat.com>
19642 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
19643 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
19644 and skip it regardless of bb boundaries. Use CALL_P macro,
19645 don't test INSN_P (insn) together with CALL_P or JUMP_P check
19646 unnecessarily, formatting fix.
19648 2018-02-06 Michael Collison <michael.collison@arm.com>
19650 * config/arm/thumb2.md:
19651 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
19652 (*thumb_mov_notscc): Ditto.
19654 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
19657 * config/rs6000/rs6000.md (su code attribute): Use "u" for
19658 unsigned_fix, not "s".
19660 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19662 * configure.ac (gcc_fn_eh_frame_ro): New function.
19663 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
19664 correct .eh_frame permissions.
19665 * configure: Regenerate.
19667 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
19669 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
19670 irrelevant options.
19672 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19674 * config/rs6000/rs6000.c (rs6000_option_override_internal):
19675 Display warning message for -mno-speculate-indirect-jumps.
19677 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
19679 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
19681 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
19683 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
19685 PR tree-optimization/84225
19686 * tree-eh.c (find_trapping_overflow): Only call
19687 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
19689 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19692 * config/i386/i386.c: Reimplement the check of possible options
19693 -mibt/-mshstk conbination. Change error messages.
19694 * doc/invoke.texi: Fix a typo: remove extra '='.
19696 2018-02-06 Marek Polacek <polacek@redhat.com>
19698 PR tree-optimization/84228
19699 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
19701 2018-02-06 Tamar Christina <tamar.christina@arm.com>
19704 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
19705 emitted arch directives.
19706 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
19707 __ARM_FEATURE_COPROC before changing architectures.
19709 2018-02-06 Richard Biener <rguenther@suse.de>
19711 * config/i386/i386.c (print_reg): Fix typo.
19712 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
19714 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
19716 * configure: Regenerate.
19718 2018-02-05 Martin Sebor <msebor@redhat.com>
19720 PR tree-optimization/83369
19721 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
19724 2018-02-05 Martin Liska <mliska@suse.cz>
19726 * doc/invoke.texi: Cherry-pick upstream r323995.
19728 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
19730 * ira.c (ira_init_register_move_cost): Adjust comment.
19732 2018-02-05 Martin Liska <mliska@suse.cz>
19734 PR gcov-profile/84137
19735 * doc/gcov.texi: Fix typo in documentation.
19737 2018-02-05 Martin Liska <mliska@suse.cz>
19739 PR gcov-profile/83879
19740 * doc/gcov.texi: Document necessity of --dynamic-list-data when
19741 using dlopen functionality.
19743 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
19745 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
19746 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
19747 _mm_maskz_range_ss, _mm_mask_range_round_ss,
19748 _mm_maskz_range_round_ss): New intrinsics.
19749 (__builtin_ia32_rangesd128_round)
19750 (__builtin_ia32_rangess128_round): Remove.
19751 (__builtin_ia32_rangesd128_mask_round,
19752 __builtin_ia32_rangess128_mask_round): New builtins.
19753 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
19754 __builtin_ia32_rangess128_round): Remove.
19755 (__builtin_ia32_rangesd128_mask_round,
19756 __builtin_ia32_rangess128_mask_round): New builtins.
19757 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
19758 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
19759 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
19760 "<round_saeonly_constraint>")): Changed to ...
19761 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
19762 "<round_saeonly_scalar_constraint>")): ... this.
19763 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
19764 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
19765 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
19766 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
19767 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
19769 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
19771 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
19773 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
19774 Remove all values except native, 8540 and 8548.
19776 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
19778 * config/i386/i386.c (ix86_output_function_return): Pass
19779 INVALID_REGNUM, instead of -1, as invalid register number to
19780 indirect_thunk_name and output_indirect_thunk.
19782 2018-02-02 Julia Koval <julia.koval@intel.com>
19784 * config.gcc: Add -march=icelake.
19785 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
19786 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
19787 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
19788 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
19789 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
19790 (processor_target_table): Add icelake.
19791 (ix86_option_override_internal): Handle new PTAs.
19792 (get_builtin_code_for_version): Handle icelake.
19793 (M_INTEL_COREI7_ICELAKE): New.
19794 (fold_builtin_cpu): Handle icelake.
19795 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
19796 * doc/invoke.texi: Add -march=icelake.
19798 2018-02-02 Julia Koval <julia.koval@intel.com>
19800 * config/i386/i386.c (ix86_option_override_internal): Change flags type
19801 to wide_int_bitmask.
19802 * wide-int-bitmask.h: New.
19804 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
19807 * config/i386/i386.md: Replace Pmode with word_mode in
19808 builtin_setjmp_setup and builtin_longjmp to support x32.
19810 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
19814 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
19816 (rs6000_supported_cpu_names): New static variable.
19817 (linux_cpu_translation_table): Likewise.
19818 (elf_platform) <cpu>: Define new static variable and use it.
19819 Translate kernel AT_PLATFORM name to canonical name if needed.
19820 Error if platform name is unknown.
19822 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
19825 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
19827 2018-02-01 Jeff Law <law@redhat.com>
19830 * config/i386/i386.c (release_scratch_register_on_entry): Add new
19831 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
19832 the scratch if RELEASE_VIA_POP is false.
19833 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
19834 If we have to save a temporary register, decrement SIZE appropriately.
19835 Pass new arguments to release_scratch_register_on_entry.
19836 (ix86_adjust_stack_and_probe): Likewise.
19837 (ix86_emit_probe_stack_range): Pass new arguments to
19838 release_scratch_register_on_entry.
19840 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
19842 PR rtl-optimization/84157
19843 * combine.c (change_zero_ext): Use REG_P predicate in
19844 front of HARD_REGISTER_P predicate.
19846 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
19848 * config/avr/avr.c (avr_option_override): Move disabling of
19849 -fdelete-null-pointer-checks to...
19850 * common/config/avr/avr-common.c (avr_option_optimization_table):
19853 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19855 PR tree-optimization/81635
19856 * tree-data-ref.c (split_constant_offset_1): For types that
19857 wrap on overflow, try to use range info to prove that wrapping
19860 2018-02-01 Renlin Li <renlin.li@arm.com>
19863 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
19864 TAILCALL_ADDR_REGS.
19865 (aarch64_register_move_cost): Likewise.
19866 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
19867 TAILCALL_ADDR_REGS.
19868 (REG_CLASS_NAMES): Likewise.
19869 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
19870 TAILCALL_ADDR_REGS. Remove IP registers.
19871 * config/aarch64/aarch64.md (Ucs): Update register constraint.
19873 2018-02-01 Richard Biener <rguenther@suse.de>
19875 * domwalk.h (dom_walker::dom_walker): Add additional constructor
19876 for specifying RPO order and allow NULL for that.
19877 * domwalk.c (dom_walker::dom_walker): Likewise.
19878 (dom_walker::walk): Handle NULL RPO order.
19879 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
19881 (rewrite_update_dom_walker): Likewise.
19882 (mark_def_dom_walker): Likewise.
19884 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19886 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
19887 (aarch64_maybe_expand_sve_subreg_move): Declare.
19888 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
19889 * config/aarch64/predicates.md (aarch64_any_register_operand): New
19891 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
19892 that are semantically a reverse operation.
19893 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
19894 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
19895 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
19897 (aarch64_can_change_mode_class): For big-endian, forbid changes
19898 between two SVE modes if they have different element sizes.
19900 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19902 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
19903 the TImode handling for big-endian targets.
19905 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19907 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
19908 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
19910 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
19911 Remove BSWAP handing for big-endian targets and use the form of
19912 LD1RQ appropariate for the mode.
19914 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19916 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
19917 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
19918 duplicated element.
19920 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
19923 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
19924 check for operands that need to go through aarch64_sve_reload_be.
19926 2018-02-01 Jakub Jelinek <jakub@redhat.com>
19928 PR tree-optimization/81661
19929 PR tree-optimization/84117
19930 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
19931 * tree-eh.c: Include gimplify.h.
19932 (find_trapping_overflow, replace_trapping_overflow,
19933 rewrite_to_non_trapping_overflow): New functions.
19934 * tree-vect-loop.c: Include tree-eh.h.
19935 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
19936 * tree-data-ref.c: Include tree-eh.h.
19937 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
19939 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
19941 PR rtl-optimization/84123
19942 * combine.c (change_zero_ext): Check if hard register satisfies
19943 can_change_dest_mode before calling gen_lowpart_SUBREG.
19945 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
19948 * ira.c (ira_init_register_move_cost): Remove assert.
19950 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19952 PR rtl-optimization/84071
19953 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
19954 * doc/tm.texi: Regenerate.
19956 2018-01-31 Richard Biener <rguenther@suse.de>
19958 PR tree-optimization/84132
19959 * tree-data-ref.c (analyze_miv_subscript): Properly
19960 check whether evolution_function_is_affine_multivariate_p
19961 before calling gcd_of_steps_may_divide_p.
19963 2018-01-31 Julia Koval <julia.koval@intel.com>
19966 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
19967 * config/i386/i386.md (rdpid_rex64) New.
19968 (rdpid): Make 32bit only.
19970 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
19973 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
19974 an IDENTIFIER_NODE for FUNCTION_TYPE's.
19976 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19979 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19981 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
19983 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
19985 PR rtl-optimization/84071
19986 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
19987 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
19989 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
19991 * config/arc/arc.c (arc_handle_aux_attribute): New function.
19992 (arc_attribute_table): Add 'aux' attribute.
19993 (arc_in_small_data_p): Consider aux like variables.
19994 (arc_is_aux_reg_p): New function.
19995 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
19996 (arc_get_aux_arg): New function.
19997 (prepare_move_operands): Handle aux-register access.
19998 (arc_handle_aux_attribute): New function.
19999 * doc/extend.texi (ARC Variable attributes): Add subsection.
20001 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
20003 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
20004 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
20005 (arc_attribute_table): Add 'uncached' attribute.
20006 (arc_print_operand): Print '.di' flag for uncached memory
20008 (arc_in_small_data_p): Do not consider for small data the uncached
20010 (arc_is_uncached_mem_p): New function.
20011 * config/arc/predicates.md (compact_store_memory_operand): Check
20012 for uncached memory accesses.
20013 (nonvol_nonimm_operand): Likewise.
20014 * doc/extend.texi (ARC Type Attribute): New subsection.
20016 2018-01-31 Jakub Jelinek <jakub@redhat.com>
20019 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
20020 falign-loops=): Add Optimization flag.
20022 2018-01-30 Jeff Law <law@redhat.com>
20025 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
20026 INT_REGISTERS_SAVED. Check it prior to calling
20027 get_scratch_register_on_entry.
20028 (ix86_adjust_stack_and_probe): Similarly.
20029 (ix86_emit_probe_stack_range): Similarly.
20030 (ix86_expand_prologue): Corresponding changes.
20032 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20035 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
20036 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
20038 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
20041 * lra-constraints.c (curr_insn_transform): Process AND in the
20044 2018-01-30 Jakub Jelinek <jakub@redhat.com>
20046 PR rtl-optimization/83986
20047 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
20048 dependence against last_pending_memory_flush in addition to
20049 pending_jump_insns.
20051 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
20053 PR tree-optimization/81611
20054 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
20057 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20060 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
20063 2018-01-30 Richard Biener <rguenther@suse.de>
20064 Jakub Jelinek <jakub@redhat.com>
20066 PR tree-optimization/84111
20067 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
20068 inner loops added during recursion, as they don't have up-to-date
20071 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20074 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
20075 (can_inline_edge_by_limits_p): ... here.
20076 (can_early_inline_edge_p, check_callers,
20077 update_caller_keys, update_callee_keys, recursive_inlining,
20078 add_new_edges_to_heap, speculation_useful_p,
20079 inline_small_functions,
20080 inline_small_functions, flatten_function,
20081 inline_to_all_callers_1): Update.
20083 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
20085 * profile-count.c (profile_count::combine_with_ipa_count): Handle
20088 2018-01-30 Richard Biener <rguenther@suse.de>
20090 PR tree-optimization/83008
20091 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
20092 invariant and constant vector uses in stmts when they need
20093 more than one stmt.
20095 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20098 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
20099 * configure: Regenerate.
20101 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20103 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
20105 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
20106 Use gen_rtx_REG rather than gen_lowpart.
20108 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20110 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
20111 rather than 0 when creating partial subregs.
20113 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
20115 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
20118 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
20121 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
20122 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
20123 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
20124 flags. This restores the settings used before the 2017-07-24.
20125 Turning off pre increment/decrement/modify allows IVOPTS to
20126 optimize DF/SF loops where the index is an int.
20128 2018-01-29 Richard Biener <rguenther@suse.de>
20129 Kelvin Nilsen <kelvin@gcc.gnu.org>
20132 * tree-vect-stmts.c (vectorizable_call): Don't call
20133 targetm.vectorize_builtin_md_vectorized_function if callee is
20136 2018-01-22 Carl Love <cel@us.ibm.com>
20138 * doc/extend.tex: Fix typo in second arg in
20139 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
20141 2018-01-29 Richard Biener <rguenther@suse.de>
20143 PR tree-optimization/84086
20144 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
20145 (flush_ssaname_freelist): When SSA names were released reset
20146 the SCEV hash table.
20148 2018-01-29 Richard Biener <rguenther@suse.de>
20150 PR tree-optimization/84057
20151 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
20152 removed paths when removing edges.
20154 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
20156 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
20157 -mfunction-return=@var{choice}.
20159 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
20161 PR diagnostic/84034
20162 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
20163 Handle CR like TAB.
20164 (layout::print_source_line): Likewise.
20165 (test_get_line_width_without_trailing_whitespace): Add test cases.
20167 2018-01-27 Jakub Jelinek <jakub@redhat.com>
20169 PR middle-end/84040
20170 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
20173 2018-01-26 Jim Wilson <jimw@sifive.com>
20175 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
20177 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
20180 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20182 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
20183 and CMP + SUB-immediate -> SUBS.
20185 2018-01-26 Martin Sebor <msebor@redhat.com>
20187 PR tree-optimization/83896
20188 * tree-ssa-strlen.c (get_string_len): Rename...
20189 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
20190 Avoid assuming length is constant.
20191 (handle_char_store): Use HOST_WIDE_INT for string length.
20193 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
20196 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
20197 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
20199 2018-01-26 Richard Biener <rguenther@suse.de>
20201 PR rtl-optimization/84003
20202 * dse.c (record_store): Only record redundant stores when
20203 the earlier store aliases at least all accesses the later one does.
20205 2018-01-26 Jakub Jelinek <jakub@redhat.com>
20207 PR rtl-optimization/83985
20208 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
20209 REG_CFA_RESTORE insns.
20210 (delete_unmarked_insns): Don't ignore separate shrink wrapping
20211 REG_CFA_RESTORE insns here.
20214 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
20215 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
20217 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20219 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
20220 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
20221 (arc_init): Likewise.
20222 (arc_override_options): Likewise.
20223 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
20225 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
20227 * config/arc/arc.h (TARGET_DBNZ): Define.
20228 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
20229 properly set the tune attribute.
20230 (dbnz): Use TARGET_DBNZ guard.
20231 * config/arc/arc.opt (mtune): Add core3 option.
20233 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20235 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
20236 recognize new pic like addresses.
20237 (arc_delegitimize_address): Clean up.
20239 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20241 * config/arc/arc-arches.def: Option mrf16 valid for all
20243 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
20244 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
20245 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
20246 * config/arc/arc-tables.opt: Regenerate.
20247 * config/arc/arc.c (arc_conditional_register_usage): Handle
20248 reduced register file case.
20249 (arc_file_start): Set must have build attributes.
20250 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
20251 mrf16 option value.
20252 * config/arc/arc.opt (mrf16): Add new option.
20253 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
20254 * config/arc/genmultilib.awk: Handle new mrf16 option.
20255 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
20256 * config/arc/t-multilib: Regenerate.
20257 * doc/invoke.texi (ARC Options): Document mrf16 option.
20259 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20261 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
20262 * config/arc/arc.c (arc_handle_secure_attribute): New function.
20263 (arc_attribute_table): Add 'secure_call' attribute.
20264 (arc_print_operand): Print secure call operand.
20265 (arc_function_ok_for_sibcall): Don't optimize tail calls when
20267 (arc_is_secure_call_p): New function. * config/arc/arc.md
20268 (call_i): Add support for sjli instruction.
20269 (call_value_i): Likewise.
20270 * config/arc/constraints.md (Csc): New constraint.
20272 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
20273 John Eric Martin <John.Martin@emmicro-us.com>
20275 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
20276 * config/arc/arc.c (_arc_jli_section): New struct.
20277 (arc_jli_section): New type.
20278 (rc_jli_sections): New static variable.
20279 (arc_handle_jli_attribute): New function.
20280 (arc_attribute_table): Add jli_always and jli_fixed attribute.
20281 (arc_file_end): New function.
20282 (TARGET_ASM_FILE_END): Define.
20283 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
20284 (arc_add_jli_section): New function.
20285 (jli_call_scan): Likewise.
20286 (arc_reorg): Call jli_call_scan.
20287 (arc_output_addsi): Remove 'S' from printing asm operand.
20288 (arc_is_jli_call_p): New function.
20289 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
20291 (movhi_insn): Likewise.
20292 (movsi_insn): Likewise.
20293 (movsi_set_cc_insn): Likewise.
20294 (loadqi_update): Likewise.
20295 (load_zeroextendqisi_update): Likewise.
20296 (load_signextendqisi_update): Likewise.
20297 (loadhi_update): Likewise.
20298 (load_zeroextendhisi_update): Likewise.
20299 (load_signextendhisi_update): Likewise.
20300 (loadsi_update): Likewise.
20301 (loadsf_update): Likewise.
20302 (movsicc_insn): Likewise.
20303 (bset_insn): Likewise.
20304 (bxor_insn): Likewise.
20305 (bclr_insn): Likewise.
20306 (bmsk_insn): Likewise.
20307 (bicsi3_insn): Likewise.
20308 (cmpsi_cc_c_insn): Likewise.
20309 (movsi_ne): Likewise.
20310 (movsi_cond_exec): Likewise.
20311 (clrsbsi2): Likewise.
20312 (norm_f): Likewise.
20322 (call_i): Remove 'S' asm letter, add jli instruction.
20323 (call_value_i): Likewise.
20324 * config/arc/arc.op (mjli-always): New option.
20325 * config/arc/constraints.md (Cji): New constraint.
20326 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
20328 (subsf3_fpx): Likewise.
20329 (mulsf3_fpx): Likewise.
20330 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
20332 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
20333 function attrbutes.
20334 * doc/invoke.texi (ARC): Document mjli-always option.
20336 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
20338 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
20339 avoid addition with 0 and use incw and decw where possible.
20341 2018-01-26 Richard Biener <rguenther@suse.de>
20343 PR tree-optimization/81082
20344 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
20345 association if it requires casting to unsigned.
20346 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
20347 from fold_plusminus_mult_expr to catch important cases late when
20348 range info is available.
20350 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20352 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
20353 * configure.ac (hidden_linkonce): New test.
20354 * configure: Regenerate.
20355 * config.in: Regenerate.
20357 2018-01-26 Julia Koval <julia.koval@intel.com>
20359 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
20360 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
20361 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
20362 _mm_mask_bitshuffle_epi64_mask): Fix type.
20363 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
20364 USI_FTYPE_V4DI_V4DI_USI): Remove.
20365 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
20366 __builtin_ia32_vpshufbitqmb256_mask,
20367 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
20368 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
20369 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
20371 2018-01-26 Alan Modra <amodra@gmail.com>
20374 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
20375 UNSPEC_VBPERMQ. Sort other unspecs.
20377 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
20379 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
20381 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
20383 PR middle-end/83055
20384 * predict.c (drop_profile): Do not push/pop cfun; update also
20386 (handle_missing_profiles): Fix logic looking for zero profiles.
20388 2018-01-25 Jakub Jelinek <jakub@redhat.com>
20390 PR middle-end/83977
20391 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
20392 on functions with #pragma omp declare simd or functions with simd
20394 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
20395 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
20396 Remove trailing \n from warning_at calls.
20398 2018-01-25 Tom de Vries <tom@codesourcery.com>
20401 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20402 for neutered workers.
20404 2018-01-24 Joseph Myers <joseph@codesourcery.com>
20407 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
20408 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
20410 2018-01-24 Jeff Law <law@redhat.com>
20413 * i386.c (get_probe_interval): Move to earlier point.
20414 (ix86_compute_frame_layout): If -fstack-clash-protection and
20415 the frame is larger than the probe interval, then use pushes
20416 to save registers rather than reg->mem moves.
20417 (ix86_expand_prologue): Remove conditional for int_registers_saved
20420 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
20423 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
20424 min/max for never referenced object.
20426 2018-01-24 Jakub Jelinek <jakub@redhat.com>
20428 PR middle-end/83977
20429 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
20431 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
20432 attributes from DECL_ATTRIBUTES (decl) without affecting
20433 DECL_ATTRIBUTES (current_function_decl).
20434 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
20435 functions with non-NULL DECL_ABSTRACT_ORIGIN.
20437 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
20439 PR tree-optimization/83979
20440 * fold-const.c (fold_comparison): Use constant_boolean_node
20441 instead of boolean_{true,false}_node.
20443 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
20445 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
20448 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20450 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20451 Simplify the clause that sets the length attribute.
20452 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20453 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
20454 clause that sets the length attribute.
20455 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20457 2018-01-24 Tom de Vries <tom@codesourcery.com>
20460 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
20461 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
20462 Add strict parameter.
20463 (prevent_branch_around_nothing): Insert dummy insn between branch to
20464 label and label with no ptx insn inbetween.
20465 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
20467 2018-01-24 Tom de Vries <tom@codesourcery.com>
20470 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20471 for neutered threads in warp.
20472 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
20474 2018-01-24 Richard Biener <rguenther@suse.de>
20476 PR tree-optimization/83176
20477 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
20480 2018-01-24 Richard Biener <rguenther@suse.de>
20482 PR tree-optimization/82819
20483 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
20484 code generating pluses that are no-ops in the target precision.
20486 2018-01-24 Richard Biener <rguenther@suse.de>
20488 PR middle-end/84000
20489 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
20491 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20493 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
20494 to merge probabilities.
20495 * predict.c (probably_never_executed): Also mark as cold functions
20496 with global 0 profile and guessed local profile.
20497 * profile-count.c (profile_probability::combine_with_count): New
20499 * profile-count.h (profile_probability::operator*,
20500 profile_probability::operator*=, profile_probability::operator/,
20501 profile_probability::operator/=): Reduce precision to adjusted
20502 and set value to guessed on contradictory divisions.
20503 (profile_probability::combine_with_freq): Remove.
20504 (profile_probability::combine_wiht_count): Declare.
20505 (profile_count::force_nonzero):: Set to adjusted.
20506 (profile_count::probability_in):: Set quality to adjusted.
20507 * tree-ssa-tail-merge.c (replace_block_by): Use
20508 combine_with_count.
20510 2018-01-23 Andrew Waterman <andrew@sifive.com>
20511 Jim Wilson <jimw@sifive.com>
20513 * config/riscv/riscv.c (riscv_stack_boundary): New.
20514 (riscv_option_override): Set riscv_stack_boundary. Handle
20515 riscv_preferred_stack_boundary_arg.
20516 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
20517 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
20518 (STACK_BOUNDARY): Set to riscv_stack_boundary.
20519 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
20520 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
20521 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
20523 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
20526 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
20527 of struct ix86_frame.
20528 (ix86_expand_epilogue): Likewise. Add a local variable for
20529 the reg_save_offset field in struct ix86_frame.
20531 2018-01-23 Bin Cheng <bin.cheng@arm.com>
20533 PR tree-optimization/82604
20534 * tree-loop-distribution.c (enum partition_kind): New enum item
20535 PKIND_PARTIAL_MEMSET.
20536 (partition_builtin_p): Support above new enum item.
20537 (generate_code_for_partition): Ditto.
20538 (compute_access_range): Differentiate cases that equality can be
20539 proven at all loops, the innermost loops or no loops.
20540 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
20541 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
20542 (finalize_partitions, distribute_loop): Don't fuse partition of
20543 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
20544 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
20545 parloop is enabled.
20547 2018-01-23 Martin Liska <mliska@suse.cz>
20549 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
20550 order to ignore the predictor.
20551 (PRED_POLYMORPHIC_CALL): Likewise.
20552 (PRED_RECURSIVE_CALL): Likewise.
20554 2018-01-23 Martin Liska <mliska@suse.cz>
20556 * tree-profile.c (tree_profiling): Print function header to
20557 aware reader which function we are working on.
20558 * value-prof.c (gimple_find_values_to_profile): Do not print
20559 not interesting value histograms.
20561 2018-01-23 Martin Liska <mliska@suse.cz>
20563 * profile-count.h (enum profile_quality): Add
20564 profile_uninitialized as the first value. Do not number values
20565 as they are zero based.
20566 (profile_count::verify): Update sanity check.
20567 (profile_probability::verify): Likewise.
20569 2018-01-23 Nathan Sidwell <nathan@acm.org>
20571 * doc/invoke.texi (ffor-scope): Deprecate.
20573 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20575 PR tree-optimization/83510
20576 * domwalk.c (set_all_edges_as_executable): New function.
20577 (dom_walker::dom_walker): Convert bool param
20578 "skip_unreachable_blocks" to enum reachability. Move setup of
20579 edge flags to set_all_edges_as_executable and only do it when
20580 reachability is REACHABLE_BLOCKS.
20581 * domwalk.h (enum dom_walker::reachability): New enum.
20582 (dom_walker::dom_walker): Convert bool param
20583 "skip_unreachable_blocks" to enum reachability.
20584 (set_all_edges_as_executable): New decl.
20585 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
20586 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
20588 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
20589 but converting true to REACHABLE_BLOCKS.
20590 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
20592 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
20593 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
20594 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
20596 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
20597 if check_all_array_refs will be called.
20599 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20601 * tree.c (selftest::test_location_wrappers): Add more test
20604 2018-01-23 David Malcolm <dmalcolm@redhat.com>
20606 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
20607 (selftest::test_bit_in_range): Likewise.
20609 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20612 * doc/sourcebuild.texi (vect_float): Say that the selector
20613 only describes the situation when -funsafe-math-optimizations is on.
20614 (vect_float_strict): Document.
20616 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
20618 PR tree-optimization/83965
20619 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
20620 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
20621 instead of checking only for a reduction.
20622 (vect_recog_widen_sum_pattern): Likewise.
20624 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20626 * predict.c (probably_never_executed): Only use precise profile info.
20627 (compute_function_frequency): Skip after inlining hack since we now
20628 have quality checking.
20630 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
20632 * profile-count.h (profile_probability::very_unlikely,
20633 profile_probability::unlikely, profile_probability::even): Set
20634 precision to guessed.
20636 2018-01-23 Richard Biener <rguenther@suse.de>
20638 PR tree-optimization/83963
20639 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
20640 Properly terminate dominator walk when crossing the exit edge not
20641 when visiting its source block.
20643 2018-01-23 Jakub Jelinek <jakub@redhat.com>
20646 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
20647 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
20649 2018-01-22 Jakub Jelinek <jakub@redhat.com>
20651 PR tree-optimization/83957
20652 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
20653 semicolon after for body surrounded by braces.
20655 PR tree-optimization/83081
20656 * profile-count.h (profile_probability::split): New method.
20657 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
20658 Use profile_probability::split.
20659 (do_compare_rtx_and_jump): Fix adjustment of probabilities
20660 when splitting a single conditional jump into 2.
20662 2018-01-22 David Malcolm <dmalcolm@redhat.com>
20664 PR tree-optimization/69452
20665 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
20668 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20670 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
20671 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
20672 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
20674 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20676 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
20678 * config/rl78/rl78.md (movdi): New define_expand.
20679 * config/rl78/rl78.c (rl78_split_movdi): New function.
20681 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
20684 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
20686 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
20687 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
20688 128-bit to produce an UNSPEC move to get the double word with the
20689 signbit and then a shift directly to do signbit.
20690 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
20691 implementation with a new version that just does either a direct
20692 move or a regular move. Move memory interface to separate insns.
20693 Move insns so they are next to the expander.
20694 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
20695 with signbit move. Split big and little endian case.
20696 (signbit<mode>2_dm_mem_le): Likewise.
20697 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
20698 (signbit<mode>2_dm2): Likewise.
20700 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20702 * config/rl78/rl78.md (anddi3): New define_expand.
20704 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20706 * config/rl78/rl78.md (umindi3): New define_expand.
20708 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20710 * config/rl78/rl78.md (smindi3): New define_expand.
20712 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20714 * config/rl78/rl78.md (smaxdi3): New define_expand.
20716 2018-01-22 Carl Love <cel@us.ibm.com>
20718 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
20719 LVX_V1TI): Add macro expansion.
20720 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
20721 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
20722 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
20723 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
20724 Change check to determine if the instruction is a byte reversing
20725 entry. Fix typo in comment.
20726 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
20727 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
20728 Add def_builtin calls for new builtins.
20729 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
20730 Add define_insn expansion.
20732 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20734 * config/rl78/rl78.md (umaxdi3): New define_expand.
20736 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
20738 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
20739 for non-QImode registers.
20741 2018-01-22 Richard Biener <rguenther@suse.de>
20743 PR tree-optimization/83963
20744 * graphite-scop-detection.c (scop_detection::get_sese): Delay
20745 including the loop exit block.
20746 (scop_detection::merge_sese): Likewise.
20747 (scop_detection::add_scop): Do it here instead.
20749 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20751 * doc/sourcebuild.texi (arm_softfloat): Document.
20753 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
20756 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
20757 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
20758 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
20760 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20761 David Edelsohn <dje.gcc@gmail.com>
20764 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20765 Change "crset eq" to "crset 2".
20766 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20767 (*call_indirect_aix<mode>_nospec): Likewise.
20768 (*call_value_indirect_aix<mode>_nospec): Likewise.
20769 (*call_indirect_elfv2<mode>_nospec): Likewise.
20770 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
20771 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
20772 change assembly output from . to $.
20773 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20774 (indirect_jump<mode>_nospec): Change assembly output from . to $.
20775 (*tablejump<mode>_internal1_nospec): Likewise.
20777 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
20780 * config/sh/sh_optimize_sett_clrt.cc:
20781 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
20783 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
20785 PR tree-optimization/83940
20786 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
20787 offset_dt to vect_constant_def rather than vect_unknown_def_type.
20788 (vect_check_load_store_mask): Add a mask_dt_out parameter and
20789 use it to pass back the definition type.
20790 (vect_check_store_rhs): Likewise rhs_dt_out.
20791 (vect_build_gather_load_calls): Add a mask_dt argument and use
20792 it instead of a call to vect_is_simple_use.
20793 (vectorizable_store): Update calls to vect_check_load_store_mask
20794 and vect_check_store_rhs. Use the dt returned by the latter instead
20795 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
20796 instead of calls to vect_is_simple_use. Pass the scalar rather
20797 than the vector operand to vect_is_simple_use when handling
20798 second and subsequent copies of an rhs value.
20799 (vectorizable_load): Update calls to vect_check_load_store_mask
20800 and vect_build_gather_load_calls. Use the cached mask_dt and
20801 gs_info.offset_dt instead of calls to vect_is_simple_use.
20803 2018-01-20 Jakub Jelinek <jakub@redhat.com>
20805 PR middle-end/83945
20806 * tree-emutls.c: Include gimplify.h.
20807 (lower_emutls_2): New function.
20808 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
20809 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
20810 it before further processing.
20813 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
20814 UINTVAL (trueop1) instead of INTVAL (op1).
20816 2018-01-19 Jakub Jelinek <jakub@redhat.com>
20820 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
20821 INCOMING_FRAME_SP_OFFSET if not defined.
20822 (scan_trace): Add ENTRY argument. If true and
20823 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
20824 emit a note to adjust the CFA offset.
20825 (create_cfi_notes): Adjust scan_trace callers.
20826 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
20827 INCOMING_FRAME_SP_OFFSET in the CIE.
20828 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
20829 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
20831 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
20832 * doc/tm.texi: Regenerated.
20834 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20836 PR rtl-optimization/83147
20837 * lra-constraints.c (remove_inheritance_pseudos): Use
20838 lra_substitute_pseudo_within_insn.
20840 2018-01-19 Tom de Vries <tom@codesourcery.com>
20841 Cesar Philippidis <cesar@codesourcery.com>
20844 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
20846 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
20849 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
20850 spaces for function labels.
20852 2018-01-19 Martin Liska <mliska@suse.cz>
20854 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
20855 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
20856 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
20857 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
20858 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
20859 (PRED_CONST_RETURN): Change from 69 to 65.
20860 (PRED_NULL_RETURN): Change from 91 to 71.
20861 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
20862 (PRED_LOOP_GUARD): Change from 66 to 73.
20864 2018-01-19 Martin Liska <mliska@suse.cz>
20866 * predict.c (predict_insn_def): Add new assert.
20867 (struct branch_predictor): Change type to signed integer.
20868 (test_prediction_value_range): Amend test to cover
20869 PROB_UNINITIALIZED.
20870 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
20871 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
20872 (PRED_LOOP_ITERATIONS_MAX): Likewise.
20873 (PRED_LOOP_IV_COMPARE): Likewise.
20874 * predict.h (PROB_UNINITIALIZED): Define new constant.
20876 2018-01-19 Martin Liska <mliska@suse.cz>
20878 * predict.c (dump_prediction): Add new format for
20879 analyze_brprob.py script which is enabled with -details
20881 * profile-count.h (precise_p): New function.
20883 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20885 PR tree-optimization/83922
20886 * tree-vect-loop.c (vect_verify_full_masking): Return false if
20887 there are no statements that need masking.
20888 (vect_active_double_reduction_p): New function.
20889 (vect_analyze_loop_operations): Use it when handling phis that
20890 are not in the loop header.
20892 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
20894 PR tree-optimization/83914
20895 * tree-vect-loop.c (vectorizable_induction): Don't convert
20896 init_expr or apply the peeling adjustment for inductions
20897 that are nested within the vectorized loop.
20899 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20901 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
20904 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20908 * function.h (gimplify_parameters): Add gimple_seq * argument.
20909 * function.c: Include gimple.h and options.h.
20910 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
20911 for the added local temporaries if needed.
20912 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
20913 if there are any parameter cleanups, wrap whole body into a
20914 try/finally with the cleanups.
20916 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
20919 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
20920 Use GET_MODE_CLASS for scalar floating point.
20922 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20926 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
20927 Fix call of call_cgraph_insertion_hooks.
20929 2018-01-18 Martin Sebor <msebor@redhat.com>
20931 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
20933 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
20936 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
20939 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
20942 * common.opt: (-ffile-prefix-map): New option.
20943 * opts.c (common_handle_option): Defer it.
20944 * opts-global.c (handle_common_deferred_options): Handle it.
20945 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
20946 * file-prefix-map.h: New file.
20947 (remap_debug_filename, add_debug_prefix_map): ...here.
20948 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
20949 * final.c (debug_prefix_map, add_debug_prefix_map
20950 remap_debug_filename): Move to...
20951 * file-prefix-map.c: New file.
20952 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
20953 generalize, get rid of alloca(), use strrchr() instead of strchr().
20954 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
20955 Implement in terms of add_prefix_map().
20956 (remap_macro_filename, remap_debug_filename): Implement in term of
20958 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
20959 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
20960 * dbxout.c: Include file-prefix-map.h.
20961 * varasm.c: Likewise.
20962 * vmsdbgout.c: Likewise.
20963 * xcoffout.c: Likewise.
20964 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
20965 * doc/cppopts.texi (-fmacro-prefix-map): Document.
20966 * doc/invoke.texi (-ffile-prefix-map): Document.
20967 (-fdebug-prefix-map): Update description.
20969 2018-01-18 Martin Liska <mliska@suse.cz>
20971 * config/i386/i386.c (indirect_thunk_name): Document that also
20973 (output_indirect_thunk): Document why both instructions
20974 (pause and lfence) are generated.
20976 2018-01-18 Richard Biener <rguenther@suse.de>
20978 PR tree-optimization/83887
20979 * graphite-scop-detection.c
20980 (scop_detection::get_nearest_dom_with_single_entry): Remove.
20981 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
20982 (scop_detection::merge_sese): Re-implement with a flood-fill
20983 algorithm that properly finds a SESE region if it exists.
20985 2018-01-18 Jakub Jelinek <jakub@redhat.com>
20988 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
20989 pointer_diff optimizations use view_convert instead of convert.
20991 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20993 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20994 Generate different code for -mno-speculate-indirect-jumps.
20995 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20996 (*call_indirect_aix<mode>): Disable for
20997 -mno-speculate-indirect-jumps.
20998 (*call_indirect_aix<mode>_nospec): New define_insn.
20999 (*call_value_indirect_aix<mode>): Disable for
21000 -mno-speculate-indirect-jumps.
21001 (*call_value_indirect_aix<mode>_nospec): New define_insn.
21002 (*sibcall_nonlocal_sysv<mode>): Generate different code for
21003 -mno-speculate-indirect-jumps.
21004 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21006 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
21008 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
21009 long double type, set the flags for noting the default long double
21010 type, even if we don't pass or return a long double type.
21012 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
21015 * ipa-inline.c (flatten_function): Do not overwrite final inlining
21018 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
21020 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
21021 support for merge[hl].
21022 (fold_mergehl_helper): New helper function.
21023 (tree-vector-builder.h): New #include for tree_vector_builder usage.
21024 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
21025 (altivec_vmrglw_direct): Add xxmrglw insn.
21027 2018-01-17 Andrew Waterman <andrew@sifive.com>
21029 * config/riscv/riscv.c (riscv_conditional_register_usage): If
21030 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
21032 2018-01-17 David Malcolm <dmalcolm@redhat.com>
21035 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
21036 call the lto_location_cache before reading the
21037 DECL_SOURCE_LOCATION of the types.
21039 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
21040 Richard Sandiford <richard.sandiford@linaro.org>
21042 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
21043 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
21044 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
21045 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
21046 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
21048 * config/aarch64/constraints.md (aarch64_movti_operand):
21050 * config/aarch64/predicates.md (Uti): Add new constraint.
21052 2018-01-17 Carl Love <cel@us.ibm.com>
21054 * config/rs6000/vsx.md (define_expand xl_len_r,
21055 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
21056 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
21058 (define_expand, define_insn): Move the shift left from the
21059 define_insn to the define_expand for lxvl and stxvl instructions.
21060 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
21061 and XL_LEN_R definitions to PURE.
21063 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
21065 * config/i386/i386.c (indirect_thunk_name): Declare regno
21066 as unsigned int. Compare regno with INVALID_REGNUM.
21067 (output_indirect_thunk): Ditto.
21068 (output_indirect_thunk_function): Ditto.
21069 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
21070 in the call to output_indirect_thunk_function.
21072 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
21074 PR middle-end/83884
21075 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
21076 rather than the size of inner_type to determine the stack slot size
21077 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
21079 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
21082 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
21085 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
21087 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
21088 endian Linux systems to optionally enable multilibs for selecting
21089 the long double type if the user configured an explicit type.
21090 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
21091 have no long double multilibs if not defined.
21092 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
21093 warn if the user used -mabi={ieee,ibm}longdouble and we built
21094 multilibs for long double.
21095 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
21096 appropriate multilib option.
21097 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
21099 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
21100 for building long double multilibs.
21101 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
21103 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
21105 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
21108 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
21110 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
21113 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
21116 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
21119 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21121 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
21122 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
21124 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
21126 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
21127 different rtl trees depending on TARGET_64BIT.
21128 (rs6000_gen_lvx): Likewise.
21130 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
21132 * config/visium/visium.md (nop): Tweak comment.
21133 (hazard_nop): Likewise.
21135 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21137 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
21138 -mspeculate-indirect-jumps.
21139 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
21140 for -mno-speculate-indirect-jumps.
21141 (*call_indirect_elfv2<mode>_nospec): New define_insn.
21142 (*call_value_indirect_elfv2<mode>): Disable for
21143 -mno-speculate-indirect-jumps.
21144 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
21145 (indirect_jump): Emit different RTL for
21146 -mno-speculate-indirect-jumps.
21147 (*indirect_jump<mode>): Disable for
21148 -mno-speculate-indirect-jumps.
21149 (*indirect_jump<mode>_nospec): New define_insn.
21150 (tablejump): Emit different RTL for
21151 -mno-speculate-indirect-jumps.
21152 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
21153 (tablejumpsi_nospec): New define_expand.
21154 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
21155 (tablejumpdi_nospec): New define_expand.
21156 (*tablejump<mode>_internal1): Disable for
21157 -mno-speculate-indirect-jumps.
21158 (*tablejump<mode>_internal1_nospec): New define_insn.
21159 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
21162 2018-01-16 Artyom Skrobov tyomitch@gmail.com
21164 * caller-save.c (insert_save): Drop unnecessary parameter. All
21167 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21168 Richard Biener <rguenth@suse.de>
21171 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
21172 return early, inline manually is_gimple_sizepos. Make sure if we
21173 call gimplify_expr we don't end up with a gimple constant.
21174 * tree.c (variably_modified_type_p): Don't return true for
21175 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
21176 * gimplify.h (is_gimple_sizepos): Remove.
21178 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21180 PR tree-optimization/83857
21181 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
21182 vectorizable_live_operation for pure SLP statements.
21183 (vectorizable_live_operation): Handle PHIs.
21185 2018-01-16 Richard Biener <rguenther@suse.de>
21187 PR tree-optimization/83867
21188 * tree-vect-stmts.c (vect_transform_stmt): Precompute
21189 nested_in_vect_loop_p since the scalar stmt may get invalidated.
21191 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21194 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
21195 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
21196 If off is not INTEGER_CST, issue a may not be aligned warning
21197 rather than isn't aligned. Use isn%'t rather than isn't.
21198 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
21200 <case MULT_EXPR>: Improve the case when bottom and one of the
21201 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
21202 operand, in that case check if the other operand is multiple of
21203 bottom divided by the INTEGER_CST operand.
21205 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21208 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
21209 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
21210 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
21211 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
21212 * config/pa/pa.c (pa_function_arg_advance): Likewise.
21213 (pa_function_arg, pa_arg_partial_bytes): Likewise.
21214 (pa_function_arg_size): New function.
21216 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21218 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
21219 in a separate statement.
21221 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
21223 PR tree-optimization/83847
21224 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
21225 group gathers and scatters.
21227 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21229 PR rtl-optimization/86620
21230 * params.def (max-sched-ready-insns): Bump minimum value to 1.
21232 PR rtl-optimization/83213
21233 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
21234 to last if both are JUMP_INSNs.
21236 PR tree-optimization/83843
21237 * gimple-ssa-store-merging.c
21238 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
21239 store_immediate_info for bswap/nop orig_stores.
21241 2018-01-15 Andrew Waterman <andrew@sifive.com>
21243 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
21245 <UDIV>: Increase cost if !TARGET_DIV.
21247 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
21249 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
21250 (define_attr "cr_logical_3op"): New.
21251 (cceq_ior_compare): Adjust.
21252 (cceq_ior_compare_complement): Adjust.
21253 (*cceq_rev_compare): Adjust.
21254 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
21255 (is_cracked_insn): Adjust.
21256 (insn_must_be_first_in_group): Adjust.
21257 * config/rs6000/40x.md: Adjust.
21258 * config/rs6000/440.md: Adjust.
21259 * config/rs6000/476.md: Adjust.
21260 * config/rs6000/601.md: Adjust.
21261 * config/rs6000/603.md: Adjust.
21262 * config/rs6000/6xx.md: Adjust.
21263 * config/rs6000/7450.md: Adjust.
21264 * config/rs6000/7xx.md: Adjust.
21265 * config/rs6000/8540.md: Adjust.
21266 * config/rs6000/cell.md: Adjust.
21267 * config/rs6000/e300c2c3.md: Adjust.
21268 * config/rs6000/e500mc.md: Adjust.
21269 * config/rs6000/e500mc64.md: Adjust.
21270 * config/rs6000/e5500.md: Adjust.
21271 * config/rs6000/e6500.md: Adjust.
21272 * config/rs6000/mpc.md: Adjust.
21273 * config/rs6000/power4.md: Adjust.
21274 * config/rs6000/power5.md: Adjust.
21275 * config/rs6000/power6.md: Adjust.
21276 * config/rs6000/power7.md: Adjust.
21277 * config/rs6000/power8.md: Adjust.
21278 * config/rs6000/power9.md: Adjust.
21279 * config/rs6000/rs64.md: Adjust.
21280 * config/rs6000/titan.md: Adjust.
21282 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21284 * config/i386/predicates.md (indirect_branch_operand): Rewrite
21285 ix86_indirect_branch_register logic.
21287 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21289 * config/i386/constraints.md (Bs): Update
21290 ix86_indirect_branch_register check. Don't check
21291 ix86_indirect_branch_register with GOT_memory_operand.
21293 * config/i386/predicates.md (GOT_memory_operand): Don't check
21294 ix86_indirect_branch_register here.
21295 (GOT32_symbol_operand): Likewise.
21297 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21299 * config/i386/predicates.md (constant_call_address_operand):
21300 Rewrite ix86_indirect_branch_register logic.
21301 (sibcall_insn_operand): Likewise.
21303 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21305 * config/i386/constraints.md (Bs): Replace
21306 ix86_indirect_branch_thunk_register with
21307 ix86_indirect_branch_register.
21309 * config/i386/i386.md (indirect_jump): Likewise.
21310 (tablejump): Likewise.
21311 (*sibcall_memory): Likewise.
21312 (*sibcall_value_memory): Likewise.
21313 Peepholes of indirect call and jump via memory: Likewise.
21314 * config/i386/i386.opt: Likewise.
21315 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21316 (GOT_memory_operand): Likewise.
21317 (call_insn_operand): Likewise.
21318 (sibcall_insn_operand): Likewise.
21319 (GOT32_symbol_operand): Likewise.
21321 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21323 PR middle-end/83837
21324 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
21325 type rather than type addr's type points to.
21326 (expand_omp_atomic_mutex): Likewise.
21327 (expand_omp_atomic): Likewise.
21329 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
21332 * config/i386/i386.c (output_indirect_thunk_function): Use
21333 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
21334 for __x86_return_thunk.
21336 2018-01-15 Richard Biener <rguenther@suse.de>
21338 PR middle-end/83850
21339 * expmed.c (extract_bit_field_1): Fix typo.
21341 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21344 * config/arm/iterators.md (VF): New mode iterator.
21345 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
21346 Remove integer-related logic from pattern.
21347 (neon_vabd<mode>_3): Likewise.
21349 2018-01-15 Jakub Jelinek <jakub@redhat.com>
21351 PR middle-end/82694
21352 * common.opt (fstrict-overflow): No longer an alias.
21353 (fwrapv-pointer): New option.
21354 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
21355 also for pointer types based on flag_wrapv_pointer.
21356 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
21357 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
21358 opts->x_flag_wrapv got set.
21359 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
21360 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
21361 POINTER_TYPE_OVERFLOW_UNDEFINED.
21362 * match.pd: Likewise in address comparison pattern.
21363 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
21365 2018-01-15 Richard Biener <rguenther@suse.de>
21368 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
21369 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
21370 Reset type names to their identifier if their TYPE_DECL doesn't
21371 have linkage (and thus is used for ODR and devirt).
21372 (save_debug_info_for_decl): Remove.
21373 (save_debug_info_for_type): Likewise.
21374 (add_tree_to_fld_list): Adjust.
21375 * tree-pretty-print.c (dump_generic_node): Make dumping of
21376 type names more robust.
21378 2018-01-15 Richard Biener <rguenther@suse.de>
21380 * BASE-VER: Bump to 8.0.1.
21382 2018-01-14 Martin Sebor <msebor@redhat.com>
21385 * builtins.c (check_access): Avoid warning when the no-warning bit
21388 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
21390 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
21391 * ira-color (allocno_hard_regs_compare): Likewise.
21393 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
21396 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
21397 Use .pushsection/.popsection.
21399 2018-01-14 Martin Sebor <msebor@redhat.com>
21402 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
21404 2018-01-14 Jakub Jelinek <jakub@redhat.com>
21406 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
21407 entry from extra_headers.
21408 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
21409 extra_headers, make the list bitwise identical to the i?86-*-* one.
21411 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21413 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
21414 -mcmodel=large with -mindirect-branch=thunk,
21415 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
21416 -mfunction-return=thunk-extern.
21417 * doc/invoke.texi: Document -mcmodel=large is incompatible with
21418 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
21419 -mfunction-return=thunk and -mfunction-return=thunk-extern.
21421 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21423 * config/i386/i386.c (print_reg): Print the name of the full
21424 integer register without '%'.
21425 (ix86_print_operand): Handle 'V'.
21426 * doc/extend.texi: Document 'V' modifier.
21428 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21430 * config/i386/constraints.md (Bs): Disallow memory operand for
21431 -mindirect-branch-register.
21433 * config/i386/predicates.md (indirect_branch_operand): Likewise.
21434 (GOT_memory_operand): Likewise.
21435 (call_insn_operand): Likewise.
21436 (sibcall_insn_operand): Likewise.
21437 (GOT32_symbol_operand): Likewise.
21438 * config/i386/i386.md (indirect_jump): Call convert_memory_address
21439 for -mindirect-branch-register.
21440 (tablejump): Likewise.
21441 (*sibcall_memory): Likewise.
21442 (*sibcall_value_memory): Likewise.
21443 Disallow peepholes of indirect call and jump via memory for
21444 -mindirect-branch-register.
21445 (*call_pop): Replace m with Bw.
21446 (*call_value_pop): Likewise.
21447 (*sibcall_pop_memory): Replace m with Bs.
21448 * config/i386/i386.opt (mindirect-branch-register): New option.
21449 * doc/invoke.texi: Document -mindirect-branch-register option.
21451 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21453 * config/i386/i386-protos.h (ix86_output_function_return): New.
21454 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
21455 set function_return_type.
21456 (indirect_thunk_name): Add ret_p to indicate thunk for function
21458 (output_indirect_thunk_function): Pass false to
21459 indirect_thunk_name.
21460 (ix86_output_indirect_branch_via_reg): Likewise.
21461 (ix86_output_indirect_branch_via_push): Likewise.
21462 (output_indirect_thunk_function): Create alias for function
21463 return thunk if regno < 0.
21464 (ix86_output_function_return): New function.
21465 (ix86_handle_fndecl_attribute): Handle function_return.
21466 (ix86_attribute_table): Add function_return.
21467 * config/i386/i386.h (machine_function): Add
21468 function_return_type.
21469 * config/i386/i386.md (simple_return_internal): Use
21470 ix86_output_function_return.
21471 (simple_return_internal_long): Likewise.
21472 * config/i386/i386.opt (mfunction-return=): New option.
21473 (indirect_branch): Mention -mfunction-return=.
21474 * doc/extend.texi: Document function_return function attribute.
21475 * doc/invoke.texi: Document -mfunction-return= option.
21477 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
21479 * config/i386/i386-opts.h (indirect_branch): New.
21480 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
21481 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
21482 with local indirect jump when converting indirect call and jump.
21483 (ix86_set_indirect_branch_type): New.
21484 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
21485 (indirectlabelno): New.
21486 (indirect_thunk_needed): Likewise.
21487 (indirect_thunk_bnd_needed): Likewise.
21488 (indirect_thunks_used): Likewise.
21489 (indirect_thunks_bnd_used): Likewise.
21490 (INDIRECT_LABEL): Likewise.
21491 (indirect_thunk_name): Likewise.
21492 (output_indirect_thunk): Likewise.
21493 (output_indirect_thunk_function): Likewise.
21494 (ix86_output_indirect_branch_via_reg): Likewise.
21495 (ix86_output_indirect_branch_via_push): Likewise.
21496 (ix86_output_indirect_branch): Likewise.
21497 (ix86_output_indirect_jmp): Likewise.
21498 (ix86_code_end): Call output_indirect_thunk_function if needed.
21499 (ix86_output_call_insn): Call ix86_output_indirect_branch if
21501 (ix86_handle_fndecl_attribute): Handle indirect_branch.
21502 (ix86_attribute_table): Add indirect_branch.
21503 * config/i386/i386.h (machine_function): Add indirect_branch_type
21504 and has_local_indirect_jump.
21505 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
21507 (tablejump): Likewise.
21508 (*indirect_jump): Use ix86_output_indirect_jmp.
21509 (*tablejump_1): Likewise.
21510 (simple_return_indirect_internal): Likewise.
21511 * config/i386/i386.opt (mindirect-branch=): New option.
21512 (indirect_branch): New.
21515 (thunk-inline): Likewise.
21516 (thunk-extern): Likewise.
21517 * doc/extend.texi: Document indirect_branch function attribute.
21518 * doc/invoke.texi: Document -mindirect-branch= option.
21520 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
21523 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
21525 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
21527 * ipa-inline.c (want_inline_small_function_p): Return false if
21528 inlining has already failed with CIF_FINAL_ERROR.
21529 (update_caller_keys): Call want_inline_small_function_p before
21531 (update_callee_keys): Likewise.
21533 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
21535 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
21537 (rs6000_quadword_masked_address_p): Likewise.
21538 (quad_aligned_load_p): Likewise.
21539 (quad_aligned_store_p): Likewise.
21540 (const_load_sequence_p): Add comment to describe the outer-most loop.
21541 (mimic_memory_attributes_and_flags): New function.
21542 (rs6000_gen_stvx): Likewise.
21543 (replace_swapped_aligned_store): Likewise.
21544 (rs6000_gen_lvx): Likewise.
21545 (replace_swapped_aligned_load): Likewise.
21546 (replace_swapped_load_constant): Capitalize argument name in
21547 comment describing this function.
21548 (rs6000_analyze_swaps): Add a third pass to search for vector loads
21549 and stores that access quad-word aligned addresses and replace
21550 with stvx or lvx instructions when appropriate.
21551 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
21552 New function prototype.
21553 (rs6000_quadword_masked_address_p): Likewise.
21554 (rs6000_gen_lvx): Likewise.
21555 (rs6000_gen_stvx): Likewise.
21556 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
21557 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
21558 when memory address is aligned.
21559 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
21560 this split to select lvx instruction when memory address is aligned.
21561 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
21562 instruction when memory address is aligned.
21563 (*vsx_le_perm_load_v16qi): Likewise.
21564 (four unnamed splitters): Modify to select the stvx instruction
21565 when memory is aligned.
21567 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
21569 * predict.c (determine_unlikely_bbs): Handle correctly BBs
21570 which appears in the queue multiple times.
21572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21573 Alan Hayward <alan.hayward@arm.com>
21574 David Sherwood <david.sherwood@arm.com>
21576 * tree-vectorizer.h (vec_lower_bound): New structure.
21577 (_loop_vec_info): Add check_nonzero and lower_bounds.
21578 (LOOP_VINFO_CHECK_NONZERO): New macro.
21579 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
21580 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
21581 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
21582 fields. Make seg_len the distance travelled, not including the
21584 (dr_direction_indicator): Declare.
21585 (dr_zero_step_indicator): Likewise.
21586 (dr_known_forward_stride_p): Likewise.
21587 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
21589 (runtime_alias_check_p): Allow runtime alias checks with
21591 (operator ==): Compare access_size and align.
21592 (prune_runtime_alias_test_list): Rework for new distinction between
21593 the access_size and seg_len.
21594 (create_intersect_range_checks_index): Likewise. Cope with polynomial
21596 (get_segment_min_max): New function.
21597 (create_intersect_range_checks): Use it.
21598 (dr_step_indicator): New function.
21599 (dr_direction_indicator): Likewise.
21600 (dr_zero_step_indicator): Likewise.
21601 (dr_known_forward_stride_p): Likewise.
21602 * tree-loop-distribution.c (data_ref_segment_size): Return
21603 DR_STEP * (niters - 1).
21604 (compute_alias_check_pairs): Update call to the dr_with_seg_len
21606 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
21607 (vect_preserves_scalar_order_p): New function, split out from...
21608 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
21609 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
21610 (vect_vfa_access_size): New function.
21611 (vect_vfa_align): Likewise.
21612 (vect_compile_time_alias): Take access_size_a and access_b arguments.
21613 (dump_lower_bound): New function.
21614 (vect_check_lower_bound): Likewise.
21615 (vect_small_gap_p): Likewise.
21616 (vectorizable_with_step_bound_p): Likewise.
21617 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
21618 depencies if the vectorization factor is 1. Convert the checks
21619 for nonzero steps into checks on the bounds of DR_STEP. Try using
21620 a bunds check for variable steps if the minimum required step is
21621 relatively small. Update calls to the dr_with_seg_len
21622 constructor and to vect_compile_time_alias.
21623 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
21625 (vect_loop_versioning): Call it.
21626 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
21628 (vect_estimate_min_profitable_iters): Account for any bounds checks.
21630 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21631 Alan Hayward <alan.hayward@arm.com>
21632 David Sherwood <david.sherwood@arm.com>
21634 * doc/sourcebuild.texi (vect_scatter_store): Document.
21635 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
21637 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
21639 * genopinit.c (main): Add supports_vec_scatter_store and
21640 supports_vec_scatter_store_cached to target_optabs.
21641 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
21642 IFN_MASK_SCATTER_STORE.
21643 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
21645 * internal-fn.h (internal_store_fn_p): Declare.
21646 (internal_fn_stored_value_index): Likewise.
21647 * internal-fn.c (scatter_store_direct): New macro.
21648 (expand_scatter_store_optab_fn): New function.
21649 (direct_scatter_store_optab_supported_p): New macro.
21650 (internal_store_fn_p): New function.
21651 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
21652 IFN_MASK_SCATTER_STORE.
21653 (internal_fn_mask_index): Likewise.
21654 (internal_fn_stored_value_index): New function.
21655 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
21656 for scatter stores.
21657 * optabs-query.h (supports_vec_scatter_store_p): Declare.
21658 * optabs-query.c (supports_vec_scatter_store_p): New function.
21659 * tree-vectorizer.h (vect_get_store_rhs): Declare.
21660 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
21661 true for scatter stores.
21662 (vect_gather_scatter_fn_p): Handle scatter stores too.
21663 (vect_check_gather_scatter): Consider using scatter stores if
21664 supports_vec_scatter_store_p.
21665 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
21666 scatter stores too.
21667 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
21668 internal_fn_stored_value_index.
21669 (check_load_store_masking): Handle scatter stores too.
21670 (vect_get_store_rhs): Make public.
21671 (vectorizable_call): Use internal_store_fn_p.
21672 (vectorizable_store): Handle scatter store internal functions.
21673 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
21674 when deciding whether the end of the group has been reached.
21675 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
21676 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
21677 (mask_scatter_store<mode>): New insns.
21679 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21680 Alan Hayward <alan.hayward@arm.com>
21681 David Sherwood <david.sherwood@arm.com>
21683 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
21684 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
21685 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
21687 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
21688 Use vect_truncate_gather_scatter_offset if we can't treat the
21689 operation as a normal gather load or scatter store.
21690 (get_group_load_store_type): Take the gather_scatter_info
21691 as argument. Try using a gather load or scatter store for
21692 single-element groups.
21693 (get_load_store_type): Update calls to get_group_load_store_type
21694 and vect_use_strided_gather_scatters_p.
21696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21697 Alan Hayward <alan.hayward@arm.com>
21698 David Sherwood <david.sherwood@arm.com>
21700 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
21701 optional tree argument.
21702 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
21704 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
21705 but continue to use the current value as a fallback.
21706 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
21707 to compare the updates.
21708 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
21709 (get_load_store_type): Use it when handling a strided access.
21710 (vect_get_strided_load_store_ops): New function.
21711 (vect_get_data_ptr_increment): Likewise.
21712 (vectorizable_load): Handle strided gather loads. Always pass
21713 a step to vect_create_data_ref_ptr and bump_vector_ptr.
21715 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21716 Alan Hayward <alan.hayward@arm.com>
21717 David Sherwood <david.sherwood@arm.com>
21719 * doc/md.texi (gather_load@var{m}): Document.
21720 (mask_gather_load@var{m}): Likewise.
21721 * genopinit.c (main): Add supports_vec_gather_load and
21722 supports_vec_gather_load_cached to target_optabs.
21723 * optabs-tree.c (init_tree_optimization_optabs): Use
21724 ggc_cleared_alloc to allocate target_optabs.
21725 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
21726 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
21728 * internal-fn.h (internal_load_fn_p): Declare.
21729 (internal_gather_scatter_fn_p): Likewise.
21730 (internal_fn_mask_index): Likewise.
21731 (internal_gather_scatter_fn_supported_p): Likewise.
21732 * internal-fn.c (gather_load_direct): New macro.
21733 (expand_gather_load_optab_fn): New function.
21734 (direct_gather_load_optab_supported_p): New macro.
21735 (direct_internal_fn_optab): New function.
21736 (internal_load_fn_p): Likewise.
21737 (internal_gather_scatter_fn_p): Likewise.
21738 (internal_fn_mask_index): Likewise.
21739 (internal_gather_scatter_fn_supported_p): Likewise.
21740 * optabs-query.c (supports_at_least_one_mode_p): New function.
21741 (supports_vec_gather_load_p): Likewise.
21742 * optabs-query.h (supports_vec_gather_load_p): Declare.
21743 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
21744 and memory_type field.
21745 (NUM_PATTERNS): Bump to 15.
21746 * tree-vect-data-refs.c: Include internal-fn.h.
21747 (vect_gather_scatter_fn_p): New function.
21748 (vect_describe_gather_scatter_call): Likewise.
21749 (vect_check_gather_scatter): Try using internal functions for
21750 gather loads. Recognize existing calls to a gather load function.
21751 (vect_analyze_data_refs): Consider using gather loads if
21752 supports_vec_gather_load_p.
21753 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
21754 (vect_get_gather_scatter_offset_type): Likewise.
21755 (vect_convert_mask_for_vectype): Likewise.
21756 (vect_add_conversion_to_patterm): Likewise.
21757 (vect_try_gather_scatter_pattern): Likewise.
21758 (vect_recog_gather_scatter_pattern): New pattern recognizer.
21759 (vect_vect_recog_func_ptrs): Add it.
21760 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
21761 internal_fn_mask_index and internal_gather_scatter_fn_p.
21762 (check_load_store_masking): Take the gather_scatter_info as an
21763 argument and handle gather loads.
21764 (vect_get_gather_scatter_ops): New function.
21765 (vectorizable_call): Check internal_load_fn_p.
21766 (vectorizable_load): Likewise. Handle gather load internal
21768 (vectorizable_store): Update call to check_load_store_masking.
21769 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
21770 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
21771 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
21772 (aarch64_gather_scale_operand_d): New predicates.
21773 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
21774 (mask_gather_load<mode>): New insns.
21776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21777 Alan Hayward <alan.hayward@arm.com>
21778 David Sherwood <david.sherwood@arm.com>
21780 * optabs.def (fold_left_plus_optab): New optab.
21781 * doc/md.texi (fold_left_plus_@var{m}): Document.
21782 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
21783 * internal-fn.c (fold_left_direct): Define.
21784 (expand_fold_left_optab_fn): Likewise.
21785 (direct_fold_left_optab_supported_p): Likewise.
21786 * fold-const-call.c (fold_const_fold_left): New function.
21787 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
21788 * tree-parloops.c (valid_reduction_p): New function.
21789 (gather_scalar_reductions): Use it.
21790 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
21791 (vect_finish_replace_stmt): Declare.
21792 * tree-vect-loop.c (fold_left_reduction_fn): New function.
21793 (needs_fold_left_reduction_p): New function, split out from...
21794 (vect_is_simple_reduction): ...here. Accept reductions that
21795 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
21796 (vect_force_simple_reduction): Also store the reduction type in
21797 the assignment's STMT_VINFO_REDUC_TYPE.
21798 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
21799 (merge_with_identity): New function.
21800 (vect_expand_fold_left): Likewise.
21801 (vectorize_fold_left_reduction): Likewise.
21802 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
21803 scalar phi in place for it. Check for target support and reject
21804 cases that would reassociate the operation. Defer the transform
21805 phase to vectorize_fold_left_reduction.
21806 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
21807 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
21808 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
21810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21812 * tree-if-conv.c (predicate_mem_writes): Remove redundant
21813 call to ifc_temp_var.
21815 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21816 Alan Hayward <alan.hayward@arm.com>
21817 David Sherwood <david.sherwood@arm.com>
21819 * target.def (legitimize_address_displacement): Take the original
21820 offset as a poly_int.
21821 * targhooks.h (default_legitimize_address_displacement): Update
21823 * targhooks.c (default_legitimize_address_displacement): Likewise.
21824 * doc/tm.texi: Regenerate.
21825 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
21826 as an argument, moving assert of ad->disp == ad->disp_term to...
21827 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
21828 Try calling targetm.legitimize_address_displacement before expanding
21829 the address rather than afterwards, and adjust for the new interface.
21830 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
21831 Match the new hook interface. Handle SVE addresses.
21832 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
21833 new hook interface.
21835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21837 * Makefile.in (OBJS): Add early-remat.o.
21838 * target.def (select_early_remat_modes): New hook.
21839 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
21840 * doc/tm.texi: Regenerate.
21841 * targhooks.h (default_select_early_remat_modes): Declare.
21842 * targhooks.c (default_select_early_remat_modes): New function.
21843 * timevar.def (TV_EARLY_REMAT): New timevar.
21844 * passes.def (pass_early_remat): New pass.
21845 * tree-pass.h (make_pass_early_remat): Declare.
21846 * early-remat.c: New file.
21847 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
21849 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
21851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21852 Alan Hayward <alan.hayward@arm.com>
21853 David Sherwood <david.sherwood@arm.com>
21855 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
21856 vfm1 with a bound_epilog parameter.
21857 (vect_do_peeling): Update calls accordingly, and move the prologue
21858 call earlier in the function. Treat the base bound_epilog as 0 for
21859 fully-masked loops and retain vf - 1 for other loops. Add 1 to
21860 this base when peeling for gaps.
21861 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
21862 with fully-masked loops.
21863 (vect_estimate_min_profitable_iters): Handle the single peeled
21864 iteration in that case.
21866 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21867 Alan Hayward <alan.hayward@arm.com>
21868 David Sherwood <david.sherwood@arm.com>
21870 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
21871 single-element interleaving even if the size is not a power of 2.
21872 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
21873 accesses for single-element interleaving if the group size is
21876 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21877 Alan Hayward <alan.hayward@arm.com>
21878 David Sherwood <david.sherwood@arm.com>
21880 * doc/md.texi (fold_extract_last_@var{m}): Document.
21881 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
21882 * optabs.def (fold_extract_last_optab): New optab.
21883 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
21884 * internal-fn.c (fold_extract_direct): New macro.
21885 (expand_fold_extract_optab_fn): Likewise.
21886 (direct_fold_extract_optab_supported_p): Likewise.
21887 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
21888 * tree-vect-loop.c (vect_model_reduction_cost): Handle
21889 EXTRACT_LAST_REDUCTION.
21890 (get_initial_def_for_reduction): Do not create an initial vector
21891 for EXTRACT_LAST_REDUCTION reductions.
21892 (vectorizable_reduction): Leave the scalar phi in place for
21893 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
21894 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
21895 epilogue code for EXTRACT_LAST_REDUCTION and defer the
21896 transform phase to vectorizable_condition.
21897 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
21899 (vect_finish_stmt_generation): ...here.
21900 (vect_finish_replace_stmt): New function.
21901 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
21902 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
21904 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
21906 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21907 Alan Hayward <alan.hayward@arm.com>
21908 David Sherwood <david.sherwood@arm.com>
21910 * doc/md.texi (extract_last_@var{m}): Document.
21911 * optabs.def (extract_last_optab): New optab.
21912 * internal-fn.def (EXTRACT_LAST): New internal function.
21913 * internal-fn.c (cond_unary_direct): New macro.
21914 (expand_cond_unary_optab_fn): Likewise.
21915 (direct_cond_unary_optab_supported_p): Likewise.
21916 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
21917 loops using EXTRACT_LAST.
21918 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
21919 (extract_last_<mode>): ...this optab.
21920 (vec_extract<mode><Vel>): Update accordingly.
21922 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21923 Alan Hayward <alan.hayward@arm.com>
21924 David Sherwood <david.sherwood@arm.com>
21926 * target.def (empty_mask_is_expensive): New hook.
21927 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
21928 * doc/tm.texi: Regenerate.
21929 * targhooks.h (default_empty_mask_is_expensive): Declare.
21930 * targhooks.c (default_empty_mask_is_expensive): New function.
21931 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
21932 if the target says that empty masks are expensive.
21933 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
21935 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
21937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21938 Alan Hayward <alan.hayward@arm.com>
21939 David Sherwood <david.sherwood@arm.com>
21941 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
21942 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
21943 (vect_use_loop_mask_for_alignment_p): New function.
21944 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
21945 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
21946 niters_skip argument. Make sure that the first niters_skip elements
21947 of the first iteration are inactive.
21948 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
21949 Update call to vect_set_loop_masks_directly.
21950 (get_misalign_in_elems): New function, split out from...
21951 (vect_gen_prolog_loop_niters): ...here.
21952 (vect_update_init_of_dr): Take a code argument that specifies whether
21953 the adjustment should be added or subtracted.
21954 (vect_update_init_of_drs): Likewise.
21955 (vect_prepare_for_masked_peels): New function.
21956 (vect_do_peeling): Skip prologue peeling if we're using a mask
21957 instead. Update call to vect_update_inits_of_drs.
21958 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21960 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
21961 alignment. Do not include the number of peeled iterations in
21962 the minimum threshold in that case.
21963 (vectorizable_induction): Adjust the start value down by
21964 LOOP_VINFO_MASK_SKIP_NITERS iterations.
21965 (vect_transform_loop): Call vect_prepare_for_masked_peels.
21966 Take the number of skipped iterations into account when calculating
21968 * tree-vect-stmts.c (vect_gen_while_not): New function.
21970 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21971 Alan Hayward <alan.hayward@arm.com>
21972 David Sherwood <david.sherwood@arm.com>
21974 * doc/sourcebuild.texi (vect_fully_masked): Document.
21975 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
21976 default value to 0.
21977 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
21979 (vect_analyze_loop_2): ...here. Don't check the vectorization
21980 factor against the number of loop iterations if the loop is
21983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
21984 Alan Hayward <alan.hayward@arm.com>
21985 David Sherwood <david.sherwood@arm.com>
21987 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
21988 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
21989 (dump_groups): Update accordingly.
21990 (iv_use::mem_type): New member variable.
21991 (address_p): New function.
21992 (record_use): Add a mem_type argument and initialize the new
21994 (record_group_use): Add a mem_type argument. Use address_p.
21995 Remove obsolete null checks of base_object. Update call to record_use.
21996 (find_interesting_uses_op): Update call to record_group_use.
21997 (find_interesting_uses_cond): Likewise.
21998 (find_interesting_uses_address): Likewise.
21999 (get_mem_type_for_internal_fn): New function.
22000 (find_address_like_use): Likewise.
22001 (find_interesting_uses_stmt): Try find_address_like_use before
22002 calling find_interesting_uses_op.
22003 (addr_offset_valid_p): Use the iv mem_type field as the type
22004 of the addressed memory.
22005 (add_autoinc_candidates): Likewise.
22006 (get_address_cost): Likewise.
22007 (split_small_address_groups_p): Use address_p.
22008 (split_address_groups): Likewise.
22009 (add_iv_candidate_for_use): Likewise.
22010 (autoinc_possible_for_pair): Likewise.
22011 (rewrite_groups): Likewise.
22012 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
22013 (determine_group_iv_cost): Update after split of USE_ADDRESS.
22014 (get_alias_ptr_type_for_ptr_address): New function.
22015 (rewrite_use_address): Rewrite address uses in calls that were
22016 identified by find_address_like_use.
22018 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22019 Alan Hayward <alan.hayward@arm.com>
22020 David Sherwood <david.sherwood@arm.com>
22022 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
22024 * gimple-expr.h (is_gimple_addressable: Likewise.
22025 * gimple-expr.c (is_gimple_address): Likewise.
22026 * internal-fn.c (expand_call_mem_ref): New function.
22027 (expand_mask_load_optab_fn): Use it.
22028 (expand_mask_store_optab_fn): Likewise.
22030 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22031 Alan Hayward <alan.hayward@arm.com>
22032 David Sherwood <david.sherwood@arm.com>
22034 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
22035 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
22036 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
22037 (cond_umax@var{mode}): Document.
22038 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
22039 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
22040 (cond_umin_optab, cond_umax_optab): New optabs.
22041 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
22042 (COND_IOR, COND_XOR): New internal functions.
22043 * internal-fn.h (get_conditional_internal_fn): Declare.
22044 * internal-fn.c (cond_binary_direct): New macro.
22045 (expand_cond_binary_optab_fn): Likewise.
22046 (direct_cond_binary_optab_supported_p): Likewise.
22047 (get_conditional_internal_fn): New function.
22048 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
22049 Cope with reduction statements that are vectorized as calls rather
22051 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
22052 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
22053 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
22054 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
22055 (UNSPEC_COND_EOR): New unspecs.
22056 (optab): Add mappings for them.
22057 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
22058 (sve_int_op, sve_fp_op): New int attributes.
22060 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22061 Alan Hayward <alan.hayward@arm.com>
22062 David Sherwood <david.sherwood@arm.com>
22064 * optabs.def (while_ult_optab): New optab.
22065 * doc/md.texi (while_ult@var{m}@var{n}): Document.
22066 * internal-fn.def (WHILE_ULT): New internal function.
22067 * internal-fn.h (direct_internal_fn_supported_p): New override
22068 that takes two types as argument.
22069 * internal-fn.c (while_direct): New macro.
22070 (expand_while_optab_fn): New function.
22071 (convert_optab_supported_p): Likewise.
22072 (direct_while_optab_supported_p): New macro.
22073 * wide-int.h (wi::udiv_ceil): New function.
22074 * tree-vectorizer.h (rgroup_masks): New structure.
22075 (vec_loop_masks): New typedef.
22076 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
22077 and fully_masked_p.
22078 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
22079 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
22080 (vect_max_vf): New function.
22081 (slpeel_make_loop_iterate_ntimes): Delete.
22082 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
22083 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
22084 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
22085 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
22086 internal-fn.h, stor-layout.h and optabs-query.h.
22087 (vect_set_loop_mask): New function.
22088 (add_preheader_seq): Likewise.
22089 (add_header_seq): Likewise.
22090 (interleave_supported_p): Likewise.
22091 (vect_maybe_permute_loop_masks): Likewise.
22092 (vect_set_loop_masks_directly): Likewise.
22093 (vect_set_loop_condition_masked): Likewise.
22094 (vect_set_loop_condition_unmasked): New function, split out from
22095 slpeel_make_loop_iterate_ntimes.
22096 (slpeel_make_loop_iterate_ntimes): Rename to..
22097 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
22098 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
22099 (vect_do_peeling): Update call accordingly.
22100 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
22102 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22103 mask_compare_type, can_fully_mask_p and fully_masked_p.
22104 (release_vec_loop_masks): New function.
22105 (_loop_vec_info): Use it to free the loop masks.
22106 (can_produce_all_loop_masks_p): New function.
22107 (vect_get_max_nscalars_per_iter): Likewise.
22108 (vect_verify_full_masking): Likewise.
22109 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
22110 retries, and free the mask rgroups before retrying. Check loop-wide
22111 reasons for disallowing fully-masked loops. Make the final decision
22112 about whether use a fully-masked loop or not.
22113 (vect_estimate_min_profitable_iters): Do not assume that peeling
22114 for the number of iterations will be needed for fully-masked loops.
22115 (vectorizable_reduction): Disable fully-masked loops.
22116 (vectorizable_live_operation): Likewise.
22117 (vect_halve_mask_nunits): New function.
22118 (vect_double_mask_nunits): Likewise.
22119 (vect_record_loop_mask): Likewise.
22120 (vect_get_loop_mask): Likewise.
22121 (vect_transform_loop): Handle the case in which the final loop
22122 iteration might handle a partial vector. Call vect_set_loop_condition
22123 instead of slpeel_make_loop_iterate_ntimes.
22124 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
22125 (check_load_store_masking): New function.
22126 (prepare_load_store_mask): Likewise.
22127 (vectorizable_store): Handle fully-masked loops.
22128 (vectorizable_load): Likewise.
22129 (supportable_widening_operation): Use vect_halve_mask_nunits for
22131 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
22132 (vect_gen_while): New function.
22133 * config/aarch64/aarch64.md (umax<mode>3): New expander.
22134 (aarch64_uqdec<mode>): New insn.
22136 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22137 Alan Hayward <alan.hayward@arm.com>
22138 David Sherwood <david.sherwood@arm.com>
22140 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
22141 (reduc_xor_scal_optab): New optabs.
22142 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
22143 (reduc_xor_scal_@var{m}): Document.
22144 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
22145 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
22146 internal functions.
22147 * fold-const-call.c (fold_const_call): Handle them.
22148 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
22149 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
22150 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
22151 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
22152 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
22153 (UNSPEC_XORV): New unspecs.
22154 (optab): Add entries for them.
22155 (BITWISEV): New int iterator.
22156 (bit_reduc_op): New int attributes.
22158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22159 Alan Hayward <alan.hayward@arm.com>
22160 David Sherwood <david.sherwood@arm.com>
22162 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
22163 * internal-fn.def (VEC_SHL_INSERT): New internal function.
22164 * optabs.def (vec_shl_insert_optab): New optab.
22165 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
22166 (duplicate_and_interleave): Likewise.
22167 * tree-vect-loop.c: Include internal-fn.h.
22168 (neutral_op_for_slp_reduction): New function, split out from
22169 get_initial_defs_for_reduction.
22170 (get_initial_def_for_reduction): Handle option 2 for variable-length
22171 vectors by loading the neutral value into a vector and then shifting
22172 the initial value into element 0.
22173 (get_initial_defs_for_reduction): Replace the code argument with
22174 the neutral value calculated by neutral_op_for_slp_reduction.
22175 Use gimple_build_vector for constant-length vectors.
22176 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
22177 but the first group_size elements have a neutral value.
22178 Use duplicate_and_interleave otherwise.
22179 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
22180 Update call to get_initial_defs_for_reduction. Handle SLP
22181 reductions for variable-length vectors by creating one vector
22182 result for each scalar result, with the elements associated
22183 with other scalar results stubbed out with the neutral value.
22184 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
22185 Require IFN_VEC_SHL_INSERT for double reductions on
22186 variable-length vectors, or SLP reductions that have
22187 a neutral value. Require can_duplicate_and_interleave_p
22188 support for variable-length unchained SLP reductions if there
22189 is no neutral value, such as for MIN/MAX reductions. Also require
22190 the number of vector elements to be a multiple of the number of
22191 SLP statements when doing variable-length unchained SLP reductions.
22192 Update call to vect_create_epilog_for_reduction.
22193 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
22194 and remove initial values.
22195 (duplicate_and_interleave): Make public.
22196 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
22197 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
22199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22200 Alan Hayward <alan.hayward@arm.com>
22201 David Sherwood <david.sherwood@arm.com>
22203 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
22204 (can_duplicate_and_interleave_p): New function.
22205 (vect_get_and_check_slp_defs): Take the vector of statements
22206 rather than just the current one. Remove excess parentheses.
22207 Restriction rejectinon of vect_constant_def and vect_external_def
22208 for variable-length vectors to boolean types, or types for which
22209 can_duplicate_and_interleave_p is false.
22210 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
22211 (duplicate_and_interleave): New function.
22212 (vect_get_constant_vectors): Use gimple_build_vector for
22213 constant-length vectors and suitable variable-length constant
22214 vectors. Use duplicate_and_interleave for other variable-length
22215 vectors. Don't defer the update when inserting new statements.
22217 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22218 Alan Hayward <alan.hayward@arm.com>
22219 David Sherwood <david.sherwood@arm.com>
22221 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
22222 min_profitable_iters doesn't go negative.
22224 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22225 Alan Hayward <alan.hayward@arm.com>
22226 David Sherwood <david.sherwood@arm.com>
22228 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
22229 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
22230 * optabs.def (vec_mask_load_lanes_optab): New optab.
22231 (vec_mask_store_lanes_optab): Likewise.
22232 * internal-fn.def (MASK_LOAD_LANES): New internal function.
22233 (MASK_STORE_LANES): Likewise.
22234 * internal-fn.c (mask_load_lanes_direct): New macro.
22235 (mask_store_lanes_direct): Likewise.
22236 (expand_mask_load_optab_fn): Handle masked operations.
22237 (expand_mask_load_lanes_optab_fn): New macro.
22238 (expand_mask_store_optab_fn): Handle masked operations.
22239 (expand_mask_store_lanes_optab_fn): New macro.
22240 (direct_mask_load_lanes_optab_supported_p): Likewise.
22241 (direct_mask_store_lanes_optab_supported_p): Likewise.
22242 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
22244 (vect_load_lanes_supported): Likewise.
22245 * tree-vect-data-refs.c (strip_conversion): New function.
22246 (can_group_stmts_p): Likewise.
22247 (vect_analyze_data_ref_accesses): Use it instead of checking
22248 for a pair of assignments.
22249 (vect_store_lanes_supported): Take a masked_p parameter.
22250 (vect_load_lanes_supported): Likewise.
22251 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
22252 vect_store_lanes_supported and vect_load_lanes_supported.
22253 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
22254 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
22255 parameter. Don't allow gaps for masked accesses.
22256 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
22257 and vect_load_lanes_supported.
22258 (get_load_store_type): Take a masked_p parameter and update
22259 call to get_group_load_store_type.
22260 (vectorizable_store): Update call to get_load_store_type.
22261 Handle IFN_MASK_STORE_LANES.
22262 (vectorizable_load): Update call to get_load_store_type.
22263 Handle IFN_MASK_LOAD_LANES.
22265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22266 Alan Hayward <alan.hayward@arm.com>
22267 David Sherwood <david.sherwood@arm.com>
22269 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
22271 * config/aarch64/aarch64-protos.h
22272 (aarch64_sve_struct_memory_operand_p): Declare.
22273 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
22274 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
22275 (VPRED, vpred): Handle SVE structure modes.
22276 * config/aarch64/constraints.md (Utx): New constraint.
22277 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
22278 (aarch64_sve_struct_nonimmediate_operand): New predicates.
22279 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
22280 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
22281 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
22282 structure modes. Split into pieces after RA.
22283 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
22284 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
22286 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
22287 SVE structure modes.
22288 (aarch64_classify_address): Likewise.
22289 (sizetochar): Move earlier in file.
22290 (aarch64_print_operand): Handle SVE register lists.
22291 (aarch64_array_mode): New function.
22292 (aarch64_sve_struct_memory_operand_p): Likewise.
22293 (TARGET_ARRAY_MODE): Redefine.
22295 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22296 Alan Hayward <alan.hayward@arm.com>
22297 David Sherwood <david.sherwood@arm.com>
22299 * target.def (array_mode): New target hook.
22300 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
22301 * doc/tm.texi: Regenerate.
22302 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
22303 * hooks.c (hook_optmode_mode_uhwi_none): New function.
22304 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
22305 targetm.array_mode.
22306 * stor-layout.c (mode_for_array): Likewise. Support polynomial
22309 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22310 Alan Hayward <alan.hayward@arm.com>
22311 David Sherwood <david.sherwood@arm.com>
22313 * fold-const.c (fold_binary_loc): Check the argument types
22314 rather than the result type when testing for a vector operation.
22316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22318 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
22319 * doc/tm.texi: Regenerate.
22321 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22322 Alan Hayward <alan.hayward@arm.com>
22323 David Sherwood <david.sherwood@arm.com>
22325 * doc/invoke.texi (-msve-vector-bits=): Document new option.
22326 (sve): Document new AArch64 extension.
22327 * doc/md.texi (w): Extend the description of the AArch64
22328 constraint to include SVE vectors.
22329 (Upl, Upa): Document new AArch64 predicate constraints.
22330 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
22332 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
22333 (msve-vector-bits=): New option.
22334 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
22335 SVE when these are disabled.
22336 (sve): New extension.
22337 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
22338 modes. Adjust their number of units based on aarch64_sve_vg.
22339 (MAX_BITSIZE_MODE_ANY_MODE): Define.
22340 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
22341 aarch64_addr_query_type.
22342 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
22343 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
22344 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
22345 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
22346 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
22347 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
22348 (aarch64_simd_imm_zero_p): Delete.
22349 (aarch64_check_zero_based_sve_index_immediate): Declare.
22350 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22351 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22352 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22353 (aarch64_sve_float_mul_immediate_p): Likewise.
22354 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22355 rather than an rtx.
22356 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
22357 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
22358 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
22359 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
22360 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
22361 (aarch64_regmode_natural_size): Likewise.
22362 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
22363 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
22365 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
22366 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
22367 for VG and the SVE predicate registers.
22368 (V_ALIASES): Add a "z"-prefixed alias.
22369 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
22370 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
22371 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
22372 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
22373 (REG_CLASS_NAMES): Add entries for them.
22374 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
22375 and the predicate registers.
22376 (aarch64_sve_vg): Declare.
22377 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
22378 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
22379 (REGMODE_NATURAL_SIZE): Define.
22380 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
22382 * config/aarch64/aarch64.c: Include cfgrtl.h.
22383 (simd_immediate_info): Add a constructor for series vectors,
22384 and an associated step field.
22385 (aarch64_sve_vg): New variable.
22386 (aarch64_dbx_register_number): Handle VG and the predicate registers.
22387 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
22388 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
22389 (VEC_ANY_DATA, VEC_STRUCT): New constants.
22390 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
22391 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
22392 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
22393 (aarch64_get_mask_mode): New functions.
22394 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
22395 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
22396 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
22397 predicate modes and predicate registers. Explicitly restrict
22398 GPRs to modes of 16 bytes or smaller. Only allow FP registers
22399 to store a vector mode if it is recognized by
22400 aarch64_classify_vector_mode.
22401 (aarch64_regmode_natural_size): New function.
22402 (aarch64_hard_regno_caller_save_mode): Return the original mode
22404 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
22405 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
22406 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
22407 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
22409 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
22410 does not overlap dest if the function is frame-related. Handle
22412 (aarch64_split_add_offset): New function.
22413 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
22414 them aarch64_add_offset.
22415 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
22416 and update call to aarch64_sub_sp.
22417 (aarch64_add_cfa_expression): New function.
22418 (aarch64_expand_prologue): Pass extra temporary registers to the
22419 functions above. Handle the case in which we need to emit new
22420 DW_CFA_expressions for registers that were originally saved
22421 relative to the stack pointer, but now have to be expressed
22422 relative to the frame pointer.
22423 (aarch64_output_mi_thunk): Pass extra temporary registers to the
22425 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
22426 IP0 and IP1 values for SVE frames.
22427 (aarch64_expand_vec_series): New function.
22428 (aarch64_expand_sve_widened_duplicate): Likewise.
22429 (aarch64_expand_sve_const_vector): Likewise.
22430 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
22431 Handle SVE constants. Use emit_move_insn to move a force_const_mem
22432 into the register, rather than emitting a SET directly.
22433 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
22434 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
22435 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
22436 (offset_9bit_signed_scaled_p): New functions.
22437 (aarch64_replicate_bitmask_imm): New function.
22438 (aarch64_bitmask_imm): Use it.
22439 (aarch64_cannot_force_const_mem): Reject expressions involving
22440 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
22441 (aarch64_classify_index): Handle SVE indices, by requiring
22442 a plain register index with a scale that matches the element size.
22443 (aarch64_classify_address): Handle SVE addresses. Assert that
22444 the mode of the address is VOIDmode or an integer mode.
22445 Update call to aarch64_classify_symbol.
22446 (aarch64_classify_symbolic_expression): Update call to
22447 aarch64_classify_symbol.
22448 (aarch64_const_vec_all_in_range_p): New function.
22449 (aarch64_print_vector_float_operand): Likewise.
22450 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
22451 "vN" for FP registers with SVE modes. Handle (const ...) vectors
22452 and the FP immediates 1.0 and 0.5.
22453 (aarch64_print_address_internal): Handle SVE addresses.
22454 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
22455 (aarch64_regno_regclass): Handle predicate registers.
22456 (aarch64_secondary_reload): Handle big-endian reloads of SVE
22458 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
22459 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
22460 (aarch64_convert_sve_vector_bits): New function.
22461 (aarch64_override_options): Use it to handle -msve-vector-bits=.
22462 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22463 rather than an rtx.
22464 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
22465 Handle SVE vector and predicate modes. Accept VL-based constants
22466 that need only one temporary register, and VL offsets that require
22467 no temporary registers.
22468 (aarch64_conditional_register_usage): Mark the predicate registers
22469 as fixed if SVE isn't available.
22470 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
22471 Return true for SVE vector and predicate modes.
22472 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
22473 rather than an unsigned int. Handle SVE modes.
22474 (aarch64_preferred_simd_mode): Update call accordingly. Handle
22476 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
22478 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22479 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22480 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22481 (aarch64_sve_float_mul_immediate_p): New functions.
22482 (aarch64_sve_valid_immediate): New function.
22483 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
22484 Explicitly reject structure modes. Check for INDEX constants.
22485 Handle PTRUE and PFALSE constants.
22486 (aarch64_check_zero_based_sve_index_immediate): New function.
22487 (aarch64_simd_imm_zero_p): Delete.
22488 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
22489 vector modes. Accept constants in the range of CNT[BHWD].
22490 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
22491 ask for an Advanced SIMD mode.
22492 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
22493 (aarch64_simd_vector_alignment): Handle SVE predicates.
22494 (aarch64_vectorize_preferred_vector_alignment): New function.
22495 (aarch64_simd_vector_alignment_reachable): Use it instead of
22497 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
22498 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
22500 (MAX_VECT_LEN): Delete.
22501 (expand_vec_perm_d): Add a vec_flags field.
22502 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
22503 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
22504 (aarch64_evpc_ext): Don't apply a big-endian lane correction
22506 (aarch64_evpc_rev): Rename to...
22507 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
22508 (aarch64_evpc_rev_global): New function.
22509 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
22510 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
22512 (aarch64_evpc_sve_tbl): New function.
22513 (aarch64_expand_vec_perm_const_1): Update after rename of
22514 aarch64_evpc_rev. Handle SVE permutes too, trying
22515 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
22516 than aarch64_evpc_tbl.
22517 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
22518 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
22519 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
22520 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
22521 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
22522 (aarch64_expand_sve_vcond): New functions.
22523 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
22524 of aarch64_vector_mode_p.
22525 (aarch64_dwarf_poly_indeterminate_value): New function.
22526 (aarch64_compute_pressure_classes): Likewise.
22527 (aarch64_can_change_mode_class): Likewise.
22528 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
22529 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
22530 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
22531 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
22532 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
22533 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
22534 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
22535 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
22537 (Dn, Dl, Dr): Accept const as well as const_vector.
22538 (Dz): Likewise. Compare against CONST0_RTX.
22539 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
22540 of "vector" where appropriate.
22541 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
22542 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
22543 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
22544 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
22545 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
22546 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
22547 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
22548 (v_int_equiv): Extend to SVE modes.
22549 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
22551 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
22552 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
22553 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
22554 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
22555 (SVE_COND_FP_CMP): New int iterators.
22556 (perm_hilo): Handle the new unpack unspecs.
22557 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
22559 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
22560 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
22561 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
22562 (aarch64_equality_operator, aarch64_constant_vector_operand)
22563 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
22564 (aarch64_sve_nonimmediate_operand): Likewise.
22565 (aarch64_sve_general_operand): Likewise.
22566 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
22567 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
22568 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
22569 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
22570 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
22571 (aarch64_sve_float_arith_immediate): Likewise.
22572 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
22573 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
22574 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
22575 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
22576 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
22577 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
22578 (aarch64_sve_float_arith_operand): Likewise.
22579 (aarch64_sve_float_arith_with_sub_operand): Likewise.
22580 (aarch64_sve_float_mul_operand): Likewise.
22581 (aarch64_sve_vec_perm_operand): Likewise.
22582 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
22583 (aarch64_mov_operand): Accept const_poly_int and const_vector.
22584 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
22585 as well as const_vector.
22586 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
22587 in file. Use CONST0_RTX and CONSTM1_RTX.
22588 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
22589 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
22590 Use aarch64_simd_imm_zero.
22591 * config/aarch64/aarch64-sve.md: New file.
22592 * config/aarch64/aarch64.md: Include it.
22593 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
22594 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
22595 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
22596 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
22597 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
22598 (sve): New attribute.
22599 (enabled): Disable instructions with the sve attribute unless
22601 (movqi, movhi): Pass CONST_POLY_INT operaneds through
22602 aarch64_expand_mov_immediate.
22603 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
22604 CNT[BHSD] immediates.
22605 (movti): Split CONST_POLY_INT moves into two halves.
22606 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
22607 Split additions that need a temporary here if the destination
22608 is the stack pointer.
22609 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
22610 (*add<mode>3_poly_1): New instruction.
22611 (set_clobber_cc): New expander.
22613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22615 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
22616 parameter and use it instead of GET_MODE_SIZE (innermode). Use
22617 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
22618 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
22619 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
22620 Change innermode from fixed_mode_size to machine_mode.
22621 (simplify_subreg): Update call accordingly. Handle a constant-sized
22622 subreg of a variable-length CONST_VECTOR.
22624 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
22625 Alan Hayward <alan.hayward@arm.com>
22626 David Sherwood <david.sherwood@arm.com>
22628 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
22629 (add_offset_to_base): New function, split out from...
22630 (create_mem_ref): ...here. When handling a scale other than 1,
22631 check first whether the address is valid without the offset.
22632 Add it into the base if so, leaving the index and scale as-is.
22634 2018-01-12 Jakub Jelinek <jakub@redhat.com>
22637 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
22638 fold_for_warn before checking if arg2 is INTEGER_CST.
22640 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
22642 * config/rs6000/predicates.md (load_multiple_operation): Delete.
22643 (store_multiple_operation): Delete.
22644 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
22645 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
22646 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
22647 guarded by TARGET_STRING.
22648 (rs6000_output_load_multiple): Delete.
22649 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
22650 OPTION_MASK_STRING / TARGET_STRING handling.
22651 (print_operand) <'N', 'O'>: Add comment that these are unused now.
22652 (const rs6000_opt_masks) <"string">: Change mask to 0.
22653 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
22654 (MASK_STRING): Delete.
22655 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
22657 (load_multiple): Delete.
22664 (store_multiple): Delete.
22671 (movmemsi_8reg): Delete.
22672 (corresponding unnamed define_insn): Delete.
22673 (movmemsi_6reg): Delete.
22674 (corresponding unnamed define_insn): Delete.
22675 (movmemsi_4reg): Delete.
22676 (corresponding unnamed define_insn): Delete.
22677 (movmemsi_2reg): Delete.
22678 (corresponding unnamed define_insn): Delete.
22679 (movmemsi_1reg): Delete.
22680 (corresponding unnamed define_insn): Delete.
22681 * config/rs6000/rs6000.opt (mno-string): New.
22682 (mstring): Replace by deprecation warning stub.
22683 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
22685 2018-01-12 Jakub Jelinek <jakub@redhat.com>
22687 * regrename.c (regrename_do_replace): If replacing the same
22688 reg multiple times, try to reuse last created gen_raw_REG.
22691 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
22692 main to workaround a bug in GDB.
22694 2018-01-12 Tom de Vries <tom@codesourcery.com>
22697 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
22699 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
22701 PR rtl-optimization/80481
22702 * ira-color.c (get_cap_member): New function.
22703 (allocnos_conflict_by_live_ranges_p): Use it.
22704 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
22705 (setup_slot_coalesced_allocno_live_ranges): Ditto.
22707 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
22710 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
22711 (*saddl_se_1): Ditto.
22712 (*ssubsi_1): Ditto.
22713 (*ssubl_se_1): Ditto.
22715 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22717 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
22718 rather than wi::to_widest for DR_INITs.
22719 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
22720 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
22721 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
22723 (vect_analyze_group_access_1): Note that here.
22725 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22727 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
22728 polynomial type sizes.
22730 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
22732 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
22733 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
22734 (gimple_add_tmp_var): Likewise.
22736 2018-01-12 Martin Liska <mliska@suse.cz>
22738 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
22739 (gimple_alloc_sizes): Likewise.
22740 (dump_gimple_statistics): Use PRIu64 in printf format.
22741 * gimple.h: Change uint64_t to int.
22743 2018-01-12 Martin Liska <mliska@suse.cz>
22745 * tree-core.h: Use uint64_t instead of int.
22746 * tree.c (tree_node_counts): Likewise.
22747 (tree_node_sizes): Likewise.
22748 (dump_tree_statistics): Use PRIu64 in printf format.
22750 2018-01-12 Martin Liska <mliska@suse.cz>
22752 * Makefile.in: As qsort_chk is implemented in vec.c, add
22753 vec.o to linkage of gencfn-macros.
22754 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
22755 passing the info to record_node_allocation_statistics.
22756 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
22758 * ggc-common.c (struct ggc_usage): Add operator== and use
22759 it in operator< and compare function.
22760 * mem-stats.h (struct mem_usage): Likewise.
22761 * vec.c (struct vec_usage): Remove operator< and compare
22762 function. Can be simply inherited.
22764 2018-01-12 Martin Jambor <mjambor@suse.cz>
22767 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
22768 * tree-ssa-math-opts.c: Include domwalk.h.
22769 (convert_mult_to_fma_1): New function.
22770 (fma_transformation_info): New type.
22771 (fma_deferring_state): Likewise.
22772 (cancel_fma_deferring): New function.
22773 (result_of_phi): Likewise.
22774 (last_fma_candidate_feeds_initial_phi): Likewise.
22775 (convert_mult_to_fma): Added deferring logic, split actual
22776 transformation to convert_mult_to_fma_1.
22777 (math_opts_dom_walker): New type.
22778 (math_opts_dom_walker::after_dom_children): New method, body moved
22779 here from pass_optimize_widening_mul::execute, added deferring logic
22781 (pass_optimize_widening_mul::execute): Moved most of code to
22782 math_opts_dom_walker::after_dom_children.
22783 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
22784 * config/i386/i386.c (ix86_option_override_internal): Added
22785 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
22787 2018-01-12 Richard Biener <rguenther@suse.de>
22790 * dwarf2out.c (gen_variable_die): Do not reset old_die for
22791 inline instance vars.
22793 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
22796 * config/rx/rx.c (rx_is_restricted_memory_address):
22797 Handle SUBREG case.
22799 2018-01-12 Richard Biener <rguenther@suse.de>
22801 PR tree-optimization/80846
22802 * target.def (split_reduction): New target hook.
22803 * targhooks.c (default_split_reduction): New function.
22804 * targhooks.h (default_split_reduction): Declare.
22805 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
22806 target requests first reduce vectors by combining low and high
22808 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
22809 (get_vectype_for_scalar_type_and_size): Export.
22810 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
22811 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
22812 * doc/tm.texi: Regenerate.
22813 * config/i386/i386.c (ix86_split_reduction): Implement
22814 TARGET_VECTORIZE_SPLIT_REDUCTION.
22816 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22819 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
22820 in PIC mode except for TARGET_VXWORKS_RTP.
22821 * config/sparc/sparc.c: Include cfgrtl.h.
22822 (TARGET_INIT_PIC_REG): Define.
22823 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
22824 (sparc_pic_register_p): New predicate.
22825 (sparc_legitimate_address_p): Use it.
22826 (sparc_legitimize_pic_address): Likewise.
22827 (sparc_delegitimize_address): Likewise.
22828 (sparc_mode_dependent_address_p): Likewise.
22829 (gen_load_pcrel_sym): Remove 4th parameter.
22830 (load_got_register): Adjust call to above. Remove obsolete stuff.
22831 (sparc_expand_prologue): Do not call load_got_register here.
22832 (sparc_flat_expand_prologue): Likewise.
22833 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
22834 (sparc_use_pseudo_pic_reg): New function.
22835 (sparc_init_pic_reg): Likewise.
22836 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22837 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
22839 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
22841 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
22842 Add item for branch_cost.
22844 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22846 PR rtl-optimization/83565
22847 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
22848 not extend the result to a larger mode for rotate operations.
22849 (num_sign_bit_copies1): Likewise.
22851 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22854 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
22856 Use values-Xc.o for -pedantic.
22857 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
22859 2018-01-12 Martin Liska <mliska@suse.cz>
22862 * ipa-devirt.c (final_warning_record::grow_type_warnings):
22864 (possible_polymorphic_call_targets): Use it.
22865 (ipa_devirt): Likewise.
22867 2018-01-12 Martin Liska <mliska@suse.cz>
22869 * profile-count.h (enum profile_quality): Use 0 as invalid
22870 enum value of profile_quality.
22872 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
22874 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
22875 -mext-string options.
22877 2018-01-12 Richard Biener <rguenther@suse.de>
22879 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
22880 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
22881 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
22883 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
22885 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
22887 * configure.ac (--with-long-double-format): Add support for the
22888 configuration option to change the default long double format on
22890 * config.gcc (powerpc*-linux*-*): Likewise.
22891 * configure: Regenerate.
22892 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
22893 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
22894 used without modification.
22896 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22898 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
22899 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
22900 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
22901 MISC_BUILTIN_SPEC_BARRIER.
22902 (rs6000_init_builtins): Likewise.
22903 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
22905 (speculation_barrier): New define_insn.
22906 * doc/extend.texi: Document __builtin_speculation_barrier.
22908 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22911 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
22912 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
22913 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
22915 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
22916 integral modes instead of "ss" and "sd".
22917 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
22918 vectors with 32-bit and 64-bit elements.
22919 (vecdupssescalarmodesuffix): New mode attribute.
22920 (vec_dup<mode>): Use it.
22922 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
22925 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
22926 frame if argument is passed on stack.
22928 2018-01-11 Jakub Jelinek <jakub@redhat.com>
22931 * ree.c (combine_reaching_defs): Optimize also
22932 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
22933 reg2=any_extend(exp); reg1=reg2;, formatting fix.
22935 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22937 PR middle-end/83189
22938 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
22940 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
22942 PR middle-end/83718
22943 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
22944 after they are computed.
22946 2018-01-11 Bin Cheng <bin.cheng@arm.com>
22948 PR tree-optimization/83695
22949 * gimple-loop-linterchange.cc
22950 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
22951 reset cached scev information after interchange.
22952 (pass_linterchange::execute): Remove call to scev_reset_htab.
22954 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22956 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
22957 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
22958 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
22959 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
22960 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
22961 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
22962 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
22963 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
22964 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
22965 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
22966 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
22967 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
22968 (V_lane_reg): Likewise.
22969 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
22971 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
22972 (vfmal_lane_low<mode>_intrinsic,
22973 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
22974 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
22975 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
22976 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
22977 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
22978 vfmsl_lane_high<mode>_intrinsic): New define_insns.
22980 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22982 * config/arm/arm-cpus.in (fp16fml): New feature.
22983 (ALL_SIMD): Add fp16fml.
22984 (armv8.2-a): Add fp16fml as an option.
22985 (armv8.3-a): Likewise.
22986 (armv8.4-a): Add fp16fml as part of fp16.
22987 * config/arm/arm.h (TARGET_FP16FML): Define.
22988 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
22990 * config/arm/arm-modes.def (V2HF): Define.
22991 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
22992 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
22993 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
22994 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
22995 vfmsl_low, vfmsl_high): New set of builtins.
22996 * config/arm/iterators.md (PLUSMINUS): New code iterator.
22997 (vfml_op): New code attribute.
22998 (VFMLHALVES): New int iterator.
22999 (VFML, VFMLSEL): New mode attributes.
23000 (V_reg): Define mapping for V2HF.
23001 (V_hi, V_lo): New mode attributes.
23002 (VF_constraint): Likewise.
23003 (vfml_half, vfml_half_selector): New int attributes.
23004 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
23006 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
23007 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
23009 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
23010 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
23011 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
23012 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
23014 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
23015 Document new effective target and option set.
23017 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23019 * config/arm/arm-cpus.in (armv8_4): New feature.
23020 (ARMv8_4a): New fgroup.
23021 (armv8.4-a): New arch.
23022 * config/arm/arm-tables.opt: Regenerate.
23023 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
23024 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
23025 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
23026 Add matching rules for -march=armv8.4-a and extensions.
23027 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
23029 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
23032 * config/rx/rx.md (BW): New mode attribute.
23033 (sync_lock_test_and_setsi): Add mode suffix to insn output.
23035 2018-01-11 Richard Biener <rguenther@suse.de>
23037 PR tree-optimization/83435
23038 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
23039 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
23040 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
23042 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23043 Alan Hayward <alan.hayward@arm.com>
23044 David Sherwood <david.sherwood@arm.com>
23046 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
23048 (aarch64_classify_address): Initialize it. Track polynomial offsets.
23049 (aarch64_print_address_internal): Use it to check for a zero offset.
23051 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23052 Alan Hayward <alan.hayward@arm.com>
23053 David Sherwood <david.sherwood@arm.com>
23055 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
23056 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
23057 Return a poly_int64 rather than a HOST_WIDE_INT.
23058 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
23059 rather than a HOST_WIDE_INT.
23060 * config/aarch64/aarch64.h (aarch64_frame): Protect with
23061 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
23062 hard_fp_offset, frame_size, initial_adjust, callee_offset and
23063 final_offset from HOST_WIDE_INT to poly_int64.
23064 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
23065 to_constant when getting the number of units in an Advanced SIMD
23067 (aarch64_builtin_vectorized_function): Check for a constant number
23069 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
23071 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
23072 attribute instead of GET_MODE_NUNITS.
23073 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
23074 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
23075 GET_MODE_SIZE for fixed-size registers.
23076 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
23077 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
23078 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
23079 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
23080 (aarch64_print_operand, aarch64_print_address_internal)
23081 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
23082 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
23083 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
23084 Handle polynomial GET_MODE_SIZE.
23085 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
23086 wider than SImode without modification.
23087 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
23088 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
23089 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
23090 passing and returning SVE modes.
23091 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
23092 rather than GEN_INT.
23093 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
23094 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
23095 (aarch64_allocate_and_probe_stack_space): Likewise.
23096 (aarch64_layout_frame): Cope with polynomial offsets.
23097 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
23098 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
23099 polynomial offsets.
23100 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
23101 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
23102 poly_int64 rather than a HOST_WIDE_INT.
23103 (aarch64_get_separate_components, aarch64_process_components)
23104 (aarch64_expand_prologue, aarch64_expand_epilogue)
23105 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
23106 (aarch64_anchor_offset): New function, split out from...
23107 (aarch64_legitimize_address): ...here.
23108 (aarch64_builtin_vectorization_cost): Handle polynomial
23109 TYPE_VECTOR_SUBPARTS.
23110 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
23112 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
23113 number of elements from the PARALLEL rather than the mode.
23114 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
23115 rather than GET_MODE_BITSIZE.
23116 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
23117 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
23118 (aarch64_expand_vec_perm_const_1): Handle polynomial
23119 d->perm.length () and d->perm elements.
23120 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
23121 Apply to_constant to d->perm elements.
23122 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
23123 polynomial CONST_VECTOR_NUNITS.
23124 (aarch64_move_pointer): Take amount as a poly_int64 rather
23126 (aarch64_progress_pointer): Avoid temporary variable.
23127 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
23128 the mode attribute instead of GET_MODE.
23130 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23131 Alan Hayward <alan.hayward@arm.com>
23132 David Sherwood <david.sherwood@arm.com>
23134 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
23135 x exists before using it.
23136 (aarch64_add_constant_internal): Rename to...
23137 (aarch64_add_offset_1): ...this. Replace regnum with separate
23138 src and dest rtxes. Handle the case in which they're different,
23139 including when the offset is zero. Replace scratchreg with an rtx.
23140 Use 2 additions if there is no spare register into which we can
23141 move a 16-bit constant.
23142 (aarch64_add_constant): Delete.
23143 (aarch64_add_offset): Replace reg with separate src and dest
23144 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
23145 Use aarch64_add_offset_1.
23146 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
23147 an rtx rather than an int. Take the delta as a poly_int64
23148 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
23149 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
23150 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
23151 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
23152 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
23153 and aarch64_add_sp.
23154 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
23155 aarch64_add_constant.
23157 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23159 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
23160 Use scalar_float_mode.
23162 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23164 * config/aarch64/aarch64-simd.md
23165 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
23166 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
23167 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
23168 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
23169 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
23170 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
23171 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
23172 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
23173 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
23174 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
23176 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23179 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
23180 targ_options->x_arm_arch_string is non NULL.
23182 2018-01-11 Tamar Christina <tamar.christina@arm.com>
23184 * config/aarch64/aarch64.h
23185 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
23187 2018-01-11 Sudakshina Das <sudi.das@arm.com>
23190 * expmed.c (emit_store_flag_force): Swap if const op0
23191 and change VOIDmode to mode of op0.
23193 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
23195 PR rtl-optimization/83761
23196 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
23197 than bytes to mode_for_size.
23199 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23201 PR middle-end/83189
23202 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
23203 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
23206 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23208 PR middle-end/83575
23209 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
23210 when in layout mode.
23211 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
23212 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
23215 2018-01-10 Michael Collison <michael.collison@arm.com>
23217 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
23218 * config/aarch64/aarch64-option-extension.def: Add
23219 AARCH64_OPT_EXTENSION of 'fp16fml'.
23220 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23221 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
23222 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
23223 * config/aarch64/constraints.md (Ui7): New constraint.
23224 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
23225 (VFMLA_SEL_W): Ditto.
23228 (VFMLA16_LOW): New int iterator.
23229 (VFMLA16_HIGH): Ditto.
23230 (UNSPEC_FMLAL): New unspec.
23231 (UNSPEC_FMLSL): Ditto.
23232 (UNSPEC_FMLAL2): Ditto.
23233 (UNSPEC_FMLSL2): Ditto.
23234 (f16mac): New code attribute.
23235 * config/aarch64/aarch64-simd-builtins.def
23236 (aarch64_fmlal_lowv2sf): Ditto.
23237 (aarch64_fmlsl_lowv2sf): Ditto.
23238 (aarch64_fmlalq_lowv4sf): Ditto.
23239 (aarch64_fmlslq_lowv4sf): Ditto.
23240 (aarch64_fmlal_highv2sf): Ditto.
23241 (aarch64_fmlsl_highv2sf): Ditto.
23242 (aarch64_fmlalq_highv4sf): Ditto.
23243 (aarch64_fmlslq_highv4sf): Ditto.
23244 (aarch64_fmlal_lane_lowv2sf): Ditto.
23245 (aarch64_fmlsl_lane_lowv2sf): Ditto.
23246 (aarch64_fmlal_laneq_lowv2sf): Ditto.
23247 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
23248 (aarch64_fmlalq_lane_lowv4sf): Ditto.
23249 (aarch64_fmlsl_lane_lowv4sf): Ditto.
23250 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
23251 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
23252 (aarch64_fmlal_lane_highv2sf): Ditto.
23253 (aarch64_fmlsl_lane_highv2sf): Ditto.
23254 (aarch64_fmlal_laneq_highv2sf): Ditto.
23255 (aarch64_fmlsl_laneq_highv2sf): Ditto.
23256 (aarch64_fmlalq_lane_highv4sf): Ditto.
23257 (aarch64_fmlsl_lane_highv4sf): Ditto.
23258 (aarch64_fmlalq_laneq_highv4sf): Ditto.
23259 (aarch64_fmlsl_laneq_highv4sf): Ditto.
23260 * config/aarch64/aarch64-simd.md:
23261 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
23262 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23263 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
23264 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23265 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
23266 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
23267 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
23268 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
23269 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
23270 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
23271 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
23272 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
23273 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
23274 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
23275 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
23276 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
23277 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
23278 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
23279 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
23280 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
23281 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
23282 (vfmlsl_low_u32): Ditto.
23283 (vfmlalq_low_u32): Ditto.
23284 (vfmlslq_low_u32): Ditto.
23285 (vfmlal_high_u32): Ditto.
23286 (vfmlsl_high_u32): Ditto.
23287 (vfmlalq_high_u32): Ditto.
23288 (vfmlslq_high_u32): Ditto.
23289 (vfmlal_lane_low_u32): Ditto.
23290 (vfmlsl_lane_low_u32): Ditto.
23291 (vfmlal_laneq_low_u32): Ditto.
23292 (vfmlsl_laneq_low_u32): Ditto.
23293 (vfmlalq_lane_low_u32): Ditto.
23294 (vfmlslq_lane_low_u32): Ditto.
23295 (vfmlalq_laneq_low_u32): Ditto.
23296 (vfmlslq_laneq_low_u32): Ditto.
23297 (vfmlal_lane_high_u32): Ditto.
23298 (vfmlsl_lane_high_u32): Ditto.
23299 (vfmlal_laneq_high_u32): Ditto.
23300 (vfmlsl_laneq_high_u32): Ditto.
23301 (vfmlalq_lane_high_u32): Ditto.
23302 (vfmlslq_lane_high_u32): Ditto.
23303 (vfmlalq_laneq_high_u32): Ditto.
23304 (vfmlslq_laneq_high_u32): Ditto.
23305 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
23306 (AARCH64_FL_FOR_ARCH8_4): New.
23307 (AARCH64_ISA_F16FML): New ISA flag.
23308 (TARGET_F16FML): New feature flag for fp16fml.
23309 (doc/invoke.texi): Document new fp16fml option.
23311 2018-01-10 Michael Collison <michael.collison@arm.com>
23313 * config/aarch64/aarch64-builtins.c:
23314 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
23315 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23316 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
23317 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
23318 (AARCH64_ISA_SHA3): New ISA flag.
23319 (TARGET_SHA3): New feature flag for sha3.
23320 * config/aarch64/iterators.md (sha512_op): New int attribute.
23321 (CRYPTO_SHA512): New int iterator.
23322 (UNSPEC_SHA512H): New unspec.
23323 (UNSPEC_SHA512H2): Ditto.
23324 (UNSPEC_SHA512SU0): Ditto.
23325 (UNSPEC_SHA512SU1): Ditto.
23326 * config/aarch64/aarch64-simd-builtins.def
23327 (aarch64_crypto_sha512hqv2di): New builtin.
23328 (aarch64_crypto_sha512h2qv2di): Ditto.
23329 (aarch64_crypto_sha512su0qv2di): Ditto.
23330 (aarch64_crypto_sha512su1qv2di): Ditto.
23331 (aarch64_eor3qv8hi): Ditto.
23332 (aarch64_rax1qv2di): Ditto.
23333 (aarch64_xarqv2di): Ditto.
23334 (aarch64_bcaxqv8hi): Ditto.
23335 * config/aarch64/aarch64-simd.md:
23336 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
23337 (aarch64_crypto_sha512su0qv2di): Ditto.
23338 (aarch64_crypto_sha512su1qv2di): Ditto.
23339 (aarch64_eor3qv8hi): Ditto.
23340 (aarch64_rax1qv2di): Ditto.
23341 (aarch64_xarqv2di): Ditto.
23342 (aarch64_bcaxqv8hi): Ditto.
23343 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
23344 (vsha512h2q_u64): Ditto.
23345 (vsha512su0q_u64): Ditto.
23346 (vsha512su1q_u64): Ditto.
23347 (veor3q_u16): Ditto.
23348 (vrax1q_u64): Ditto.
23349 (vxarq_u64): Ditto.
23350 (vbcaxq_u16): Ditto.
23351 * config/arm/types.md (crypto_sha512): New type attribute.
23352 (crypto_sha3): Ditto.
23353 (doc/invoke.texi): Document new sha3 option.
23355 2018-01-10 Michael Collison <michael.collison@arm.com>
23357 * config/aarch64/aarch64-builtins.c:
23358 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
23359 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23360 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
23361 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
23362 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
23363 (AARCH64_ISA_SM4): New ISA flag.
23364 (TARGET_SM4): New feature flag for sm4.
23365 * config/aarch64/aarch64-simd-builtins.def
23366 (aarch64_sm3ss1qv4si): Ditto.
23367 (aarch64_sm3tt1aq4si): Ditto.
23368 (aarch64_sm3tt1bq4si): Ditto.
23369 (aarch64_sm3tt2aq4si): Ditto.
23370 (aarch64_sm3tt2bq4si): Ditto.
23371 (aarch64_sm3partw1qv4si): Ditto.
23372 (aarch64_sm3partw2qv4si): Ditto.
23373 (aarch64_sm4eqv4si): Ditto.
23374 (aarch64_sm4ekeyqv4si): Ditto.
23375 * config/aarch64/aarch64-simd.md:
23376 (aarch64_sm3ss1qv4si): Ditto.
23377 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
23378 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
23379 (aarch64_sm4eqv4si): Ditto.
23380 (aarch64_sm4ekeyqv4si): Ditto.
23381 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
23382 (sm3part_op): Ditto.
23383 (CRYPTO_SM3TT): Ditto.
23384 (CRYPTO_SM3PART): Ditto.
23385 (UNSPEC_SM3SS1): New unspec.
23386 (UNSPEC_SM3TT1A): Ditto.
23387 (UNSPEC_SM3TT1B): Ditto.
23388 (UNSPEC_SM3TT2A): Ditto.
23389 (UNSPEC_SM3TT2B): Ditto.
23390 (UNSPEC_SM3PARTW1): Ditto.
23391 (UNSPEC_SM3PARTW2): Ditto.
23392 (UNSPEC_SM4E): Ditto.
23393 (UNSPEC_SM4EKEY): Ditto.
23394 * config/aarch64/constraints.md (Ui2): New constraint.
23395 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
23396 * config/arm/types.md (crypto_sm3): New type attribute.
23397 (crypto_sm4): Ditto.
23398 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
23399 (vsm3tt1aq_u32): Ditto.
23400 (vsm3tt1bq_u32): Ditto.
23401 (vsm3tt2aq_u32): Ditto.
23402 (vsm3tt2bq_u32): Ditto.
23403 (vsm3partw1q_u32): Ditto.
23404 (vsm3partw2q_u32): Ditto.
23405 (vsm4eq_u32): Ditto.
23406 (vsm4ekeyq_u32): Ditto.
23407 (doc/invoke.texi): Document new sm4 option.
23409 2018-01-10 Michael Collison <michael.collison@arm.com>
23411 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
23412 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
23413 (AARCH64_FL_FOR_ARCH8_4): New.
23414 (AARCH64_FL_V8_4): New flag.
23415 (doc/invoke.texi): Document new armv8.4-a option.
23417 2018-01-10 Michael Collison <michael.collison@arm.com>
23419 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23420 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
23421 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
23422 * config/aarch64/aarch64-option-extension.def: Add
23423 AARCH64_OPT_EXTENSION of 'sha2'.
23424 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
23425 (crypto): Disable sha2 and aes if crypto disabled.
23426 (crypto): Enable aes and sha2 if enabled.
23427 (simd): Disable sha2 and aes if simd disabled.
23428 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
23430 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
23431 (TARGET_SHA2): New feature flag for sha2.
23432 (TARGET_AES): New feature flag for aes.
23433 * config/aarch64/aarch64-simd.md:
23434 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
23435 conditional on TARGET_AES.
23436 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
23437 (aarch64_crypto_sha1hsi): Make pattern conditional
23439 (aarch64_crypto_sha1hv4si): Ditto.
23440 (aarch64_be_crypto_sha1hv4si): Ditto.
23441 (aarch64_crypto_sha1su1v4si): Ditto.
23442 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
23443 (aarch64_crypto_sha1su0v4si): Ditto.
23444 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
23445 (aarch64_crypto_sha256su0v4si): Ditto.
23446 (aarch64_crypto_sha256su1v4si): Ditto.
23447 (doc/invoke.texi): Document new aes and sha2 options.
23449 2018-01-10 Martin Sebor <msebor@redhat.com>
23451 PR tree-optimization/83781
23452 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
23455 2018-01-11 Martin Sebor <msebor@gmail.com>
23456 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23458 PR tree-optimization/83501
23459 PR tree-optimization/81703
23461 * tree-ssa-strlen.c (get_string_cst): Rename...
23462 (get_string_len): ...to this. Handle global constants.
23463 (handle_char_store): Adjust.
23465 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
23466 Jim Wilson <jimw@sifive.com>
23468 * config/riscv/riscv-protos.h (riscv_output_return): New.
23469 * config/riscv/riscv.c (struct machine_function): New naked_p field.
23470 (riscv_attribute_table, riscv_output_return),
23471 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
23472 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
23473 (riscv_compute_frame_info): Only compute frame->mask if not a naked
23475 (riscv_expand_prologue): Add early return for naked function.
23476 (riscv_expand_epilogue): Likewise.
23477 (riscv_function_ok_for_sibcall): Return false for naked function.
23478 (riscv_set_current_function): New.
23479 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
23480 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
23481 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
23482 * doc/extend.texi (RISC-V Function Attributes): New.
23484 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
23486 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
23487 check for 128-bit long double before checking TCmode.
23488 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
23489 128-bit long doubles before checking TFmode or TCmode.
23490 (FLOAT128_IBM_P): Likewise.
23492 2018-01-10 Martin Sebor <msebor@redhat.com>
23494 PR tree-optimization/83671
23495 * builtins.c (c_strlen): Unconditionally return zero for the empty
23497 Use -Warray-bounds for warnings.
23498 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
23499 for non-constant array indices with COMPONENT_REF, arrays of
23500 arrays, and pointers to arrays.
23501 (gimple_fold_builtin_strlen): Determine and set length range for
23502 non-constant character arrays.
23504 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
23506 PR middle-end/81897
23507 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
23510 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
23512 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
23514 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23517 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
23518 VECTOR_MEM_ALTIVEC_OR_VSX_P.
23519 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
23520 indexed_or_indirect_operand predicate.
23521 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
23522 (*vsx_le_perm_load_v8hi): Likewise.
23523 (*vsx_le_perm_load_v16qi): Likewise.
23524 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
23525 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
23526 (*vsx_le_perm_store_v8hi): Likewise.
23527 (*vsx_le_perm_store_v16qi): Likewise.
23528 (eight unnamed splitters): Likewise.
23530 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
23532 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
23533 * config/rs6000/emmintrin.h: Likewise.
23534 * config/rs6000/mmintrin.h: Likewise.
23535 * config/rs6000/xmmintrin.h: Likewise.
23537 2018-01-10 David Malcolm <dmalcolm@redhat.com>
23540 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
23542 * tree.c (tree_nop_conversion): Return true for location wrapper
23544 (maybe_wrap_with_location): New function.
23545 (selftest::check_strip_nops): New function.
23546 (selftest::test_location_wrappers): New function.
23547 (selftest::tree_c_tests): Call it.
23548 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
23549 (maybe_wrap_with_location): New decl.
23550 (EXPR_LOCATION_WRAPPER_P): New macro.
23551 (location_wrapper_p): New inline function.
23552 (tree_strip_any_location_wrapper): New inline function.
23554 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
23557 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
23558 stack_realign_offset for the largest alignment of stack slot
23560 (ix86_find_max_used_stack_alignment): New function.
23561 (ix86_finalize_stack_frame_flags): Use it. Set
23562 max_used_stack_alignment if we don't realign stack.
23563 * config/i386/i386.h (machine_function): Add
23564 max_used_stack_alignment.
23566 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
23568 * config/arm/arm.opt (-mbranch-cost): New option.
23569 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
23572 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
23575 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
23576 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
23578 2018-01-10 Richard Biener <rguenther@suse.de>
23581 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
23582 early out so it also covers the case where we have a non-NULL
23585 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23587 PR tree-optimization/83753
23588 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
23589 for non-strided grouped accesses if the number of elements is 1.
23591 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
23594 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
23595 * i386.h (TARGET_USE_GATHER): Define.
23596 * x86-tune.def (X86_TUNE_USE_GATHER): New.
23598 2018-01-10 Martin Liska <mliska@suse.cz>
23601 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
23602 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
23604 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
23605 CLEANUP_NO_PARTITIONING is not set.
23607 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
23609 * doc/rtl.texi: Remove documentation of (const ...) wrappers
23610 for vectors, as a partial revert of r254296.
23611 * rtl.h (const_vec_p): Delete.
23612 (const_vec_duplicate_p): Don't test for vector CONSTs.
23613 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
23614 * expmed.c (make_tree): Likewise.
23617 * common.md (E, F): Use CONSTANT_P instead of checking for
23619 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
23620 checking for CONST_VECTOR.
23622 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23624 PR middle-end/83575
23625 * predict.c (force_edge_cold): Handle in more sane way edges
23626 with no prediction.
23628 2018-01-09 Carl Love <cel@us.ibm.com>
23630 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
23632 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
23633 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
23634 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
23635 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
23636 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
23637 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
23638 * config/rs6000/rs6000-protos.h: Add extern defition for
23639 rs6000_generate_float2_double_code.
23640 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
23642 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
23643 (float2_v2df): Add define_expand.
23645 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
23648 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
23649 op_mode in the force_to_mode call.
23651 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23653 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
23654 instead of checking each element individually.
23655 (aarch64_evpc_uzp): Likewise.
23656 (aarch64_evpc_zip): Likewise.
23657 (aarch64_evpc_ext): Likewise.
23658 (aarch64_evpc_rev): Likewise.
23659 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
23660 instead of checking each element individually. Return true without
23662 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
23663 whether all selected elements come from the same input, instead of
23664 checking each element individually. Remove calls to gen_rtx_REG,
23665 start_sequence and end_sequence and instead assert that no rtl is
23668 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23670 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
23671 order of HIGH and CONST checks.
23673 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
23675 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
23676 if the destination isn't an SSA_NAME.
23678 2018-01-09 Richard Biener <rguenther@suse.de>
23680 PR tree-optimization/83668
23681 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
23683 (canonicalize_loop_form): ... here, renamed from ...
23684 (canonicalize_loop_closed_ssa_form): ... this and amended to
23685 swap successor edges for loop exit blocks to make us use
23686 the RPO order we need for initial schedule generation.
23688 2018-01-09 Joseph Myers <joseph@codesourcery.com>
23690 PR tree-optimization/64811
23691 * match.pd: When optimizing comparisons with Inf, avoid
23692 introducing or losing exceptions from comparisons with NaN.
23694 2018-01-09 Martin Liska <mliska@suse.cz>
23697 * asan.c (shadow_mem_size): Add gcc_assert.
23699 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
23701 Don't save registers in main().
23704 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
23705 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
23706 * config/avr/avr.c (avr_set_current_function): Don't error if
23707 naked, OS_task or OS_main are specified at the same time.
23708 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
23710 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
23712 * common/config/avr/avr-common.c (avr_option_optimization_table):
23713 Switch on -mmain-is-OS_task for optimizing compilations.
23715 2018-01-09 Richard Biener <rguenther@suse.de>
23717 PR tree-optimization/83572
23718 * graphite.c: Include cfganal.h.
23719 (graphite_transform_loops): Connect infinite loops to exit
23720 and remove fake edges at the end.
23722 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23724 * ipa-inline.c (edge_badness): Revert accidental checkin.
23726 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
23729 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
23730 symbols; not inline clones.
23732 2018-01-09 Jakub Jelinek <jakub@redhat.com>
23735 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
23736 hard registers. Formatting fixes.
23738 PR preprocessor/83722
23739 * gcc.c (try_generate_repro): Pass
23740 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
23741 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
23744 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
23745 Kito Cheng <kito.cheng@gmail.com>
23747 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
23748 (riscv_leaf_function_p): Delete.
23749 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
23751 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23753 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
23755 (do_ifelse): New function.
23756 (do_isel): New function.
23757 (do_sub3): New function.
23758 (do_add3): New function.
23759 (do_load_mask_compare): New function.
23760 (do_overlap_load_compare): New function.
23761 (expand_compare_loop): New function.
23762 (expand_block_compare): Call expand_compare_loop() when appropriate.
23763 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
23764 option description.
23765 (-mblock-compare-inline-loop-limit): New option.
23767 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23770 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
23771 Reverse order of second and third operands in first alternative.
23772 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
23773 of first and second elements in UNSPEC_VPERMR vector.
23774 (altivec_expand_vec_perm_le): Likewise.
23776 2018-01-08 Jeff Law <law@redhat.com>
23778 PR rtl-optimizatin/81308
23779 * tree-switch-conversion.c (cfg_altered): New file scoped static.
23780 (process_switch): If group_case_labels makes a change, then set
23782 (pass_convert_switch::execute): If a switch is converted, then
23783 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
23785 PR rtl-optimization/81308
23786 * recog.c (split_all_insns): Conditionally cleanup the CFG after
23789 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
23791 PR target/83663 - Revert r255946
23792 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
23793 generation for cases where splatting a value is not useful.
23794 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
23795 across a vec_duplicate and a paradoxical subreg forming a vector
23796 mode to a vec_concat.
23798 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23800 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
23801 -march=armv8.3-a variants.
23802 * config/arm/t-multilib: Likewise.
23803 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
23805 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23807 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
23809 (cceq_ior_compare_complement): Give it a name so I can use it, and
23810 change boolean_or_operator predicate to boolean_operator so it can
23811 be used to generate a crand.
23812 (eqne): New code iterator.
23813 (bd/bd_neg): New code_attrs.
23814 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
23815 a single define_insn.
23816 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
23817 decrement (bdnzt/bdnzf/bdzt/bdzf).
23818 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
23819 with the new names of the branch decrement patterns, and added the
23820 names of the branch decrement conditional patterns.
23822 2018-01-08 Richard Biener <rguenther@suse.de>
23824 PR tree-optimization/83563
23825 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
23828 2018-01-08 Richard Biener <rguenther@suse.de>
23830 PR middle-end/83713
23831 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
23833 2018-01-08 Richard Biener <rguenther@suse.de>
23835 PR tree-optimization/83685
23836 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
23837 references to abnormals.
23839 2018-01-08 Richard Biener <rguenther@suse.de>
23842 * dwarf2out.c (output_indirect_strings): Handle empty
23843 skeleton_debug_str_hash.
23844 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
23846 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23848 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
23849 (emit_store_direct): Likewise.
23850 (arc_trampoline_adjust_address): Likewise.
23851 (arc_asm_trampoline_template): New function.
23852 (arc_initialize_trampoline): Use asm_trampoline_template.
23853 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
23854 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
23855 * config/arc/arc.md (flush_icache): Delete pattern.
23857 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
23859 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
23860 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
23863 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23866 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
23867 by not USED_FOR_TARGET.
23868 (make_pass_resolve_sw_modes): Likewise.
23870 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
23872 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
23875 2018-01-08 Richard Biener <rguenther@suse.de>
23877 PR middle-end/83580
23878 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
23880 2018-01-08 Richard Biener <rguenther@suse.de>
23882 PR middle-end/83517
23883 * match.pd ((t * 2) / 2) -> t): Add missing :c.
23885 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
23887 PR middle-end/81897
23888 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
23889 basic blocks with a small number of successors.
23890 (convert_control_dep_chain_into_preds): Improve handling of
23892 (dump_predicates): Split apart into...
23893 (dump_pred_chain): ...here...
23894 (dump_pred_info): ...and here.
23895 (can_one_predicate_be_invalidated_p): Add debugging printfs.
23896 (can_chain_union_be_invalidated_p): Improve check for invalidation
23898 (uninit_uses_cannot_happen): Avoid unnecessary if
23899 convert_control_dep_chain_into_preds yielded nothing.
23901 2018-01-06 Martin Sebor <msebor@redhat.com>
23903 PR tree-optimization/83640
23904 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
23905 subtracting negative offset from size.
23906 (builtin_access::overlap): Adjust offset bounds of the access to fall
23907 within the size of the object if possible.
23909 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
23911 PR rtl-optimization/83699
23912 * expmed.c (extract_bit_field_1): Restrict the vector usage of
23913 extract_bit_field_as_subreg to cases in which the extracted
23914 value is also a vector.
23916 * lra-constraints.c (process_alt_operands): Test for the equivalence
23917 substitutions when detecting a possible reload cycle.
23919 2018-01-06 Jakub Jelinek <jakub@redhat.com>
23922 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
23923 by default if flag_selective_schedling{,2}. Formatting fixes.
23925 PR rtl-optimization/83682
23926 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
23927 if it has non-VECTOR_MODE element mode.
23928 (vec_duplicate_p): Likewise.
23930 PR middle-end/83694
23931 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
23932 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
23934 2018-01-05 Jakub Jelinek <jakub@redhat.com>
23937 * config/i386/i386-builtin.def
23938 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
23939 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
23940 Require also OPTION_MASK_ISA_AVX512F in addition to
23941 OPTION_MASK_ISA_GFNI.
23942 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
23943 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
23944 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
23945 to OPTION_MASK_ISA_GFNI.
23946 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
23947 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
23948 OPTION_MASK_ISA_AVX512BW.
23949 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
23950 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
23951 addition to OPTION_MASK_ISA_GFNI.
23952 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
23953 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
23954 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
23955 to OPTION_MASK_ISA_GFNI.
23956 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
23957 a requirement for all ISAs rather than any of them with a few
23959 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
23961 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
23962 bitmasks to be enabled with 3 exceptions, instead of requiring any
23963 enabled ISA with lots of exceptions.
23964 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
23965 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
23966 Change avx512bw in isa attribute to avx512f.
23967 * config/i386/sgxintrin.h: Add license boilerplate.
23968 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
23969 to __AVX512F__ and __AVX512VL to __AVX512VL__.
23970 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
23971 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
23973 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
23974 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
23975 temporarily sse2 rather than sse if not enabled already.
23978 * config/i386/sse.md (VI248_VLBW): Rename to ...
23979 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
23980 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
23981 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
23982 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
23983 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
23984 mode iterator instead of VI248_VLBW.
23986 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
23988 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
23989 (record_modified): Skip clobbers; add debug output.
23990 (param_change_prob): Use sreal frequencies.
23992 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23994 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
23995 punt for user-aligned variables.
23997 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
23999 * tree-chrec.c (chrec_contains_symbols): Return true for
24002 2018-01-05 Sudakshina Das <sudi.das@arm.com>
24005 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
24006 of (x|y) == x for BICS pattern.
24008 2018-01-05 Jakub Jelinek <jakub@redhat.com>
24010 PR tree-optimization/83605
24011 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
24012 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
24015 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
24017 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
24018 * config/epiphany/rtems.h: New file.
24020 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24021 Uros Bizjak <ubizjak@gmail.com>
24024 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
24025 QIreg_operand instead of register_operand predicate.
24026 * config/i386/i386.c (ix86_rop_should_change_byte_p,
24027 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
24028 comments instead of -fmitigate[-_]rop.
24030 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24033 * cgraphunit.c (symbol_table::compile): Switch to text_section
24034 before calling assembly_start debug hook.
24035 * run-rtl-passes.c (run_rtl_passes): Likewise.
24038 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24040 * tree-vrp.c (extract_range_from_binary_expr_1): Check
24041 range_int_cst_p rather than !symbolic_range_p before calling
24042 extract_range_from_multiplicative_op_1.
24044 2018-01-04 Jeff Law <law@redhat.com>
24046 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
24047 redundant test in assertion.
24049 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24051 * doc/rtl.texi: Document machine_mode wrapper classes.
24053 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24055 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
24056 using tree_to_uhwi.
24058 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24060 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
24061 the VEC_PERM_EXPR fold to fail.
24063 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24066 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
24067 to switched_sections.
24069 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24072 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
24073 test for d.testing.
24075 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
24078 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
24079 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
24081 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24084 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
24085 is BLKmode and bitpos not zero or mode change is needed.
24087 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
24090 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
24093 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
24096 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
24097 instead of MULT rtx. Update all corresponding splitters.
24098 (*saddl_se): Ditto.
24099 (*ssub<modesuffix>): Ditto.
24100 (*ssubl_se): Ditto.
24101 (*cmp_sadd_di): Update split patterns.
24102 (*cmp_sadd_si): Ditto.
24103 (*cmp_sadd_sidi): Ditto.
24104 (*cmp_ssub_di): Ditto.
24105 (*cmp_ssub_si): Ditto.
24106 (*cmp_ssub_sidi): Ditto.
24107 * config/alpha/predicates.md (const23_operand): New predicate.
24108 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
24109 Look for ASHIFT, not MULT inner operand.
24110 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
24112 2018-01-04 Martin Liska <mliska@suse.cz>
24114 PR gcov-profile/83669
24115 * gcov.c (output_intermediate_file): Add version to intermediate
24117 * doc/gcov.texi: Document new field 'version' in intermediate
24118 file format. Fix location of '-k' option of gcov command.
24120 2018-01-04 Martin Liska <mliska@suse.cz>
24123 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
24125 2018-01-04 Jakub Jelinek <jakub@redhat.com>
24127 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
24129 2018-01-03 Martin Sebor <msebor@redhat.com>
24131 PR tree-optimization/83655
24132 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
24133 checking calls with invalid arguments.
24135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24137 * tree-vect-stmts.c (vect_get_store_rhs): New function.
24138 (vectorizable_mask_load_store): Delete.
24139 (vectorizable_call): Return false for masked loads and stores.
24140 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
24141 instead of gimple_assign_rhs1.
24142 (vectorizable_load): Handle IFN_MASK_LOAD.
24143 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
24145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24147 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
24149 (vectorizable_mask_load_store): ...here.
24150 (vectorizable_load): ...and here.
24152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24154 * tree-vect-stmts.c (vect_build_all_ones_mask)
24155 (vect_build_zero_merge_argument): New functions, split out from...
24156 (vectorizable_load): ...here.
24158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24160 * tree-vect-stmts.c (vect_check_store_rhs): New function,
24162 (vectorizable_mask_load_store): ...here.
24163 (vectorizable_store): ...and here.
24165 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24167 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
24169 (vectorizable_mask_load_store): ...here.
24171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24173 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
24174 (vect_model_store_cost): Take a vec_load_store_type instead of a
24176 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
24177 (vect_model_store_cost): Take a vec_load_store_type instead of a
24179 (vectorizable_mask_load_store): Update accordingly.
24180 (vectorizable_store): Likewise.
24181 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
24183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24185 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
24186 IFN_MASK_LOAD calls here rather than...
24187 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
24189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24190 Alan Hayward <alan.hayward@arm.com>
24191 David Sherwood <david.sherwood@arm.com>
24193 * expmed.c (extract_bit_field_1): For vector extracts,
24194 fall back to extract_bit_field_as_subreg if vec_extract
24197 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24198 Alan Hayward <alan.hayward@arm.com>
24199 David Sherwood <david.sherwood@arm.com>
24201 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
24202 they are variable or constant sized.
24203 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
24204 slots for constant-sized data.
24206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24207 Alan Hayward <alan.hayward@arm.com>
24208 David Sherwood <david.sherwood@arm.com>
24210 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
24211 handling COND_EXPRs with boolean comparisons, try to find a better
24212 basis for the mask type than the boolean itself.
24214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24216 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
24217 is calculated and how it can be overridden.
24218 * genmodes.c (max_bitsize_mode_any_mode): New variable.
24219 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
24221 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
24224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24225 Alan Hayward <alan.hayward@arm.com>
24226 David Sherwood <david.sherwood@arm.com>
24228 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
24229 Remove the mode argument.
24230 (aarch64_simd_valid_immediate): Remove the mode and inverse
24232 * config/aarch64/iterators.md (bitsize): New iterator.
24233 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
24234 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
24235 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
24236 aarch64_simd_valid_immediate.
24237 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
24238 (aarch64_reg_or_bic_imm): Likewise.
24239 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
24240 with an insn_type enum and msl with a modifier_type enum.
24241 Replace element_width with a scalar_mode. Change the shift
24242 to unsigned int. Add constructors for scalar_float_mode and
24243 scalar_int_mode elements.
24244 (aarch64_vect_float_const_representable_p): Delete.
24245 (aarch64_can_const_movi_rtx_p)
24246 (aarch64_simd_scalar_immediate_valid_for_move)
24247 (aarch64_simd_make_constant): Update call to
24248 aarch64_simd_valid_immediate.
24249 (aarch64_advsimd_valid_immediate_hs): New function.
24250 (aarch64_advsimd_valid_immediate): Likewise.
24251 (aarch64_simd_valid_immediate): Remove mode and inverse
24252 arguments. Rewrite to use the above. Use const_vec_duplicate_p
24253 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
24254 and aarch64_float_const_representable_p on the result.
24255 (aarch64_output_simd_mov_immediate): Remove mode argument.
24256 Update call to aarch64_simd_valid_immediate and use of
24257 simd_immediate_info.
24258 (aarch64_output_scalar_simd_mov_immediate): Update call
24261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24262 Alan Hayward <alan.hayward@arm.com>
24263 David Sherwood <david.sherwood@arm.com>
24265 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
24266 (mode_nunits): Likewise CONST_MODE_NUNITS.
24267 * machmode.def (ADJUST_NUNITS): Document.
24268 * genmodes.c (mode_data::need_nunits_adj): New field.
24269 (blank_mode): Update accordingly.
24270 (adj_nunits): New variable.
24271 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
24273 (emit_mode_size_inline): Set need_bytesize_adj for all modes
24274 listed in adj_nunits.
24275 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
24276 listed in adj_nunits. Don't emit case statements for such modes.
24277 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
24278 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
24279 nothing if adj_nunits is nonnull.
24280 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
24281 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
24282 (emit_mode_fbit): Update use of print_maybe_const_decl.
24283 (emit_move_size): Likewise. Treat the array as non-const
24285 (emit_mode_adjustments): Handle adj_nunits.
24287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24289 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
24290 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
24291 (VECTOR_MODES): Use it.
24292 (make_vector_modes): Take the prefix as an argument.
24294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24295 Alan Hayward <alan.hayward@arm.com>
24296 David Sherwood <david.sherwood@arm.com>
24298 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
24299 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
24300 for MODE_VECTOR_BOOL.
24301 * machmode.def (VECTOR_BOOL_MODE): Document.
24302 * genmodes.c (VECTOR_BOOL_MODE): New macro.
24303 (make_vector_bool_mode): New function.
24304 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
24306 * lto-streamer-in.c (lto_input_mode_table): Likewise.
24307 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
24309 * stor-layout.c (int_mode_for_mode): Likewise.
24310 * tree.c (build_vector_type_for_mode): Likewise.
24311 * varasm.c (output_constant_pool_2): Likewise.
24312 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
24313 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
24314 for MODE_VECTOR_BOOL.
24315 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
24316 of mode class checks.
24317 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
24318 instead of a list of mode class checks.
24319 (expand_vector_scalar_condition): Likewise.
24320 (type_for_widest_vector_mode): Handle BImode as an inner mode.
24322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24323 Alan Hayward <alan.hayward@arm.com>
24324 David Sherwood <david.sherwood@arm.com>
24326 * machmode.h (mode_size): Change from unsigned short to
24328 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
24329 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24330 or if measurement_type is not polynomial.
24331 (fixed_size_mode::includes_p): Check for constant-sized modes.
24332 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
24333 return a poly_uint16 rather than an unsigned short.
24334 (emit_mode_size): Change the type of mode_size from unsigned short
24335 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
24336 (emit_mode_adjustments): Cope with polynomial vector sizes.
24337 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24339 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24341 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
24342 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
24343 * caller-save.c (setup_save_areas): Likewise.
24344 (replace_reg_with_saved_mem): Likewise.
24345 * calls.c (emit_library_call_value_1): Likewise.
24346 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
24347 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
24348 (gen_lowpart_for_combine): Likewise.
24349 * convert.c (convert_to_integer_1): Likewise.
24350 * cse.c (equiv_constant, cse_insn): Likewise.
24351 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
24352 (cselib_subst_to_values): Likewise.
24353 * dce.c (word_dce_process_block): Likewise.
24354 * df-problems.c (df_word_lr_mark_ref): Likewise.
24355 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
24356 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
24357 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
24358 (rtl_for_decl_location): Likewise.
24359 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
24360 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
24361 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
24362 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
24363 (expand_expr_real_1): Likewise.
24364 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
24365 (pad_below): Likewise.
24366 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24367 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
24368 * ira.c (get_subreg_tracking_sizes): Likewise.
24369 * ira-build.c (ira_create_allocno_objects): Likewise.
24370 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
24371 (ira_sort_regnos_for_alter_reg): Likewise.
24372 * ira-costs.c (record_operand_costs): Likewise.
24373 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
24374 (resolve_simple_move): Likewise.
24375 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
24376 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
24377 (lra_constraints): Likewise.
24378 (CONST_POOL_OK_P): Reject variable-sized modes.
24379 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
24380 (add_pseudo_to_slot, lra_spill): Likewise.
24381 * omp-low.c (omp_clause_aligned_alignment): Likewise.
24382 * optabs-query.c (get_best_extraction_insn): Likewise.
24383 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24384 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
24385 (expand_mult_highpart, valid_multiword_target_p): Likewise.
24386 * recog.c (offsettable_address_addr_space_p): Likewise.
24387 * regcprop.c (maybe_mode_change): Likewise.
24388 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
24389 * regrename.c (build_def_use): Likewise.
24390 * regstat.c (dump_reg_info): Likewise.
24391 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
24392 (find_reloads, find_reloads_subreg_address): Likewise.
24393 * reload1.c (eliminate_regs_1): Likewise.
24394 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
24395 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
24396 (simplify_binary_operation_1, simplify_subreg): Likewise.
24397 * targhooks.c (default_function_arg_padding): Likewise.
24398 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
24399 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
24400 (verify_gimple_assign_ternary): Likewise.
24401 * tree-inline.c (estimate_move_cost): Likewise.
24402 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24403 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
24404 (get_address_cost_ainc): Likewise.
24405 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
24406 (vect_supportable_dr_alignment): Likewise.
24407 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24408 (vectorizable_reduction): Likewise.
24409 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
24410 (vectorizable_operation, vectorizable_load): Likewise.
24411 * tree.c (build_same_sized_truth_vector_type): Likewise.
24412 * valtrack.c (cleanup_auto_inc_dec): Likewise.
24413 * var-tracking.c (emit_note_insn_var_location): Likewise.
24414 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
24415 (ADDR_VEC_ALIGN): Likewise.
24417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24418 Alan Hayward <alan.hayward@arm.com>
24419 David Sherwood <david.sherwood@arm.com>
24421 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
24423 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24424 or if measurement_type is polynomial.
24425 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
24426 * combine.c (make_extraction): Likewise.
24427 * dse.c (find_shift_sequence): Likewise.
24428 * dwarf2out.c (mem_loc_descriptor): Likewise.
24429 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
24430 (extract_bit_field, extract_low_bits): Likewise.
24431 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
24432 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
24433 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
24434 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
24435 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24436 * reload.c (find_reloads): Likewise.
24437 * reload1.c (alter_reg): Likewise.
24438 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
24439 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
24440 * tree-if-conv.c (predicate_mem_writes): Likewise.
24441 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
24442 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
24443 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
24444 * valtrack.c (dead_debug_insert_temp): Likewise.
24445 * varasm.c (mergeable_constant_section): Likewise.
24446 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
24448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24449 Alan Hayward <alan.hayward@arm.com>
24450 David Sherwood <david.sherwood@arm.com>
24452 * expr.c (expand_assignment): Cope with polynomial mode sizes
24453 when assigning to a CONCAT.
24455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24456 Alan Hayward <alan.hayward@arm.com>
24457 David Sherwood <david.sherwood@arm.com>
24459 * machmode.h (mode_precision): Change from unsigned short to
24461 (mode_to_precision): Return a poly_uint16 rather than an unsigned
24463 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
24464 or if measurement_type is not polynomial.
24465 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
24466 in which the mode is already known to be a scalar_int_mode.
24467 * genmodes.c (emit_mode_precision): Change the type of mode_precision
24468 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
24470 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24471 for GET_MODE_PRECISION.
24472 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24473 for GET_MODE_PRECISION.
24474 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
24476 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
24477 (expand_field_assignment, make_extraction): Likewise.
24478 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
24479 (get_last_value): Likewise.
24480 * convert.c (convert_to_integer_1): Likewise.
24481 * cse.c (cse_insn): Likewise.
24482 * expr.c (expand_expr_real_1): Likewise.
24483 * lra-constraints.c (simplify_operand_subreg): Likewise.
24484 * optabs-query.c (can_atomic_load_p): Likewise.
24485 * optabs.c (expand_atomic_load): Likewise.
24486 (expand_atomic_store): Likewise.
24487 * ree.c (combine_reaching_defs): Likewise.
24488 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
24489 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
24490 * tree.h (type_has_mode_precision_p): Likewise.
24491 * ubsan.c (instrument_si_overflow): Likewise.
24493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24494 Alan Hayward <alan.hayward@arm.com>
24495 David Sherwood <david.sherwood@arm.com>
24497 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
24498 polynomial numbers of units.
24499 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
24500 (valid_vector_subparts_p): New function.
24501 (build_vector_type): Remove temporary shim and take the number
24502 of units as a poly_uint64 rather than an int.
24503 (build_opaque_vector_type): Take the number of units as a
24504 poly_uint64 rather than an int.
24505 * tree.c (build_vector_from_ctor): Handle polynomial
24506 TYPE_VECTOR_SUBPARTS.
24507 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
24508 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
24509 (build_vector_from_val): If the number of units is variable,
24510 use build_vec_duplicate_cst for constant operands and
24511 VEC_DUPLICATE_EXPR otherwise.
24512 (make_vector_type): Remove temporary is_constant ().
24513 (build_vector_type, build_opaque_vector_type): Take the number of
24514 units as a poly_uint64 rather than an int.
24515 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
24517 * cfgexpand.c (expand_debug_expr): Likewise.
24518 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
24519 (store_constructor, expand_expr_real_1): Likewise.
24520 (const_scalar_mask_from_tree): Likewise.
24521 * fold-const-call.c (fold_const_reduction): Likewise.
24522 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
24523 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
24524 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
24525 (fold_relational_const): Likewise.
24526 (native_interpret_vector): Likewise. Change the size from an
24527 int to an unsigned int.
24528 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
24529 TYPE_VECTOR_SUBPARTS.
24530 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
24531 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
24532 duplicating a non-constant operand into a variable-length vector.
24533 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
24534 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
24535 * ipa-icf.c (sem_variable::equals): Likewise.
24536 * match.pd: Likewise.
24537 * omp-simd-clone.c (simd_clone_subparts): Likewise.
24538 * print-tree.c (print_node): Likewise.
24539 * stor-layout.c (layout_type): Likewise.
24540 * targhooks.c (default_builtin_vectorization_cost): Likewise.
24541 * tree-cfg.c (verify_gimple_comparison): Likewise.
24542 (verify_gimple_assign_binary): Likewise.
24543 (verify_gimple_assign_ternary): Likewise.
24544 (verify_gimple_assign_single): Likewise.
24545 * tree-pretty-print.c (dump_generic_node): Likewise.
24546 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24547 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
24548 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
24549 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
24550 (vect_shift_permute_load_chain): Likewise.
24551 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
24552 (expand_vector_condition, optimize_vector_constructor): Likewise.
24553 (lower_vec_perm, get_compute_type): Likewise.
24554 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24555 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
24556 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
24557 (vect_recog_mask_conversion_pattern): Likewise.
24558 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
24559 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
24560 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24561 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
24562 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
24563 (vectorizable_shift, vectorizable_operation, vectorizable_store)
24564 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
24565 (supportable_widening_operation): Likewise.
24566 (supportable_narrowing_operation): Likewise.
24567 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
24569 * varasm.c (output_constant): Likewise.
24571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24572 Alan Hayward <alan.hayward@arm.com>
24573 David Sherwood <david.sherwood@arm.com>
24575 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
24576 so that both the length == 3 and length != 3 cases set up their
24577 own permute vectors. Add comments explaining why we know the
24578 number of elements is constant.
24579 (vect_permute_load_chain): Likewise.
24581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24582 Alan Hayward <alan.hayward@arm.com>
24583 David Sherwood <david.sherwood@arm.com>
24585 * machmode.h (mode_nunits): Change from unsigned char to
24587 (ONLY_FIXED_SIZE_MODES): New macro.
24588 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
24589 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
24590 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
24592 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
24593 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
24594 or if measurement_type is not polynomial.
24595 * genmodes.c (ZERO_COEFFS): New macro.
24596 (emit_mode_nunits_inline): Make mode_nunits_inline return a
24598 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
24599 Use ZERO_COEFFS when emitting initializers.
24600 * data-streamer.h (bp_pack_poly_value): New function.
24601 (bp_unpack_poly_value): Likewise.
24602 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24603 for GET_MODE_NUNITS.
24604 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24605 for GET_MODE_NUNITS.
24606 * tree.c (make_vector_type): Remove temporary shim and make
24607 the real function take the number of units as a poly_uint64
24608 rather than an int.
24609 (build_vector_type_for_mode): Handle polynomial nunits.
24610 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
24611 * emit-rtl.c (const_vec_series_p_1): Likewise.
24612 (gen_rtx_CONST_VECTOR): Likewise.
24613 * fold-const.c (test_vec_duplicate_folding): Likewise.
24614 * genrecog.c (validate_pattern): Likewise.
24615 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
24616 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24617 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
24618 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
24619 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
24620 * rtlanal.c (subreg_get_info): Likewise.
24621 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24622 (vect_grouped_load_supported): Likewise.
24623 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
24624 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
24625 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
24626 (simplify_const_unary_operation, simplify_binary_operation_1)
24627 (simplify_const_binary_operation, simplify_ternary_operation)
24628 (test_vector_ops_duplicate, test_vector_ops): Likewise.
24629 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
24630 instead of CONST_VECTOR_NUNITS.
24631 * varasm.c (output_constant_pool_2): Likewise.
24632 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
24633 explicit-encoded elements in the XVEC for variable-length vectors.
24635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24637 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
24639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24640 Alan Hayward <alan.hayward@arm.com>
24641 David Sherwood <david.sherwood@arm.com>
24643 * coretypes.h (fixed_size_mode): Declare.
24644 (fixed_size_mode_pod): New typedef.
24645 * builtins.h (target_builtins::x_apply_args_mode)
24646 (target_builtins::x_apply_result_mode): Change type to
24647 fixed_size_mode_pod.
24648 * builtins.c (apply_args_size, apply_result_size, result_vector)
24649 (expand_builtin_apply_args_1, expand_builtin_apply)
24650 (expand_builtin_return): Update accordingly.
24652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24654 * cse.c (hash_rtx_cb): Hash only the encoded elements.
24655 * cselib.c (cselib_hash_rtx): Likewise.
24656 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
24657 CONST_VECTOR encoding.
24659 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24660 Jeff Law <law@redhat.com>
24663 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
24664 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
24665 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
24666 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
24669 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
24670 explicitly probe *sp in a noreturn function if there were any callee
24671 register saves or frame pointer is needed.
24673 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24676 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
24677 BLKmode for ternary, binary or unary expressions.
24680 * var-tracking.c (delete_vta_debug_insn): New inline function.
24681 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
24682 insns from get_insns () to NULL instead of each bb separately.
24683 Use delete_vta_debug_insn. No longer static.
24684 (vt_debug_insns_local, variable_tracking_main_1): Adjust
24685 delete_vta_debug_insns callers.
24686 * rtl.h (delete_vta_debug_insns): Declare.
24687 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
24688 instead of variable_tracking_main.
24690 2018-01-03 Martin Sebor <msebor@redhat.com>
24692 PR tree-optimization/83603
24693 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
24694 arguments past the endof the argument list in functions declared
24695 without a prototype.
24696 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
24697 Avoid checking when arguments are null.
24699 2018-01-03 Martin Sebor <msebor@redhat.com>
24702 * doc/extend.texi (attribute const): Fix a typo.
24703 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
24704 issuing -Wsuggest-attribute for void functions.
24706 2018-01-03 Martin Sebor <msebor@redhat.com>
24708 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
24709 offset_int::from instead of wide_int::to_shwi.
24710 (maybe_diag_overlap): Remove assertion.
24711 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
24712 * gimple-ssa-sprintf.c (format_directive): Same.
24713 (parse_directive): Same.
24714 (sprintf_dom_walker::compute_format_length): Same.
24715 (try_substitute_return_value): Same.
24717 2018-01-03 Jeff Law <law@redhat.com>
24719 PR middle-end/83654
24720 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
24721 non-constant residual for zero at runtime and avoid probing in
24722 that case. Reorganize code for trailing problem to mirror handling
24725 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24727 PR tree-optimization/83501
24728 * tree-ssa-strlen.c (get_string_cst): New.
24729 (handle_char_store): Call get_string_cst.
24731 2018-01-03 Martin Liska <mliska@suse.cz>
24733 PR tree-optimization/83593
24734 * tree-ssa-strlen.c: Include tree-cfg.h.
24735 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
24736 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
24737 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
24739 (strlen_dom_walker::before_dom_children): Call
24740 gimple_purge_dead_eh_edges. Dump tranformation with details
24742 (strlen_dom_walker::before_dom_children): Update call by adding
24743 new argument cleanup_eh.
24744 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
24746 2018-01-03 Martin Liska <mliska@suse.cz>
24749 * cif-code.def (VARIADIC_THUNK): New enum value.
24750 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
24753 2018-01-03 Jan Beulich <jbeulich@suse.com>
24755 * sse.md (mov<mode>_internal): Tighten condition for when to use
24756 vmovdqu<ssescalarsize> for TI and OI modes.
24758 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24760 Update copyright years.
24762 2018-01-03 Martin Liska <mliska@suse.cz>
24765 * ipa-visibility.c (function_and_variable_visibility): Skip
24766 functions with noipa attribure.
24768 2018-01-03 Jakub Jelinek <jakub@redhat.com>
24770 * gcc.c (process_command): Update copyright notice dates.
24771 * gcov-dump.c (print_version): Ditto.
24772 * gcov.c (print_version): Ditto.
24773 * gcov-tool.c (print_version): Ditto.
24774 * gengtype.c (create_file): Ditto.
24775 * doc/cpp.texi: Bump @copying's copyright year.
24776 * doc/cppinternals.texi: Ditto.
24777 * doc/gcc.texi: Ditto.
24778 * doc/gccint.texi: Ditto.
24779 * doc/gcov.texi: Ditto.
24780 * doc/install.texi: Ditto.
24781 * doc/invoke.texi: Ditto.
24783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24785 * vector-builder.h (vector_builder::m_full_nelts): Change from
24786 unsigned int to poly_uint64.
24787 (vector_builder::full_nelts): Update prototype accordingly.
24788 (vector_builder::new_vector): Likewise.
24789 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
24790 (vector_builder::operator ==): Likewise.
24791 (vector_builder::finalize): Likewise.
24792 * int-vector-builder.h (int_vector_builder::int_vector_builder):
24793 Take the number of elements as a poly_uint64 rather than an
24795 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
24796 from unsigned int to poly_uint64.
24797 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
24798 (vec_perm_indices::new_vector): Likewise.
24799 (vec_perm_indices::length): Likewise.
24800 (vec_perm_indices::nelts_per_input): Likewise.
24801 (vec_perm_indices::input_nelts): Likewise.
24802 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
24803 number of elements per input as a poly_uint64 rather than an
24804 unsigned int. Use the original encoding for variable-length
24805 vectors, rather than clamping each individual element.
24806 For the second and subsequent elements in each pattern,
24807 clamp the step and base before clamping their sum.
24808 (vec_perm_indices::series_p): Handle polynomial element counts.
24809 (vec_perm_indices::all_in_range_p): Likewise.
24810 (vec_perm_indices_to_tree): Likewise.
24811 (vec_perm_indices_to_rtx): Likewise.
24812 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
24813 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
24814 (tree_vector_builder::new_binary_operation): Handle polynomial
24815 element counts. Return false if we need to know the number
24816 of elements at compile time.
24817 * fold-const.c (fold_vec_perm): Punt if the number of elements
24818 isn't known at compile time.
24820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24822 * vec-perm-indices.h (vec_perm_builder): Change element type
24823 from HOST_WIDE_INT to poly_int64.
24824 (vec_perm_indices::element_type): Update accordingly.
24825 (vec_perm_indices::clamp): Handle polynomial element_types.
24826 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24827 (vec_perm_indices::all_in_range_p): Likewise.
24828 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
24830 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
24831 polynomial vec_perm_indices element types.
24832 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
24833 * fold-const.c (fold_vec_perm): Likewise.
24834 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
24835 * tree-vect-generic.c (lower_vec_perm): Likewise.
24836 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24837 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
24838 element type to HOST_WIDE_INT.
24840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24841 Alan Hayward <alan.hayward@arm.com>
24842 David Sherwood <david.sherwood@arm.com>
24844 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
24845 rather than an int. Use plus_constant.
24846 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
24847 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
24849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24850 Alan Hayward <alan.hayward@arm.com>
24851 David Sherwood <david.sherwood@arm.com>
24853 * calls.c (emit_call_1, expand_call): Change struct_value_size from
24854 a HOST_WIDE_INT to a poly_int64.
24856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24857 Alan Hayward <alan.hayward@arm.com>
24858 David Sherwood <david.sherwood@arm.com>
24860 * calls.c (load_register_parameters): Cope with polynomial
24861 mode sizes. Require a constant size for BLKmode parameters
24862 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
24863 forces a parameter to be padded at the lsb end in order to
24864 fill a complete number of words, require the parameter size
24865 to be ordered wrt UNITS_PER_WORD.
24867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24868 Alan Hayward <alan.hayward@arm.com>
24869 David Sherwood <david.sherwood@arm.com>
24871 * reload1.c (spill_stack_slot_width): Change element type
24872 from unsigned int to poly_uint64_pod.
24873 (alter_reg): Treat mode sizes as polynomial.
24875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24876 Alan Hayward <alan.hayward@arm.com>
24877 David Sherwood <david.sherwood@arm.com>
24879 * reload.c (complex_word_subreg_p): New function.
24880 (reload_inner_reg_of_subreg, push_reload): Use it.
24882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24883 Alan Hayward <alan.hayward@arm.com>
24884 David Sherwood <david.sherwood@arm.com>
24886 * lra-constraints.c (process_alt_operands): Reject matched
24887 operands whose sizes aren't ordered.
24888 (match_reload): Refer to this check here.
24890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24891 Alan Hayward <alan.hayward@arm.com>
24892 David Sherwood <david.sherwood@arm.com>
24894 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
24895 that the mode size is in the set {1, 2, 4, 8, 16}.
24897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24898 Alan Hayward <alan.hayward@arm.com>
24899 David Sherwood <david.sherwood@arm.com>
24901 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
24902 Use plus_constant instead of gen_rtx_PLUS.
24904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24905 Alan Hayward <alan.hayward@arm.com>
24906 David Sherwood <david.sherwood@arm.com>
24908 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
24909 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
24910 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
24911 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
24912 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
24913 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
24914 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
24915 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
24916 * config/i386/i386.c (ix86_push_rounding): ...this new function.
24917 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
24919 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
24920 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
24921 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
24922 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
24923 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
24924 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
24925 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
24926 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
24927 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
24928 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
24930 * expr.c (emit_move_resolve_push): Treat the input and result
24931 of PUSH_ROUNDING as a poly_int64.
24932 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
24933 (emit_push_insn): Likewise.
24934 * lra-eliminations.c (mark_not_eliminable): Likewise.
24935 * recog.c (push_operand): Likewise.
24936 * reload1.c (elimination_effects): Likewise.
24937 * rtlanal.c (nonzero_bits1): Likewise.
24938 * calls.c (store_one_arg): Likewise. Require the padding to be
24939 known at compile time.
24941 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24942 Alan Hayward <alan.hayward@arm.com>
24943 David Sherwood <david.sherwood@arm.com>
24945 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
24946 Use plus_constant instead of gen_rtx_PLUS.
24948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24949 Alan Hayward <alan.hayward@arm.com>
24950 David Sherwood <david.sherwood@arm.com>
24952 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
24953 rather than an int.
24955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24956 Alan Hayward <alan.hayward@arm.com>
24957 David Sherwood <david.sherwood@arm.com>
24959 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
24960 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
24961 via stack temporaries. Treat the mode size as polynomial too.
24963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24964 Alan Hayward <alan.hayward@arm.com>
24965 David Sherwood <david.sherwood@arm.com>
24967 * expr.c (expand_expr_real_2): When handling conversions involving
24968 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
24969 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
24970 as a poly_uint64 too.
24972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24973 Alan Hayward <alan.hayward@arm.com>
24974 David Sherwood <david.sherwood@arm.com>
24976 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
24978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24979 Alan Hayward <alan.hayward@arm.com>
24980 David Sherwood <david.sherwood@arm.com>
24982 * combine.c (can_change_dest_mode): Handle polynomial
24983 REGMODE_NATURAL_SIZE.
24984 * expmed.c (store_bit_field_1): Likewise.
24985 * expr.c (store_constructor): Likewise.
24986 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
24987 and polynomial REGMODE_NATURAL_SIZE.
24988 (gen_lowpart_common): Likewise.
24989 * reginfo.c (record_subregs_of_mode): Likewise.
24990 * rtlanal.c (read_modify_subreg_p): Likewise.
24992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
24993 Alan Hayward <alan.hayward@arm.com>
24994 David Sherwood <david.sherwood@arm.com>
24996 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
24997 numbers of elements.
24999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25000 Alan Hayward <alan.hayward@arm.com>
25001 David Sherwood <david.sherwood@arm.com>
25003 * match.pd: Cope with polynomial numbers of vector elements.
25005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25006 Alan Hayward <alan.hayward@arm.com>
25007 David Sherwood <david.sherwood@arm.com>
25009 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
25010 in a POINTER_PLUS_EXPR.
25012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25013 Alan Hayward <alan.hayward@arm.com>
25014 David Sherwood <david.sherwood@arm.com>
25016 * omp-simd-clone.c (simd_clone_subparts): New function.
25017 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
25018 (ipa_simd_modify_function_body): Likewise.
25020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25021 Alan Hayward <alan.hayward@arm.com>
25022 David Sherwood <david.sherwood@arm.com>
25024 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
25025 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
25026 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
25027 (expand_vector_condition, vector_element): Likewise.
25028 (subparts_gt): New function.
25029 (get_compute_type): Use subparts_gt.
25030 (count_type_subparts): Delete.
25031 (expand_vector_operations_1): Use subparts_gt instead of
25032 count_type_subparts.
25034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25035 Alan Hayward <alan.hayward@arm.com>
25036 David Sherwood <david.sherwood@arm.com>
25038 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
25039 (vect_compile_time_alias): ...this new function. Do the calculation
25040 on poly_ints rather than trees.
25041 (vect_prune_runtime_alias_test_list): Update call accordingly.
25043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25044 Alan Hayward <alan.hayward@arm.com>
25045 David Sherwood <david.sherwood@arm.com>
25047 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
25049 (vect_schedule_slp_instance): Likewise.
25051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25052 Alan Hayward <alan.hayward@arm.com>
25053 David Sherwood <david.sherwood@arm.com>
25055 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
25056 constant and extern definitions for variable-length vectors.
25057 (vect_get_constant_vectors): Note that the number of units
25058 is known to be constant.
25060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25061 Alan Hayward <alan.hayward@arm.com>
25062 David Sherwood <david.sherwood@arm.com>
25064 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
25065 of units as polynomial. Choose between WIDE and NARROW based
25068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25069 Alan Hayward <alan.hayward@arm.com>
25070 David Sherwood <david.sherwood@arm.com>
25072 * tree-vect-stmts.c (simd_clone_subparts): New function.
25073 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
25075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25076 Alan Hayward <alan.hayward@arm.com>
25077 David Sherwood <david.sherwood@arm.com>
25079 * tree-vect-stmts.c (vectorizable_call): Treat the number of
25080 vectors as polynomial. Use build_index_vector for
25081 IFN_GOMP_SIMD_LANE.
25083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25084 Alan Hayward <alan.hayward@arm.com>
25085 David Sherwood <david.sherwood@arm.com>
25087 * tree-vect-stmts.c (get_load_store_type): Treat the number of
25088 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
25089 for variable-length vectors.
25090 (vectorizable_mask_load_store): Treat the number of units as
25091 polynomial, asserting that it is constant if the condition has
25092 already been enforced.
25093 (vectorizable_store, vectorizable_load): Likewise.
25095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25096 Alan Hayward <alan.hayward@arm.com>
25097 David Sherwood <david.sherwood@arm.com>
25099 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
25100 of units as polynomial. Punt if we can't tell at compile time
25101 which vector contains the final result.
25103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25104 Alan Hayward <alan.hayward@arm.com>
25105 David Sherwood <david.sherwood@arm.com>
25107 * tree-vect-loop.c (vectorizable_induction): Treat the number
25108 of units as polynomial. Punt on SLP inductions. Use an integer
25109 VEC_SERIES_EXPR for variable-length integer reductions. Use a
25110 cast of such a series for variable-length floating-point
25113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25114 Alan Hayward <alan.hayward@arm.com>
25115 David Sherwood <david.sherwood@arm.com>
25117 * tree.h (build_index_vector): Declare.
25118 * tree.c (build_index_vector): New function.
25119 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
25120 of units as polynomial, forcibly converting it to a constant if
25121 vectorizable_reduction has already enforced the condition.
25122 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
25123 to create a {1,2,3,...} vector.
25124 (vectorizable_reduction): Treat the number of units as polynomial.
25125 Choose vectype_in based on the largest scalar element size rather
25126 than the smallest number of units. Enforce the restrictions
25129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25130 Alan Hayward <alan.hayward@arm.com>
25131 David Sherwood <david.sherwood@arm.com>
25133 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
25134 number of units as polynomial.
25136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25137 Alan Hayward <alan.hayward@arm.com>
25138 David Sherwood <david.sherwood@arm.com>
25140 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
25141 * target.def (autovectorize_vector_sizes): Return the vector sizes
25142 by pointer, using vector_sizes rather than a bitmask.
25143 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
25144 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
25145 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
25147 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
25148 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
25149 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
25150 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
25151 * omp-general.c (omp_max_vf): Likewise.
25152 * omp-low.c (omp_clause_aligned_alignment): Likewise.
25153 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
25154 * tree-vect-loop.c (vect_analyze_loop): Likewise.
25155 * tree-vect-slp.c (vect_slp_bb): Likewise.
25156 * doc/tm.texi: Regenerate.
25157 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
25159 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
25160 the vector size as a poly_uint64 rather than an unsigned int.
25161 (current_vector_size): Change from an unsigned int to a poly_uint64.
25162 (get_vectype_for_scalar_type): Update accordingly.
25163 * tree.h (build_truth_vector_type): Take the size and number of
25164 units as a poly_uint64 rather than an unsigned int.
25165 (build_vector_type): Add a temporary overload that takes
25166 the number of units as a poly_uint64 rather than an unsigned int.
25167 * tree.c (make_vector_type): Likewise.
25168 (build_truth_vector_type): Take the number of units as a poly_uint64
25169 rather than an unsigned int.
25171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25172 Alan Hayward <alan.hayward@arm.com>
25173 David Sherwood <david.sherwood@arm.com>
25175 * target.def (get_mask_mode): Take the number of units and length
25176 as poly_uint64s rather than unsigned ints.
25177 * targhooks.h (default_get_mask_mode): Update accordingly.
25178 * targhooks.c (default_get_mask_mode): Likewise.
25179 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
25180 * doc/tm.texi: Regenerate.
25182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25183 Alan Hayward <alan.hayward@arm.com>
25184 David Sherwood <david.sherwood@arm.com>
25186 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
25187 * omp-general.c (omp_max_vf): Likewise.
25188 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
25189 (expand_omp_simd): Handle polynomial safelen.
25190 * omp-low.c (omplow_simd_context): Add a default constructor.
25191 (omplow_simd_context::max_vf): Change from int to poly_uint64.
25192 (lower_rec_simd_input_clauses): Update accordingly.
25193 (lower_rec_input_clauses): Likewise.
25195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25196 Alan Hayward <alan.hayward@arm.com>
25197 David Sherwood <david.sherwood@arm.com>
25199 * tree-vectorizer.h (vect_nunits_for_cost): New function.
25200 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
25201 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
25202 (vect_analyze_slp_cost): Likewise.
25203 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
25204 (vect_model_load_cost): Likewise.
25206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25207 Alan Hayward <alan.hayward@arm.com>
25208 David Sherwood <david.sherwood@arm.com>
25210 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
25211 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
25212 from an unsigned int * to a poly_uint64_pod *.
25213 (calculate_unrolling_factor): New function.
25214 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
25216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25217 Alan Hayward <alan.hayward@arm.com>
25218 David Sherwood <david.sherwood@arm.com>
25220 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
25221 from an unsigned int to a poly_uint64.
25222 (_loop_vec_info::slp_unrolling_factor): Likewise.
25223 (_loop_vec_info::vectorization_factor): Change from an int
25225 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
25226 (vect_get_num_vectors): New function.
25227 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
25228 (vect_get_num_copies): Use vect_get_num_vectors.
25229 (vect_analyze_data_ref_dependences): Change max_vf from an int *
25230 to an unsigned int *.
25231 (vect_analyze_data_refs): Change min_vf from an int * to a
25233 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25234 than an unsigned HOST_WIDE_INT.
25235 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
25236 (vect_analyze_data_ref_dependence): Change max_vf from an int *
25237 to an unsigned int *.
25238 (vect_analyze_data_ref_dependences): Likewise.
25239 (vect_compute_data_ref_alignment): Handle polynomial vf.
25240 (vect_enhance_data_refs_alignment): Likewise.
25241 (vect_prune_runtime_alias_test_list): Likewise.
25242 (vect_shift_permute_load_chain): Likewise.
25243 (vect_supportable_dr_alignment): Likewise.
25244 (dependence_distance_ge_vf): Take the vectorization factor as a
25245 poly_uint64 rather than an unsigned HOST_WIDE_INT.
25246 (vect_analyze_data_refs): Change min_vf from an int * to a
25248 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
25249 vfm1 as a poly_uint64 rather than an int. Make the same change
25250 for the returned bound_scalar.
25251 (vect_gen_vector_loop_niters): Handle polynomial vf.
25252 (vect_do_peeling): Likewise. Update call to
25253 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
25254 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
25256 * tree-vect-loop.c (vect_determine_vectorization_factor)
25257 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
25258 (vect_get_known_peeling_cost): Likewise.
25259 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
25260 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
25261 (vect_transform_loop): Likewise. Use the lowest possible VF when
25262 updating the upper bounds of the loop.
25263 (vect_min_worthwhile_factor): Make static. Return an unsigned int
25264 rather than an int.
25265 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
25266 polynomial unroll factors.
25267 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
25268 (vect_make_slp_decision): Likewise.
25269 (vect_supported_load_permutation_p): Likewise, and polynomial
25271 (vect_analyze_slp_cost): Handle polynomial vf.
25272 (vect_slp_analyze_node_operations): Likewise.
25273 (vect_slp_analyze_bb_1): Likewise.
25274 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25275 than an unsigned HOST_WIDE_INT.
25276 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
25277 (vectorizable_load): Handle polynomial vf.
25278 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
25280 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
25282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25283 Alan Hayward <alan.hayward@arm.com>
25284 David Sherwood <david.sherwood@arm.com>
25286 * match.pd: Handle bit operations involving three constants
25287 and try to fold one pair.
25289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25291 * tree-vect-loop-manip.c: Include gimple-fold.h.
25292 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
25293 niters_maybe_zero parameters. Handle other cases besides a step of 1.
25294 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
25295 Add a path that uses a step of VF instead of 1, but disable it
25297 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
25298 and niters_no_overflow parameters. Update calls to
25299 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
25300 Create a new SSA name if the latter choses to use a ste other
25301 than zero, and return it via niters_vector_mult_vf_var.
25302 * tree-vect-loop.c (vect_transform_loop): Update calls to
25303 vect_do_peeling, vect_gen_vector_loop_niters and
25304 slpeel_make_loop_iterate_ntimes.
25305 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
25306 (vect_gen_vector_loop_niters): Update declarations after above changes.
25308 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
25310 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
25311 128-bit round to integer instructions.
25312 (ceil<mode>2): Likewise.
25313 (btrunc<mode>2): Likewise.
25314 (round<mode>2): Likewise.
25316 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25318 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
25319 unaligned VSX load/store on P8/P9.
25320 (expand_block_clear): Allow the use of unaligned VSX
25321 load/store on P8/P9.
25323 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25325 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
25327 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
25328 swap associated with both a load and a store.
25330 2018-01-02 Andrew Waterman <andrew@sifive.com>
25332 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
25333 * config/riscv/riscv.md (clear_cache): Use it.
25335 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
25337 * web.c: Remove out-of-date comment.
25339 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25341 * expr.c (fixup_args_size_notes): Check that any existing
25342 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
25343 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
25344 (emit_single_push_insn): ...here.
25346 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25348 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
25349 (const_vector_encoded_nelts): New function.
25350 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
25351 (const_vector_int_elt, const_vector_elt): Declare.
25352 * emit-rtl.c (const_vector_int_elt_1): New function.
25353 (const_vector_elt): Likewise.
25354 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
25355 of CONST_VECTOR_ELT.
25357 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25359 * expr.c: Include rtx-vector-builder.h.
25360 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
25361 directly on the tree encoding.
25362 (const_vector_from_tree): Likewise.
25363 * optabs.c: Include rtx-vector-builder.h.
25364 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
25365 sequence of "u" values.
25366 * vec-perm-indices.c: Include rtx-vector-builder.h.
25367 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
25368 directly on the vec_perm_indices encoding.
25370 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25372 * doc/rtl.texi (const_vector): Describe new encoding scheme.
25373 * Makefile.in (OBJS): Add rtx-vector-builder.o.
25374 * rtx-vector-builder.h: New file.
25375 * rtx-vector-builder.c: Likewise.
25376 * rtl.h (rtx_def::u2): Add a const_vector field.
25377 (CONST_VECTOR_NPATTERNS): New macro.
25378 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
25379 (CONST_VECTOR_DUPLICATE_P): Likewise.
25380 (CONST_VECTOR_STEPPED_P): Likewise.
25381 (CONST_VECTOR_ENCODED_ELT): Likewise.
25382 (const_vec_duplicate_p): Check for a duplicated vector encoding.
25383 (unwrap_const_vec_duplicate): Likewise.
25384 (const_vec_series_p): Check for a non-duplicated vector encoding.
25385 Say that the function only returns true for integer vectors.
25386 * emit-rtl.c: Include rtx-vector-builder.h.
25387 (gen_const_vec_duplicate_1): Delete.
25388 (gen_const_vector): Call gen_const_vec_duplicate instead of
25389 gen_const_vec_duplicate_1.
25390 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
25391 (gen_const_vec_duplicate): Use rtx_vector_builder.
25392 (gen_const_vec_series): Likewise.
25393 (gen_rtx_CONST_VECTOR): Likewise.
25394 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
25395 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25396 Build a new vector rather than modifying a CONST_VECTOR in-place.
25397 (handle_special_swappables): Update call accordingly.
25398 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
25399 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25400 Build a new vector rather than modifying a CONST_VECTOR in-place.
25401 (handle_special_swappables): Update call accordingly.
25403 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25405 * simplify-rtx.c (simplify_const_binary_operation): Use
25406 CONST_VECTOR_ELT instead of XVECEXP.
25408 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25410 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
25411 the selector elements to be different from the data elements
25412 if the selector is a VECTOR_CST.
25413 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
25414 ssizetype for the selector.
25416 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25418 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
25419 before testing each element individually.
25420 * tree-vect-generic.c (lower_vec_perm): Likewise.
25422 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25424 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
25425 * selftest-run-tests.c (selftest::run_tests): Call it.
25426 * vector-builder.h (vector_builder::operator ==): New function.
25427 (vector_builder::operator !=): Likewise.
25428 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
25429 (vec_perm_indices::all_from_input_p): New function.
25430 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25431 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
25432 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
25433 instead of reading the VECTOR_CST directly. Detect whether both
25434 vector inputs are the same before constructing the vec_perm_indices,
25435 and update the number of inputs argument accordingly. Use the
25436 utility functions added above. Only construct sel2 if we need to.
25438 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25440 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
25441 the broadcast of the low byte.
25442 (expand_mult_highpart): Use an explicit encoding for the permutes.
25443 * optabs-query.c (can_mult_highpart_p): Likewise.
25444 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25445 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25446 (vectorizable_bswap): Likewise.
25447 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
25448 explicit encoding for the power-of-2 permutes.
25449 (vect_permute_store_chain): Likewise.
25450 (vect_grouped_load_supported): Likewise.
25451 (vect_permute_load_chain): Likewise.
25453 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25455 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
25456 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
25457 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
25458 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25459 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
25460 (vect_gen_perm_mask_any): Likewise.
25462 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25464 * int-vector-builder.h: New file.
25465 * vec-perm-indices.h: Include int-vector-builder.h.
25466 (vec_perm_indices): Redefine as an int_vector_builder.
25467 (auto_vec_perm_indices): Delete.
25468 (vec_perm_builder): Redefine as a stand-alone class.
25469 (vec_perm_indices::vec_perm_indices): New function.
25470 (vec_perm_indices::clamp): Likewise.
25471 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
25472 (vec_perm_indices::new_vector): New function.
25473 (vec_perm_indices::new_expanded_vector): Update for new
25474 vec_perm_indices class.
25475 (vec_perm_indices::rotate_inputs): New function.
25476 (vec_perm_indices::all_in_range_p): Operate directly on the
25477 encoded form, without computing elided elements.
25478 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
25479 encoding. Update for new vec_perm_indices class.
25480 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
25481 the given vec_perm_builder.
25482 (expand_vec_perm_var): Update vec_perm_builder constructor.
25483 (expand_mult_highpart): Use vec_perm_builder instead of
25484 auto_vec_perm_indices.
25485 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
25486 vec_perm_indices instead of auto_vec_perm_indices. Use a single
25487 or double series encoding as appropriate.
25488 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
25489 vec_perm_indices instead of auto_vec_perm_indices.
25490 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25491 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25492 (vect_permute_store_chain): Likewise.
25493 (vect_grouped_load_supported): Likewise.
25494 (vect_permute_load_chain): Likewise.
25495 (vect_shift_permute_load_chain): Likewise.
25496 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25497 (vect_transform_slp_perm_load): Likewise.
25498 (vect_schedule_slp_instance): Likewise.
25499 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25500 (vectorizable_mask_load_store): Likewise.
25501 (vectorizable_bswap): Likewise.
25502 (vectorizable_store): Likewise.
25503 (vectorizable_load): Likewise.
25504 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
25505 vec_perm_indices instead of auto_vec_perm_indices. Use
25506 tree_to_vec_perm_builder to read the vector from a tree.
25507 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
25508 vec_perm_builder instead of a vec_perm_indices.
25509 (have_whole_vector_shift): Use vec_perm_builder and
25510 vec_perm_indices instead of auto_vec_perm_indices. Leave the
25511 truncation to calc_vec_perm_mask_for_shift.
25512 (vect_create_epilog_for_reduction): Likewise.
25513 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
25514 from auto_vec_perm_indices to vec_perm_indices.
25515 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25516 instead of changing individual elements.
25517 (aarch64_vectorize_vec_perm_const): Use new_vector to install
25518 the vector in d.perm.
25519 * config/arm/arm.c (expand_vec_perm_d::perm): Change
25520 from auto_vec_perm_indices to vec_perm_indices.
25521 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25522 instead of changing individual elements.
25523 (arm_vectorize_vec_perm_const): Use new_vector to install
25524 the vector in d.perm.
25525 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
25526 Update vec_perm_builder constructor.
25527 (rs6000_expand_interleave): Likewise.
25528 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
25529 (rs6000_expand_interleave): Likewise.
25531 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25533 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
25534 to qimode could truncate the indices.
25535 * optabs.c (expand_vec_perm_var): Likewise.
25537 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25539 * Makefile.in (OBJS): Add vec-perm-indices.o.
25540 * vec-perm-indices.h: New file.
25541 * vec-perm-indices.c: Likewise.
25542 * target.h (vec_perm_indices): Replace with a forward class
25544 (auto_vec_perm_indices): Move to vec-perm-indices.h.
25545 * optabs.h: Include vec-perm-indices.h.
25546 (expand_vec_perm): Delete.
25547 (selector_fits_mode_p, expand_vec_perm_var): Declare.
25548 (expand_vec_perm_const): Declare.
25549 * target.def (vec_perm_const_ok): Replace with...
25550 (vec_perm_const): ...this new hook.
25551 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
25552 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
25553 * doc/tm.texi: Regenerate.
25554 * optabs.def (vec_perm_const): Delete.
25555 * doc/md.texi (vec_perm_const): Likewise.
25556 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
25557 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
25558 expand_vec_perm for constant permutation vectors. Assert that
25559 the mode of variable permutation vectors is the integer equivalent
25560 of the mode that is being permuted.
25561 * optabs-query.h (selector_fits_mode_p): Declare.
25562 * optabs-query.c: Include vec-perm-indices.h.
25563 (selector_fits_mode_p): New function.
25564 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
25565 is defined, instead of checking whether the vec_perm_const_optab
25566 exists. Use targetm.vectorize.vec_perm_const instead of
25567 targetm.vectorize.vec_perm_const_ok. Check whether the indices
25568 fit in the vector mode before using a variable permute.
25569 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
25570 vec_perm_indices instead of an rtx.
25571 (expand_vec_perm): Replace with...
25572 (expand_vec_perm_const): ...this new function. Take the selector
25573 as a vec_perm_indices rather than an rtx. Also take the mode of
25574 the selector. Update call to shift_amt_for_vec_perm_mask.
25575 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
25576 Use vec_perm_indices::new_expanded_vector to expand the original
25577 selector into bytes. Check whether the indices fit in the vector
25578 mode before using a variable permute.
25579 (expand_vec_perm_var): Make global.
25580 (expand_mult_highpart): Use expand_vec_perm_const.
25581 * fold-const.c: Includes vec-perm-indices.h.
25582 * tree-ssa-forwprop.c: Likewise.
25583 * tree-vect-data-refs.c: Likewise.
25584 * tree-vect-generic.c: Likewise.
25585 * tree-vect-loop.c: Likewise.
25586 * tree-vect-slp.c: Likewise.
25587 * tree-vect-stmts.c: Likewise.
25588 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
25590 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
25591 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
25592 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
25593 (aarch64_vectorize_vec_perm_const): ...this new function.
25594 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25595 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25596 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
25597 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
25598 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25599 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25600 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
25602 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
25603 check for NEON modes.
25604 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
25605 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
25606 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
25607 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
25609 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
25610 the old VEC_PERM_CONST conditions.
25611 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
25612 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
25613 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
25614 (ia64_vectorize_vec_perm_const_ok): Merge into...
25615 (ia64_vectorize_vec_perm_const): ...this new function.
25616 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
25617 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
25618 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
25619 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
25620 * config/mips/mips.c (mips_expand_vec_perm_const)
25621 (mips_vectorize_vec_perm_const_ok): Merge into...
25622 (mips_vectorize_vec_perm_const): ...this new function.
25623 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
25624 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
25625 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
25626 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
25627 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
25628 (rs6000_expand_vec_perm_const): Delete.
25629 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
25631 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25632 (altivec_expand_vec_perm_const_le): Take each operand individually.
25633 Operate on constant selectors rather than rtxes.
25634 (altivec_expand_vec_perm_const): Likewise. Update call to
25635 altivec_expand_vec_perm_const_le.
25636 (rs6000_expand_vec_perm_const): Delete.
25637 (rs6000_vectorize_vec_perm_const_ok): Delete.
25638 (rs6000_vectorize_vec_perm_const): New function.
25639 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25640 an element count and rtx array.
25641 (rs6000_expand_extract_even): Update call accordingly.
25642 (rs6000_expand_interleave): Likewise.
25643 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
25644 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
25645 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
25646 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
25647 (rs6000_expand_vec_perm_const): Delete.
25648 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25649 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25650 (altivec_expand_vec_perm_const_le): Take each operand individually.
25651 Operate on constant selectors rather than rtxes.
25652 (altivec_expand_vec_perm_const): Likewise. Update call to
25653 altivec_expand_vec_perm_const_le.
25654 (rs6000_expand_vec_perm_const): Delete.
25655 (rs6000_vectorize_vec_perm_const_ok): Delete.
25656 (rs6000_vectorize_vec_perm_const): New function. Remove stray
25657 reference to the SPE evmerge intructions.
25658 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25659 an element count and rtx array.
25660 (rs6000_expand_extract_even): Update call accordingly.
25661 (rs6000_expand_interleave): Likewise.
25662 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
25663 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
25665 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25667 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25669 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
25670 vector mode and that that mode matches the mode of the data
25672 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
25673 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
25674 directly using expand_vec_perm_1 when forcing selectors into
25676 (expand_vec_perm_var): New function, split out from expand_vec_perm.
25678 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25680 * optabs-query.h (can_vec_perm_p): Delete.
25681 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
25682 * optabs-query.c (can_vec_perm_p): Split into...
25683 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
25684 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
25685 particular selector is valid.
25686 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25687 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25688 (vect_grouped_load_supported): Likewise.
25689 (vect_shift_permute_load_chain): Likewise.
25690 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25691 (vect_transform_slp_perm_load): Likewise.
25692 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25693 (vectorizable_bswap): Likewise.
25694 (vect_gen_perm_mask_checked): Likewise.
25695 * fold-const.c (fold_ternary_loc): Likewise. Don't take
25696 implementations of variable permutation vectors into account
25697 when deciding which selector to use.
25698 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
25699 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
25700 with a false third argument.
25701 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
25702 to test whether the constant selector is valid and can_vec_perm_var_p
25703 to test whether a variable selector is valid.
25705 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25707 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
25708 * optabs-query.c (can_vec_perm_p): Likewise.
25709 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
25710 instead of vec_perm_indices.
25711 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
25712 (vect_gen_perm_mask_checked): Likewise,
25713 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
25714 (vect_gen_perm_mask_checked): Likewise,
25716 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
25718 * optabs-query.h (qimode_for_vec_perm): Declare.
25719 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
25720 (qimode_for_vec_perm): ...this new function.
25721 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
25723 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25725 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
25726 does not have a conditional at the top.
25728 2018-01-02 Richard Biener <rguenther@suse.de>
25730 * ipa-inline.c (big_speedup_p): Fix expression.
25732 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
25735 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
25738 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
25742 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
25743 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
25744 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
25745 cond_taken_branch_cost 3->4.
25747 2018-01-01 Jakub Jelinek <jakub@redhat.com>
25749 PR tree-optimization/83581
25750 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
25751 TODO_cleanup_cfg if any changes have been made.
25753 PR middle-end/83608
25754 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
25755 convert_modes if target mode has the right side, but different mode
25758 PR middle-end/83609
25759 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
25760 last argument when extracting from CONCAT. If either from_real or
25761 from_imag is NULL, use expansion through memory. If result is not
25762 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
25763 the parts directly to inner mode, if even that fails, use expansion
25766 PR middle-end/83623
25767 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
25768 check for bswap in mode rather than HImode and use that in expand_unop
25771 Copyright (C) 2018 Free Software Foundation, Inc.
25773 Copying and distribution of this file, with or without modification,
25774 are permitted in any medium without royalty provided the copyright
25775 notice and this notice are preserved.