1 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
3 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
6 2018-08-22 Martin Sebor <msebor@redhat.com>
9 * tree-pretty-print.c (pretty_print_string): Add argument.
10 (dump_generic_node): Call to pretty_print_string with string size.
12 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
14 PR rtl-optimization/86771
15 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
16 of two SETs into those two SETs, one to be placed at i2, if that SETs
17 destination is modified between i2 and i3.
19 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
21 PR tree-optimization/86725
22 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
24 (vect_analyze_scalar_cycles_1): Check it.
26 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
28 PR tree-optimization/86725
29 * tree-vect-loop.c (vect_is_simple_reduction): When treating
30 an outer loop phi as a double reduction, make sure that the
31 single user of the phi result is an inner loop phi.
33 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
35 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
36 grouped stores with gaps to a strided group.
38 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
40 * tree-vect-stmts.c (get_group_load_store_type)
41 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
42 first statement in a group.
44 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
46 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
47 the sequence used in gcc/gcc.c.
49 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
52 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
55 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
57 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
58 Darwin10-specific unwinder-shim.
59 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
60 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
61 New to cater for Darwin10 Rosetta.
63 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
65 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
66 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
68 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
73 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
74 (dwarf2out_switch_text_section): Generate a local label for the
75 second function sub-section and apply it as the second FDE start
77 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
78 second sub-section start.
80 2018-08-22 Richard Biener <rguenther@suse.de>
82 PR tree-optimization/86988
83 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
85 2018-08-22 Richard Biener <rguenther@suse.de>
87 PR tree-optimization/86945
88 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
90 2018-08-22 Alexandre Oliva <oliva@adacore.com>
92 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
93 a comment about how uses of r2 for .sdata2 come about.
95 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
97 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
99 2018-08-21 Marek Polacek <polacek@redhat.com>
101 PR c++/86981, Implement -Wpessimizing-move.
102 * doc/invoke.texi: Document -Wpessimizing-move.
104 2018-08-21 Jan Hubicka <jh@suse.cz>
106 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
107 * tree.h (is_redundant_typedef): Remove.
108 * dwarf2out.c (is_redundant_typedef): Turn into static function.
110 2018-08-21 Jan Hubicka <jh@suse.cz>
112 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
115 2018-08-21 Tamar Christina <tamar.christina@arm.com>
117 * expmed.c (extract_low_bits): Reject invalid subregs early.
119 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
122 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
125 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
127 * config/vxworks.h: Guard vxworks_asm_out_constructor and
128 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
129 * config/vxworks.c: Likewise.
131 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
133 * config/vxworks.c: Set targetm.have_ctors_dtors
134 if HAVE_INITFINI_ARRAY_SUPPORT.
135 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
136 if HAVE_INITFINI_ARRAY_SUPPORT.
138 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
140 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
141 default search path for VxWorks < 7.
143 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
145 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
146 (get_format_string): Refer to c_getstr.
148 2018-08-21 Tom de Vries <tdevries@suse.de>
150 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
151 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
152 (debuginfo_early_stop): Declare.
153 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
154 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
155 (debuginfo_early_stop): New function.
156 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
157 and debuginfo_early_stop.
158 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
159 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
160 (general_init): Call debuginfo_early_init.
161 (finalize): Call debuginfo_fini.
162 (do_compile): Call debuginfo_init.
163 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
165 (@item -fdump-debug, @item -fdump-earlydebug): Add.
167 2018-08-21 Tom de Vries <tdevries@suse.de>
169 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
170 flag_dump_noaddr and flag_dump_unnumbered.
172 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
174 * wide-int-range.cc (wide_int_range_abs): New.
175 (wide_int_range_order_set): Rename from wide_int_range_min_max.
176 * wide-int-range.h (wide_int_range_abs): New.
177 (wide_int_range_min_max): New.
178 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
179 case to call wide_int_range_abs.
180 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
181 (extract_range_from_abs_expr): Delete.
183 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
186 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
187 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
190 2018-08-20 Nathan Sidwell <nathan@acm.org>
191 Jeff Law <law@redhat.com>
193 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
194 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
196 2018-08-20 David Malcolm <dmalcolm@redhat.com>
199 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
200 (decl_attributes): Likewise.
201 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
203 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
204 * diagnostic-core.h (class auto_diagnostic_group): New class.
205 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
206 (diagnostic_report_diagnostic): Handle the first diagnostics within
208 (emit_diagnostic): Add auto_diagnostic_group instance.
210 (inform_n): Likewise.
212 (warning_at): Likewise.
213 (warning_n): Likewise.
215 (permerror): Likewise.
218 (error_at): Likewise.
220 (fatal_error): Likewise.
221 (internal_error): Likewise.
222 (internal_error_no_backtrace): Likewise.
223 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
224 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
225 * diagnostic.h (struct diagnostic_context): Add fields
226 "diagnostic_group_nesting_depth",
227 "diagnostic_group_emission_count", "begin_group_cb",
229 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
230 Add auto_diagnostic_group instance(s).
231 (find_explicit_erroneous_behavior): Likewise.
232 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
233 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
234 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
235 (gimplify_va_arg_expr): Likewise.
236 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
237 (HSA_SORRY_AT): Likewise.
238 * ipa-devirt.c (compare_virtual_tables): Likewise.
239 (warn_odr): Likewise.
240 * multiple_target.c (expand_target_clones): Likewise.
241 * opts-common.c (cmdline_handle_error): Likewise.
242 * reginfo.c (globalize_reg): Likewise.
243 * substring-locations.c (format_warning_n_va): Likewise.
244 * tree-inline.c (expand_call_inline): Likewise.
245 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
246 * tree-ssa-loop-niter.c
247 (do_warn_aggressive_loop_optimizations): Likewise.
248 * tree-ssa-uninit.c (warn_uninit): Likewise.
249 * tree.c (warn_deprecated_use): Likewise.
251 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
254 * config/i386/i386.md (eh_return): Always update EH return
255 address in word_mode.
257 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
259 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
260 TARGET_SPLIT_COMPLEX_ARG is defined.
262 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
264 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
266 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
269 * expr.c (expand_assignment): Assert that bitpos is positive.
270 (store_field): Likewise
271 (expand_expr_real_1): Make sure that bitpos is positive.
272 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
275 2018-08-20 Nathan Sidwell <nathan@acm.org>
277 * Makefile.in (CPP_ID_DATA_H): Delete.
278 (CPP_INTERNAL_H): Don't add it.
279 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
280 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
282 2018-08-20 Richard Biener <rguenther@suse.de>
284 PR tree-optimization/78655
285 * tree-vrp.c (extract_range_from_binary_expr_1): Make
286 pointer + offset nonnull if either operand is nonnull work.
288 2018-08-20 Tom de Vries <tdevries@suse.de>
290 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
291 unless the referenced die describes the added property using
292 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
293 Otherwise, add a DW_AT_location to the referenced die.
295 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
298 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
299 register_operand when calling ix86_set_reg_reg_cost.
300 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
301 Set *total to 0 for operands that satisfy x86_64_immediate_operand
302 predicate and to 1 otherwise.
304 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
306 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
307 emit a diagnostic that it is not supported and reset the option.
308 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
309 supported and consume the option. (ASM_FINAL_SPEC): New.
311 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
313 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
316 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
318 C-SKY port: Documentation
320 * doc/extend.texi (C-SKY Function Attributes): New section.
321 * doc/invoke.texi (Option Summary): Add C-SKY options.
322 (C-SKY Options): New section.
323 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
325 2018-08-17 Jojo <jijie_rong@c-sky.com>
326 Huibin Wang <huibin_wang@c-sky.com>
327 Sandra Loosemore <sandra@codesourcery.com>
328 Chung-Lin Tang <cltang@codesourcery.com>
330 C-SKY port: Backend implementation
332 * config/csky/*: New.
333 * common/config/csky/*: New.
335 2018-08-17 Jojo <jijie_rong@c-sky.com>
336 Huibin Wang <huibin_wang@c-sky.com>
337 Sandra Loosemore <sandra@codesourcery.com>
338 Chung-Lin Tang <cltang@codesourcery.com>
339 Andrew Jenner <andrew@codesourcery.com>
341 C-SKY port: Configury
343 * config.gcc (csky-*-*): New.
344 * configure.ac: Add csky to targets for dwarf2 debug_line support.
345 * configure: Regenerated.
347 2018-08-17 David Malcolm <dmalcolm@redhat.com>
349 * dump-context.h: Include "dumpfile.h".
350 (dump_context::dump_printf_va): Convert final param from va_list
351 to va_list *. Convert from ATTRIBUTE_PRINTF to
352 ATTRIBUTE_GCC_DUMP_PRINTF.
353 (dump_context::dump_printf_loc_va): Likewise.
354 * dumpfile.c: Include "stringpool.h".
355 (make_item_for_dump_printf_va): Delete.
356 (make_item_for_dump_printf): Delete.
357 (class dump_pretty_printer): New class.
358 (dump_pretty_printer::dump_pretty_printer): New ctor.
359 (dump_pretty_printer::emit_items): New member function.
360 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
362 (dump_pretty_printer::emit_item): New member function.
363 (dump_pretty_printer::stash_item): New member function.
364 (dump_pretty_printer::format_decoder_cb): New member function.
365 (dump_pretty_printer::decode_format): New member function.
366 (dump_context::dump_printf_va): Reimplement in terms of
368 (dump_context::dump_printf_loc_va): Convert final param from va_list
370 (dump_context::begin_scope): Reimplement call to
371 make_item_for_dump_printf.
372 (dump_printf): Update for change to dump_printf_va.
373 (dump_printf_loc): Likewise.
374 (selftest::test_capture_of_dump_calls): Convert "stmt" from
375 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
377 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
378 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
379 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
380 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
381 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
382 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
383 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
384 within a dump_printf_loc call to "%wu".
385 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
386 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
387 missing space after "=".
388 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
389 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
390 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
391 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
392 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
393 duplicate "vectorized" from message.
395 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
397 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
398 polyNxK_t element's TYPE_STRING_FLAG.
400 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
402 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
403 (they were unnamed before). Fix comments.
405 2018-08-17 Nathan Sidwell <nathan@acm.org>
407 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
409 2018-08-17 Richard Biener <rguenther@suse.de>
411 PR tree-optimization/86841
412 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
414 2018-08-17 Martin Liska <mliska@suse.cz>
416 * common.opt: Remove Warn, Init and Report for options with
417 Ignore/Deprecated flag. Warning is done automatically for
419 * config/i386/i386.opt: Likewise.
420 * config/ia64/ia64.opt: Likewise.
421 * config/rs6000/rs6000.opt: Likewise.
422 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
423 Remove usage of flag_check_pointer_bounds.
424 * lto-wrapper.c (merge_and_complain): Do not handle
425 OPT_fcheck_pointer_bounds.
426 (append_compiler_options): Likewise.
427 * opt-functions.awk: Do not handle Deprecated.
428 * optc-gen.awk: Check that Var, Report and Init are not
429 used for an option with Ignore/Deprecated flag.
430 * opts-common.c (decode_cmdline_option): Do not report
432 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
434 * opts.h (struct cl_option): Remove cl_deprecated flag.
435 (CL_ERR_DEPRECATED): Remove error enum value.
437 2018-08-17 Richard Biener <rguenther@suse.de>
440 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
441 across a va-arg-pack using call adjust its return value accordingly.
443 2018-08-16 Martin Sebor <msebor@redhat.com>
445 PR tree-optimization/86853
446 * gimple-ssa-sprintf.c (struct format_result): Rename member.
447 (struct fmtresult): Add member and initialize it in ctors.
448 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
449 (format_string): Handle %S the same as %ls. Set MAYFAIL.
450 (format_directive): Set POSUNDER4K when MAYFAIL is set.
451 (parse_directive): Handle %C same as %c.
452 (sprintf_dom_walker::compute_format_length): Adjust.
453 (is_call_safe): Adjust.
455 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
457 * builtins.c (c_strlen): Add new parameter eltsize. Use it
458 for determining how to count the elements.
459 * builtins.h (c_strlen): Adjust prototype.
460 * expr.c (string_constant): Add new parameter mem_size.
461 Set *mem_size appropriately.
462 * expr.h (string_constant): Adjust protoype.
463 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
464 * gimple-fold.h (get_range_strlen): Adjust prototype.
465 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
466 (format_string): Call get_string_length with eltsize.
468 2018-08-16 David Malcolm <dmalcolm@redhat.com>
470 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
471 to emit the span, rather than setting it as the prefix.
473 2018-08-16 David Malcolm <dmalcolm@redhat.com>
475 * diagnostic-show-locus.c (layout::start_annotation_line): Add
476 "margin_char" parameter, defaulting to space. Use it in place
477 of pp_space for the initial part of the margin.
478 (layout::print_leading_fixits): Use '+' when filling the margin
479 of line-insertion fix-it hints.
481 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
483 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
486 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
488 * config/rs6000/altivec.md: Don't set length attribute to the default
490 * config/rs6000/darwin.md: Ditto.
491 * config/rs6000/dfp.md: Ditto.
492 * config/rs6000/htm.md: Ditto.
493 * config/rs6000/rs6000.md: Ditto.
494 * config/rs6000/sync.md: Ditto.
495 * config/rs6000/vsx.md: Ditto.
497 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
499 * config/rs6000/altivec.md: Don't set length attribute to the default
500 value, for branch instructions.
501 * config/rs6000/darwin.md: Ditto.
502 * config/rs6000/rs6000.md: Ditto.
504 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
506 * config/rs6000/rs6000.md (length): Always define as const_int 4.
507 (unnamed conditional branch define_insn): Set length to 4 or 8
509 (<bd>_<mode>): Similar, for alternative 0.
510 (<bd>tf_<mode>): Ditto.
512 2018-08-16 Tamar Christina <tamar.christina@arm.com>
514 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
516 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
518 * doc/rtl.texi: Replace old RTX class names with new names.
521 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
523 * expmed.h (canonicalize_comparison): New declaration.
524 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
525 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
526 * optabs.c (prepare_cmp_insn): Likewise.
527 * rtl.h (unsigned_condition_p): New function which checks if a
528 comparison operator is unsigned.
530 2018-08-16 Nathan Sidwell <nathan@acm.org>
532 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
533 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
535 2018-08-16 Tamar Christina <tamar.christina@arm.com>
538 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
539 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
540 (mov<mov>): ..this and enable unconditionally.
542 2018-08-16 Tamar Christina <tamar.christina@arm.com>
544 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
546 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
548 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
549 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
550 "Common" with "Target".
552 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
554 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
555 * doc/invoke.texi (mmitigate-rop): Remove.
556 * config/i386/i386.c: Do not include "regrename.h".
557 (ix86_rop_should_change_byte_p, reg_encoded_number)
558 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
560 (ix86_reorg): Remove call to ix86_mitigate_rop.
561 * config/i386/i386.md (attr "modrm_class"): Remove.
562 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
563 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
564 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
566 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
568 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
569 allow folding of mergeh() and mergel() for the float and double types.
570 (fold_mergehl_helper): Rework to handle building a permute tree
573 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
575 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
578 2018-08-15 David Malcolm <dmalcolm@redhat.com>
580 * common.opt (fdiagnostics-show-labels): New option.
581 * diagnostic-show-locus.c (class layout_range): Add field
583 (class layout): Add field "m_show_labels_p".
584 (layout_range::layout_range): Add param "label" and use it to
586 (make_range): Pass in NULL for new "label" param of layout_range's
588 (layout::layout): Initialize m_show_labels_p.
589 (layout::maybe_add_location_range): Pass in loc_range->m_label
590 when constructing layout_range instances.
591 (struct line_label): New struct.
592 (layout::print_any_labels): New member function.
593 (layout::print_line): Call it if label-printing is enabled.
594 (selftest::test_one_liner_labels): New test.
595 (selftest::test_diagnostic_show_locus_one_liner): Call it.
596 * diagnostic.c (diagnostic_initialize): Initialize
597 context->show_labels_p.
598 * diagnostic.h (struct diagnostic_context): Add field
600 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
601 -fno-diagnostics-show-labels.
602 * dwarf2out.c (gen_producer_string): Add
603 OPT_fdiagnostics_show_labels to the ignored options.
604 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
606 (gcc_rich_location::maybe_add_expr): Likewise.
607 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
608 label" param, defaulting to NULL.
609 (gcc_rich_location::add_expr): Add "label" param.
610 (gcc_rich_location::maybe_add_expr): Likewise.
611 (class text_range_label): New class.
612 (class range_label_for_type_mismatch): New class.
613 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
614 of format_warning_va.
615 (fmtwarn_n): Likewise for new params of format_warning_n_va.
616 * lto-wrapper.c (merge_and_complain): Add
617 OPT_fdiagnostics_show_labels to the "pick one setting" options.
618 (append_compiler_options): Likewise to the dropped options.
619 (append_diag_options): Likewise to the passed-on options.
620 * opts.c (common_handle_option): Handle the new option.
621 * selftest-diagnostic.c
622 (test_diagnostic_context::test_diagnostic_context): Enable
624 * substring-locations.c: Include "gcc-rich-location.h".
625 (format_warning_n_va): Add "fmt_label" and "param_label" params
626 and use them as appropriate.
627 (format_warning_va): Add "fmt_label" and "param_label" params,
628 passing them on to format_warning_n_va.
629 (format_warning_at_substring): Likewise.
630 (format_warning_at_substring_n): Likewise.
631 * substring-locations.h (format_warning_va): Add "fmt_label" and
632 "param_label" params.
633 (format_warning_n_va): Likewise.
634 (format_warning_at_substring): Likewise.
635 (format_warning_at_substring_n): Likewise.
636 * toplev.c (general_init): Initialize global_dc->show_labels_p.
638 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
641 * builtins.c (expand_builtin_memcmp): Do not expand the call
642 when overflow is detected.
644 2018-08-15 Martin Sebor <msebor@redhat.com>
646 PR tree-optimization/71625
647 * config/aarch64/aarch64-builtins.c
648 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
650 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
652 * config/s390/s390.c (s390_reorg): Remove loop.
654 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
657 (darwin_function_switched_text_sections): Delete.
658 * gcc/config/darwin.h
659 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
661 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
664 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
665 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
666 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
668 2018-08-15 Martin Liska <mliska@suse.cz>
670 PR tree-optimization/86925
671 * predict.c (expr_expected_value_1): When taking
672 later predictor, assign also probability.
673 Use fold_build2_initializer_loc in order to fold
674 the expression in -frounding-math.
676 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
678 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
680 (expand_vec_perm_1): Try the new method.
682 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
685 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
686 Check whether lra_live_max_point is 0 before dividing.
688 2018-08-14 Martin Sebor <msebor@redhat.com>
690 PR tree-optimization/86650
691 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
692 (vrp_prop::check_mem_ref): Same.
694 2018-08-13 Liu Hao <lh_mouse@126.com>
696 * pretty-print.c (eat_esc_sequence): Swap the foreground and
697 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
698 and clear it thereafter, as it only works for DBCS.
700 2018-08-13 Liu Hao <lh_mouse@126.com>
702 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
703 handle returned by _get_osf_handle().
705 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
707 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
708 for folding vec_perm.
710 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
712 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
713 Add support for gimple-folding of vec_pack() and vec_unpack()
716 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
718 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
719 vec_xst variants to the list.
720 (rs6000_gimple_fold_builtin): Add support for folding unaligned
721 vector loads and stores.
723 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
725 * config.gcc (rs6000-ibm-aix4.x): Delete.
726 (rs6000-ibm-aix5.1): Delete.
727 (rs6000-ibm-aix5.2): Delete.
728 (rs6000-ibm-aix5.3): Delete.
729 * config/rs6000/aix43.h: Delete.
730 * config/rs6000/aix51.h: Delete.
731 * config/rs6000/aix52.h: Delete.
732 * config/rs6000/t-aix43: Delete.
734 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
736 * config/s390/s390.c (s390_decompose_constant_pool_ref):
738 (s390_decompose_address): Factor out constant pool ref
741 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
743 * config/nds32/nds32-predicates.c
744 (nds32_can_use_bclr_p): Change return type as bool.
745 (nds32_can_use_bset_p): Ditto.
746 (nds32_can_use_btgl_p): Ditto.
747 (nds32_can_use_bitci_p): Ditto.
748 * config/nds32/nds32-protos.h
749 (nds32_can_use_bclr_p): Change declaration.
750 (nds32_can_use_bset_p): Ditto.
751 (nds32_can_use_btgl_p): Ditto.
752 (nds32_can_use_bitci_p): Ditto.
754 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
756 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
757 Support -msched-prolog-epilog option.
758 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
760 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
762 * common/config/nds32/nds32-common.c
763 (nds32_option_optimization_table): Enalbe -malways-align.
765 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
767 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
769 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
770 OPT_misr_secure_ case.
771 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
772 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
773 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
775 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
776 (nds32_isr_info): New field security_level.
777 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
778 * config/nds32/nds32.md (return_internal): Consider critical attribute.
779 * config/nds32/nds32.opt (misr-secure): New option.
780 * config/nds32/nds32_init.inc: New file.
781 * config/nds32/nds32_isr.h: New file.
783 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
785 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
786 Update comment for atomic instructions.
787 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
788 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
790 (atomic_loaddi): Revise fence expansion to only emit fence prior to
791 load for __ATOMIC_SEQ_CST model.
792 (atomic_loaddi_1): Remove float register target.
793 (atomic_storedi): Handle CONST_INT values.
794 (atomic_storedi_1): Remove float register source. Add special case
796 (memory_barrier): New expander and insn.
798 2018-08-11 Jakub Jelinek <jakub@redhat.com>
800 PR tree-optimization/86835
801 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
802 new_stmt after def_gsi, make sure to insert new_square_stmt after
803 that stmt, not 2 stmts before it.
805 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
808 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
811 2018-08-10 Martin Liska <mliska@suse.cz>
814 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
816 * builtins.c (expand_builtin_expect_with_probability):
818 (expand_builtin_expect_with_probability): New function.
819 (build_builtin_expect_predicate): Add new argumnet probability
820 for BUILT_IN_EXPECT_WITH_PROBABILITY.
821 (fold_builtin_expect):
824 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
825 * builtins.h (fold_builtin_expect): Set new argument.
826 * doc/extend.texi: Document __builtin_expect_with_probability.
827 * doc/invoke.texi: Likewise.
828 * gimple-fold.c (gimple_fold_call): Pass new argument.
829 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
830 also BUILT_IN_EXPECT_WITH_PROBABILITY.
831 * predict.c (get_predictor_value): New function.
832 (expr_expected_value): Add new argument probability. Assume
833 that predictor and probability are always non-null.
834 (expr_expected_value_1): Likewise. For __builtin_expect and
835 __builtin_expect_with_probability set probability. Handle
836 combination in binary expressions.
837 (tree_predict_by_opcode): Simplify code by simply calling
839 (pass_strip_predict_hints::execute): Add handling of
840 BUILT_IN_EXPECT_WITH_PROBABILITY.
841 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
843 * tree.h (DECL_BUILT_IN_P): New function.
845 2018-08-10 Martin Liska <mliska@suse.cz>
847 PR tree-optimization/85799
848 * passes.def: Add argument for pass_strip_predict_hints.
849 * predict.c (class pass_strip_predict_hints): Add new argument
851 (strip_predictor_early): New function.
852 (pass_strip_predict_hints::execute): Call the function to
854 (strip_predict_hints): New function.
855 * predict.def: Fix comment.
857 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
859 * Makefile.in: Clarify which tm.texi to copy over to assert the
860 right to grant a GFDL license for all.
862 2018-08-09 Jeff Law <law@redhat.com>
864 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
867 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
869 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
872 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
874 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
875 reductions for variable-length vectors.
877 2018-08-09 David Malcolm <dmalcolm@redhat.com>
880 * common.opt (fdiagnostics-show-line-numbers): New option.
881 * diagnostic-show-locus.c (class layout): Add fields
882 "m_show_line_numbers_p" and "m_linenum_width";
883 (num_digits): New function.
884 (test_num_digits): New function.
885 (layout::layout): Initialize new fields. Update m_x_offset
886 logic to handle any left margin.
887 (layout::print_source_line): Print line number when requested.
888 (layout::start_annotation_line): New member function.
889 (layout::print_annotation_line): Call it.
890 (layout::print_leading_fixits): Likewise.
891 (layout::print_trailing_fixits): Likewise. Update calls to
892 move_to_column for new parameter.
893 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
894 it to potentially call start_annotation_line.
895 (layout::show_ruler): Call start_annotation_line.
896 (selftest::test_line_numbers_multiline_range): New selftest.
897 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
898 and selftest::test_line_numbers_multiline_range.
899 * diagnostic.c (diagnostic_initialize): Initialize
901 * diagnostic.h (struct diagnostic_context): Add field
902 "show_line_numbers_p".
903 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
904 -fno-diagnostics-show-line-numbers.
905 * dwarf2out.c (gen_producer_string): Add
906 OPT_fdiagnostics_show_line_numbers to the ignored options.
907 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
908 one setting" options.
909 (append_compiler_options): Likewise to the dropped options.
910 (append_diag_options): Likewise to the passed-on options.
911 * opts.c (common_handle_option): Handle the new option.
912 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
914 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
916 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
917 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
918 third argument of type "const signed char" to descriptions of
919 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
920 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
921 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
922 __builtin_bcdsub_ov functions.
924 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
926 PR tree-optimization/86858
927 * tree-vect-loop.c (vect_is_simple_reduction): Restore
928 flow_bb_inside_loop_p calls.
930 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
932 PR tree-optimization/86871
933 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
934 instead of gimple_assign_lhs.
936 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
939 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
940 register constraint to operand 0.
941 (add<mode>3_carryinC): Likewise.
942 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
944 2018-08-09 Martin Liska <mliska@suse.cz>
947 * common.opt: Remove extra line.
949 2018-08-09 Martin Liska <mliska@suse.cz>
951 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
952 at the end of a line, make first letter capital and end up
953 a sentence with a dot.
954 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
955 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
956 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
957 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
958 (PARAM_MAX_ISL_OPERATIONS): Likewise.
959 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
960 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
961 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
962 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
963 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
964 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
965 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
966 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
967 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
968 (PARAM_TREE_REASSOC_WIDTH): Likewise.
969 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
970 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
971 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
973 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
976 * config/s390/s390.c (s390_option_override_internal): Reduce the
977 stack-clash-protection-probe-interval param if it would be too big
980 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
983 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
984 * config/m68k/m68k.c (handle_move_double): Don't call
985 m68k_final_prescan_insn.
986 (m68k_adjust_decorated_operand): Renamed from
987 m68k_final_prescan_insn, remove first and third operand and
989 (print_operand): Call it.
990 (print_operand_address): Call it.
992 2018-08-08 Nathan Sidwell <nathan@acm.org>
994 * diagnostic.c (diagnostic_report_current_module): Use
995 linemap_included_from & linemap_included_from_linemap.
997 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
999 * config/aarch64/aarch64-cores.def: Add phecda core.
1000 * config/aarch64/aarch64-tune.md: Regenerate.
1001 * doc/invoke.texi: Add phecda core.
1003 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
1006 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
1008 * config/s390/s390.md ("movti"): Add more alternatives for
1009 constant to GPR copies.
1011 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
1013 * config/s390/s390.c: Fix whitespace damage throughout the file.
1014 * config/s390/s390.h: Likewise.
1015 * config/s390/tpf.h: Likewise.
1017 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
1019 * config/s390/s390.c (s390_loadrelative_operand_p):
1020 Remove TARGET_CPU_ZARCH usages.
1021 (s390_rtx_costs): Likewise.
1022 (s390_legitimate_constant_p): Likewise.
1023 (s390_cannot_force_const_mem): Likewise.
1024 (legitimate_reload_constant_p): Likewise.
1025 (s390_preferred_reload_class): Likewise.
1026 (legitimize_pic_address): Likewise.
1027 (legitimize_tls_address): Likewise.
1028 (s390_split_branches): Removed.
1029 (s390_add_execute): Removed.
1030 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
1031 (s390_mainpool_start): Likewise.
1032 (s390_mainpool_finish): Likewise.
1033 (s390_mainpool_cancel): Removed.
1034 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
1035 (s390_chunkify_cancel): Likewise.
1036 (s390_return_addr_rtx): Likewise.
1037 (s390_register_info): Remove split_branches_pending_p uages.
1038 (s390_optimize_register_info): Likewise.
1039 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
1040 split_branches_pending_p usages.
1041 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
1042 (s390_load_got): Likewise.
1043 (s390_expand_split_stack_prologue): Likewise.
1044 (output_asm_nops): Likewise.
1045 (s390_function_profiler): Likewise.
1046 (s390_emit_call): Likewise.
1047 (s390_conditional_register_usage): Likewise.
1048 (s390_optimize_prologue): Likewise.
1049 (s390_reorg): Remove TARGET_CPU_ZARCH and
1050 split_branches_pending_p usages.
1051 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
1053 (s390_output_indirect_thunk_function): Likewise.
1054 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
1055 (TARGET_CPU_ZARCH_P): Removed.
1056 (struct machine_function): Remove split_branches_pending_p.
1057 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
1059 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
1061 * common/config/s390/s390-common.c (processor_flags_table):
1063 * config.gcc: Remove with_arch/with_tune support.
1064 * config/s390/2064.md: Remove cpu attribute comparisons.
1065 * config/s390/driver-native.c (s390_host_detect_local_cpu):
1067 * config/s390/linux.h (ASM_SPEC):
1068 Remove -march support.
1069 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
1070 Use a table to get an arch level.
1071 * config/s390/s390-opts.h (enum processor_type):
1073 * config/s390/s390.c
1074 (processor_table): Remove entries, add arch_level values.
1075 (s390_issue_rate): Remove cases.
1076 (s390_option_override): Adjust
1077 s390_option_override_internal() call.
1078 (s390_option_override_internal): Remove deprecation warning.
1079 (s390_valid_target_attribute_tree): Adjust
1080 s390_option_override_internal() call.
1081 * config/s390/s390.h (struct s390_processor):
1082 Share with s390-c.c, add arch_level field.
1083 * config/s390/s390.md:
1084 Remove occurrences in cpu attribute.
1085 * config/s390/s390.opt: Remove -march/-mtune support.
1086 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
1087 * doc/invoke.texi: Remove deprecation warning.
1089 2018-08-08 Luis Machado <luis.machado@linaro.org>
1091 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
1093 (qdf24xx_tunings): Set vector cost structure to
1094 qdf24xx_vector_cost.
1096 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
1097 <register_sextend>: Set to 3.
1099 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
1102 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
1103 * config/aarch64/aarch64-simd.md
1104 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
1105 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
1106 (aarch64_frecpx<mode>): ...this new pattern.
1107 * config/aarch64/aarch64-simd-builtins.def: Remove comment
1108 about aarch64_frecp<FRECP:frecp_suffix><mode>.
1110 2018-08-07 Martin Liska <mliska@suse.cz>
1113 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
1114 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
1115 * predict.def (PRED_MALLOC_NONNULL): New predictor.
1116 * doc/extend.texi: Document that malloc attribute adds
1119 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
1122 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1123 Define to speculation_safe_value_not_needed.
1125 2018-08-06 Jeff Law <law@redhat.com>
1127 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
1128 the vr_values instance to cprop_into_stmt.
1129 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
1130 (cprop_operand): Also query EVRP to determine if OP is a constant.
1132 2018-08-06 Nathan Sidwell <nathan@acm.org>
1134 * diagnostic.c (diagnostic_report_current_module): Reroll
1135 included-at loop. Translate text.
1137 2018-08-06 David Malcolm <dmalcolm@redhat.com>
1139 * function-tests.c (selftest::test_expansion_to_rtl): Call
1140 free_after_compilation.
1142 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1144 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
1146 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
1148 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
1149 loops with memory block operations from getting unrolled.
1151 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
1154 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1155 Define to speculation_safe_value_not_needed.
1157 2018-08-06 Jeff Law <law@redhat.com>
1159 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
1162 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1165 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
1166 with all enabled __intN types.
1168 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
1170 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1172 * alias.c (record_set): Check for clobber high.
1173 * cfgexpand.c (expand_gimple_stmt): Likewise.
1174 * combine-stack-adj.c (single_set_for_csa): Likewise.
1175 * combine.c (find_single_use_1): Likewise.
1176 (set_nonzero_bits_and_sign_copies): Likewise.
1177 (get_combine_src_dest): Likewise.
1178 (is_parallel_of_n_reg_sets): Likewise.
1179 (try_combine): Likewise.
1180 (record_dead_and_set_regs_1): Likewise.
1181 (reg_dead_at_p_1): Likewise.
1182 (reg_dead_at_p): Likewise.
1183 * dce.c (deletable_insn_p): Likewise.
1184 (mark_nonreg_stores_1): Likewise.
1185 (mark_nonreg_stores_2): Likewise.
1186 * df-scan.c (df_find_hard_reg_defs): Likewise.
1187 (df_uses_record): Likewise.
1188 (df_get_call_refs): Likewise.
1189 * dwarf2out.c (mem_loc_descriptor): Likewise.
1190 * haifa-sched.c (haifa_classify_rtx): Likewise.
1191 * ira-build.c (create_insn_allocnos): Likewise.
1192 * ira-costs.c (scan_one_insn): Likewise.
1193 * ira.c (equiv_init_movable_p): Likewise.
1194 (rtx_moveable_p): Likewise.
1195 (interesting_dest_for_shprep): Likewise.
1196 * jump.c (mark_jump_label_1): Likewise.
1197 * postreload-gcse.c (record_opr_changes): Likewise.
1198 * postreload.c (reload_cse_simplify): Likewise.
1199 (struct reg_use): Add source expr.
1200 (reload_combine): Check for clobber high.
1201 (reload_combine_note_use): Likewise.
1202 (reload_cse_move2add): Likewise.
1203 (move2add_note_store): Likewise.
1204 * print-rtl.c (print_pattern): Likewise.
1205 * recog.c (decode_asm_operands): Likewise.
1206 (store_data_bypass_p): Likewise.
1207 (if_test_bypass_p): Likewise.
1208 * regcprop.c (kill_clobbered_value): Likewise.
1209 (kill_set_value): Likewise.
1210 * reginfo.c (reg_scan_mark_refs): Likewise.
1211 * reload1.c (maybe_fix_stack_asms): Likewise.
1212 (eliminate_regs_1): Likewise.
1213 (elimination_effects): Likewise.
1214 (mark_not_eliminable): Likewise.
1215 (scan_paradoxical_subregs): Likewise.
1216 (forget_old_reloads_1): Likewise.
1217 * reorg.c (find_end_label): Likewise.
1218 (try_merge_delay_insns): Likewise.
1219 (redundant_insn): Likewise.
1220 (own_thread_p): Likewise.
1221 (fill_simple_delay_slots): Likewise.
1222 (fill_slots_from_thread): Likewise.
1223 (dbr_schedule): Likewise.
1224 * resource.c (update_live_status): Likewise.
1225 (mark_referenced_resources): Likewise.
1226 (mark_set_resources): Likewise.
1227 * rtl.c (copy_rtx): Likewise.
1228 * rtlanal.c (reg_referenced_p): Likewise.
1229 (single_set_2): Likewise.
1230 (noop_move_p): Likewise.
1231 (note_stores): Likewise.
1232 * sched-deps.c (sched_analyze_reg): Likewise.
1233 (sched_analyze_insn): Likewise.
1235 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1237 * cse.c (invalidate_reg): New function extracted from...
1238 (invalidate): ...here.
1239 (canonicalize_insn): Check for clobber high.
1240 (invalidate_from_clobbers): invalidate clobber highs.
1241 (invalidate_from_sets_and_clobbers): Likewise.
1242 (count_reg_usage): Check for clobber high.
1243 (insn_live_p): Likewise.
1244 * cselib.c (cselib_expand_value_rtx_1):Likewise.
1245 (cselib_invalidate_regno): Check for clobber in setter.
1246 (cselib_invalidate_rtx): Pass through setter.
1247 (cselib_invalidate_rtx_note_stores):
1248 (cselib_process_insn): Check for clobber high.
1249 * cselib.h (cselib_invalidate_rtx): Add operand.
1251 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1253 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
1254 (mark_not_eliminable): Likewise.
1255 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
1256 * lra-lives.c (process_bb_lives): Check for clobber high.
1257 * lra.c (new_insn_reg): Remember clobber highs.
1258 (collect_non_operand_hard_regs): Check for clobber high.
1259 (lra_set_insn_recog_data): Likewise.
1260 (add_regs_to_insn_regno_info): Likewise.
1261 (lra_update_insn_regno_info): Likewise.
1263 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1265 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
1266 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
1268 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1270 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
1271 (copy_insn_1): Likewise.
1272 (gen_hard_reg_clobber_high): New gen function.
1273 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
1274 * genemit.c (gen_exp): Likewise.
1275 (gen_emit_seq): Pass through info.
1276 (gen_insn): Check for CLOBBER_HIGH.
1277 (gen_expand): Pass through info.
1278 (gen_split): Likewise.
1279 (output_add_clobbers): Likewise.
1280 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
1281 (remove_clobbers): Likewise.
1282 * rtl.h (gen_hard_reg_clobber_high): New declaration.
1284 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1286 * doc/rtl.texi (clobber_high): Add.
1287 (parallel): Add in clobber high
1288 * rtl.c (rtl_check_failed_code3): Add function.
1289 * rtl.def (CLOBBER_HIGH): Add expression.
1290 * rtl.h (RTL_CHECKC3): Add macro.
1291 (rtl_check_failed_code3): Add declaration.
1292 (XC3EXP): Add macro.
1294 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
1297 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
1298 cfun->machine->max_used_stack_alignment if needed.
1300 2018-08-04 Martin Sebor <msebor@redhat.com>
1302 PR tree-optimization/86571
1303 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
1306 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
1309 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1312 2018-08-03 Jeff Law <law@redhat.com>
1315 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1316 Define to speculation_safe_value_not_needed.
1318 2018-08-03 David Malcolm <dmalcolm@redhat.com>
1320 * doc/gcov.texi (-x): Remove duplicate "to".
1321 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
1322 (-Wif-not-aligned): Remove duplicate "is".
1323 (-flto): Remove duplicate "the".
1324 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
1325 duplicate "v5.00.b".
1326 (MSP430 Options): Remove duplicate "and" from the description
1327 of "-mgprel-sec=regexp".
1328 (x86 Options): Remove duplicate copies of "vmldLog102" and
1329 vmlsLog104 from description of "-mveclibabi=type".
1331 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
1333 * internal-fn.h (first_commutative_argument): Declare.
1334 * internal-fn.c (first_commutative_argument): New function.
1335 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
1336 restrictions for pattern statements. Use first_commutative_argument
1337 to look for commutative operands in calls to internal functions.
1339 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
1341 * Makefile.in (wide-int-range.o): New.
1342 * tree-vrp.c: Move all the wide_int_* functions to...
1343 * wide-int-range.cc: ...here.
1344 * tree-vrp.h: Move all the wide_int_* prototypes to...
1345 * wide-int-range.h: ...here.
1347 2018-08-03 Tom de Vries <tdevries@suse.de>
1349 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
1351 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
1352 * except.c (output_function_exception_table): Do early exit if
1353 targetm_common.except_unwind_info (&global_options) == UI_NONE.
1355 2018-08-03 Martin Liska <mliska@suse.cz>
1357 * predict.c (dump_prediction): Change to 2 digits
1360 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
1362 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
1364 (falkor_am_1_gtov_gtov): ... a new insn reservation.
1366 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
1368 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
1369 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
1370 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
1371 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
1372 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
1374 2018-08-02 David Malcolm <dmalcolm@redhat.com>
1376 * diagnostic-show-locus.c (diagnostic_show_locus): Use
1377 pp_take_prefix when saving the existing prefix.
1378 * diagnostic.c (diagnostic_append_note): Likewise.
1379 * langhooks.c (lhd_print_error_function): Likewise.
1380 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
1381 param's type. Free the existing prefix.
1382 (pp_take_prefix): New function.
1383 (pretty_printer::pretty_printer): Drop the prefix parameter.
1384 Rename the length parameter to match the comment.
1385 (pretty_printer::~pretty_printer): Free the prefix.
1386 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
1388 (struct pretty_printer): Drop the "const" from "prefix" field's
1389 type and clarify memory management.
1390 (pp_set_prefix): Drop the "const" from the 2nd param.
1391 (pp_take_prefix): New decl.
1393 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
1395 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
1396 for word_mode_ok here instead of passing as argument.
1397 (expand_block_compare): Change select_block_compare_mode() call.
1398 (expand_strncmp_gpr_sequence): New function.
1399 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
1401 2018-08-02 Jeff Law <law@redhat.com>
1404 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1405 Define to speculation_safe_value_not_needed.
1408 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1409 Define to speculation_safe_value_not_needed.
1411 2018-08-02 Tom de Vries <tdevries@suse.de>
1414 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
1415 function. Return UI_TARGET unconditionally.
1416 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
1417 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
1419 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1421 * genemit.c (print_overload_test): Fix typo.
1423 2018-08-02 Richard Biener <rguenther@suse.de>
1425 PR tree-optimization/86816
1426 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
1427 which checks for value availability before querying it.
1428 (gvn_uses_equal): Use it.
1429 (same_succ_hash): Likewise.
1430 (gimple_equal_p): Likewise.
1432 2018-08-02 Nick Clifton <nickc@redhat.com>
1435 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1436 Define to speculation_safe_value_not_needed.
1439 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1440 Define to speculation_safe_value_not_needed.
1443 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1444 Define to speculation_safe_value_not_needed.
1447 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1448 Define to speculation_safe_value_not_needed.
1451 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1452 Define to speculation_safe_value_not_needed.
1455 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1456 Define to speculation_safe_value_not_needed.
1459 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1460 Define to speculation_safe_value_not_needed.
1463 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1464 Define to speculation_safe_value_not_needed.
1467 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1468 speculation_safe_value_not_needed.
1471 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
1472 to speculation_safe_value_not_needed.
1474 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1476 * doc/md.texi: Expand the documentation of instruction names
1477 to mention port-local uses. Document '@' in pattern names.
1478 * read-md.h (overloaded_instance, overloaded_name): New structs.
1480 (md_reader::handle_overloaded_name): New member function.
1481 (md_reader::get_overloads): Likewise.
1482 (md_reader::m_first_overload): New member variable.
1483 (md_reader::m_next_overload_ptr): Likewise.
1484 (md_reader::m_overloads_htab): Likewise.
1485 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
1486 m_next_overload_ptr and m_overloads_htab.
1487 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
1488 (get_mode_token, get_code_token, get_int_token): New functions.
1489 (map_attr_string): Add an optional argument that passes back
1490 the associated iterator.
1491 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
1492 (md_reader::handle_overloaded_name, add_overload_instance): New
1494 (apply_iterators): Handle '@' names. Report an error if '@'
1495 is used without iterators.
1496 (initialize_iterators): Initialize the new iterator_group fields.
1497 * genopinit.c (handle_overloaded_code_for)
1498 (handle_overloaded_gen): New functions.
1499 (main): Use them to print declarations of maybe_code_for_* and
1500 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
1501 * genemit.c (print_overload_arguments, print_overload_test)
1502 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
1503 (main): Use it to print definitions of maybe_code_for_* and
1504 maybe_gen_* functions.
1505 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
1506 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
1507 instead of explicit mode checks.
1508 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
1509 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
1510 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
1511 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
1512 (aarch64_expand_compare_and_swap): Likewise
1513 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
1514 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
1515 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
1516 (aarch64_constant_pool_reload_icode): Delete.
1517 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
1518 instead of aarch64_constant_pool_reload_icode. Use
1519 code_for_aarch64_reload_mov instead of explicit mode checks.
1520 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
1521 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
1522 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
1523 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
1524 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
1525 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
1526 (aarch64_atomic_load_op_code): Delete.
1527 (aarch64_emit_atomic_load_op): Likewise.
1528 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
1529 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
1530 instead of aarch64_emit_atomic_load_op.
1531 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
1532 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
1533 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
1534 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
1535 character before the pattern name.
1536 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
1537 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
1538 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
1539 (aarch64_frecps<mode>): Likewise.
1540 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
1541 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
1542 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
1543 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
1544 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
1546 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1548 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
1549 Allow HFmode constants if TARGET_FP_F16INST.
1551 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
1554 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
1555 No longer check last store for clobber of address register.
1557 2018-08-02 Martin Liska <mliska@suse.cz>
1559 PR gcov-profile/86817
1560 * gcov.c (process_all_functions): New function.
1562 (process_file): Move functions processing to
1563 process_all_functions.
1565 2018-08-02 David Malcolm <dmalcolm@redhat.com>
1567 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
1568 "const" to the "gimple *" and "rtx_insn *" parameters.
1569 * dumpfile.h (dump_user_location_t::dump_user_location_t):
1571 (dump_location_t::dump_location_t): Likewise.
1573 2018-08-01 Martin Sebor <msebor@redhat.com>
1575 PR tree-optimization/86650
1576 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
1577 rather than a "gcall *". Directly pass the data of interest
1578 to percent_K_format, rather than building a temporary CALL_EXPR
1580 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
1581 (gimple_fold_builtin_strncat): Adjust.
1582 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
1583 gcall* argument with gimple*.
1584 * gimple-ssa-warn-restrict.c (check_call): Same.
1585 (wrestrict_dom_walker::before_dom_children): Same.
1586 (builtin_access::builtin_access): Same.
1587 (check_bounds_or_overlap): Same
1588 (maybe_diag_overlap): Same.
1589 (maybe_diag_offset_bounds): Same.
1590 * tree-diagnostic.c (default_tree_printer): Move usage of
1591 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
1593 * tree-pretty-print.c (percent_K_format): Add argument.
1594 * tree-pretty-print.h: Add argument.
1595 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
1596 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
1597 (maybe_diag_stxncpy_trunc): Same.
1598 (handle_builtin_stxncpy): Same.
1599 (handle_builtin_strcat): Same.
1601 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1603 * match.pd: Optimise pointer range checks.
1605 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1607 PR tree-optimization/86758
1608 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
1609 to remove pattern statements.
1611 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1613 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
1614 result of dfs_enumerate_from when constructing stmt_vec_infos,
1615 instead of additionally calling get_loop_body.
1617 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1619 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
1621 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
1622 When creating an iv, assert that the step is not known to be zero.
1623 (vect_setup_realignment): Update call accordingly.
1624 * tree-vect-stmts.c (vectorizable_store): Likewise.
1625 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
1627 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1629 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
1630 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
1631 (vectorizable_reduction): Likewise.
1632 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1633 (vect_detect_hybrid_slp_stmts): Likewise.
1634 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
1636 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
1638 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
1639 (wide_int_set_zero_nonzero_bits): ...this.
1640 (zero_nonzero_bits_from_vr): Rename to...
1641 (vrp_set_zero_nonzero_bits): ...this.
1642 (extract_range_from_multiplicative_op_1): Abstract wide int
1644 (wide_int_range_multiplicative_op): ...here.
1645 (extract_range_from_binary_expr_1): Extract wide int binary
1646 operations into their own functions.
1647 (wide_int_range_lshift): New.
1648 (wide_int_range_can_optimize_bit_op): New.
1649 (wide_int_range_shift_undefined_p): New.
1650 (wide_int_range_bit_xor): New.
1651 (wide_int_range_bit_ior): New.
1652 (wide_int_range_bit_and): New.
1653 (wide_int_range_trunc_mod): New.
1654 (extract_range_into_wide_ints): New.
1655 (vrp_shift_undefined_p): New.
1656 (extract_range_from_multiplicative_op): New.
1657 (vrp_can_optimize_bit_op): New.
1658 * tree-vrp.h (value_range::dump): New.
1659 (wide_int_range_multiplicative_op): New.
1660 (wide_int_range_lshift):New.
1661 (wide_int_range_shift_undefined_p): New.
1662 (wide_int_range_bit_xor): New.
1663 (wide_int_range_bit_ior): New.
1664 (wide_int_range_bit_and): New.
1665 (wide_int_range_trunc_mod): New.
1666 (zero_nonzero_bits_from_bounds): Rename to...
1667 (wide_int_set_zero_nonzero_bits): ...this.
1668 (zero_nonzero_bits_from_vr): Rename to...
1669 (vrp_set_zero_nonzero_bits): ...this.
1670 (range_easy_mask_min_max): Rename to...
1671 (wide_int_range_can_optimize_bit_op): this.
1672 * vr-values.c (simplify_bit_ops_using_ranges): Rename
1673 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
1675 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1677 * tree-vectorizer.h (vect_orig_stmt): New function.
1678 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
1679 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
1680 (vect_create_epilog_for_reduction): Likewise.
1681 (vectorizable_live_operation): Likewise.
1682 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1683 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
1684 * tree-vect-stmts.c (vectorizable_call): Likewise.
1685 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
1687 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1689 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
1691 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
1692 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
1693 (vect_transform_loop): Likewise.
1694 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1696 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1698 * tree-vectorizer.h (vect_schedule_slp): Return void.
1699 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1700 (vect_schedule_slp): Likewise.
1702 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1704 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
1706 (vect_transform_loop): Update calls accordingly. Schedule SLP
1707 instances before the main loop, if any exist.
1709 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1711 PR tree-optimization/86749
1712 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
1713 If the lhs is used in a COND_EXPR, check that it is being used
1714 as the "then" or "else" value.
1716 2018-08-01 Tom de Vries <tdevries@suse.de>
1719 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1720 speculation_safe_value_not_needed.
1722 2018-08-01 Richard Biener <rguenther@suse.de>
1724 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
1727 2018-08-01 Martin Liska <mliska@suse.cz>
1729 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
1730 format how successful transformation is dumped.
1731 (gimple_mod_pow2_value_transform): Likewise.
1732 (gimple_mod_subtract_transform): Likewise.
1733 (gimple_stringops_transform): Likewise.
1735 2018-08-01 Martin Liska <mliska@suse.cz>
1738 * value-prof.c (interesting_stringop_to_profile_p):
1739 Simplify the code and add BUILT_IN_MEMMOVE.
1740 (gimple_stringops_transform): Likewise.
1742 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
1744 * config/aarch64/aarch64-simd.md
1745 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
1746 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
1747 use GPI iterator instead of SI mode.
1749 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1751 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
1752 rs6000_speculation_barrier.
1753 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
1754 new barrier pattern name.
1756 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1758 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
1759 (speculation_barrier): New insn.
1761 2018-08-01 Richard Biener <rguenther@suse.de>
1764 * graphite.h: Include isl/id.h and isl/space.h to allow build
1767 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
1770 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1771 mode for COFF targets.
1772 * defaults.h (TARGET_COFF): Define.
1773 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
1774 TARGET_COFF): Define.
1775 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
1776 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
1777 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
1779 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1781 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
1782 * gimple-streamer-out.c (output_bb): Save it.
1783 * lto-streamer-in.c (input_struct_function_base): Restore
1784 instance discriminator if available. Create map on demand.
1785 * lto-streamer-out.c (output_struct_function_base): Save it if
1787 * final.c (decl_to_instance_map): Document LTO strategy.
1789 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1790 Olivier Hainque <hainque@adacore.com>
1792 * debug.h (decl_to_instance_map_t): New type.
1793 (decl_to_instance_map): Declare.
1794 (maybe_create_decl_to_instance_map): New inline function.
1795 * final.c (bb_discriminator, last_bb_discriminator): New statics,
1796 to track basic block discriminators.
1797 (final_start_function_1): Initialize them.
1798 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
1800 (decl_to_instance_map): New variable.
1801 (map_decl_to_instance, maybe_set_discriminator): New functions.
1802 (notice_source_line): Set discriminator.
1804 2018-07-31 Ian Lance Taylor <iant@golang.org>
1806 * targhooks.c (default_have_speculation_safe_value): Add
1809 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1811 * dump-context.h: Include "pretty-print.h".
1812 (dump_context::refresh_dumps_are_enabled): New decl.
1813 (dump_context::emit_item): New decl.
1814 (class dump_context): Add fields "m_test_pp" and
1816 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1817 (temp_dump_context::get_dumped_text): New decl.
1818 (class temp_dump_context): Add field "m_pp".
1819 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
1820 (dump_context::refresh_dumps_are_enabled): ...and add a test for
1822 (set_dump_file): Update for above change.
1823 (set_alt_dump_file): Likewise.
1824 (dump_loc): New overload, taking a pretty_printer *.
1825 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
1826 to any test pretty-printer.
1827 (make_item_for_dump_gimple_stmt): New function, adapted from
1828 optinfo::add_gimple_stmt.
1829 (dump_context::dump_gimple_stmt): Call it, and use the result,
1830 eliminating the direct usage of dump_file and alt_dump_file in
1831 favor of indirectly using them via emit_item.
1832 (make_item_for_dump_gimple_expr): New function, adapted from
1833 optinfo::add_gimple_expr.
1834 (dump_context::dump_gimple_expr): Call it, and use the result,
1835 eliminating the direct usage of dump_file and alt_dump_file in
1836 favor of indirectly using them via emit_item.
1837 (make_item_for_dump_generic_expr): New function, adapted from
1839 (dump_context::dump_generic_expr): Call it, and use the result,
1840 eliminating the direct usage of dump_file and alt_dump_file in
1841 favor of indirectly using them via emit_item.
1842 (make_item_for_dump_printf_va): New function, adapted from
1843 optinfo::add_printf_va.
1844 (make_item_for_dump_printf): New function.
1845 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
1846 and use the result, eliminating the direct usage of dump_file and
1847 alt_dump_file in favor of indirectly using them via emit_item.
1848 (make_item_for_dump_dec): New function.
1849 (dump_context::dump_dec): Call it, and use the result,
1850 eliminating the direct usage of dump_file and alt_dump_file in
1851 favor of indirectly using them via emit_item.
1852 (make_item_for_dump_symtab_node): New function, adapted from
1853 optinfo::add_symtab_node.
1854 (dump_context::dump_symtab_node): Call it, and use the result,
1855 eliminating the direct usage of dump_file and alt_dump_file in
1856 favor of indirectly using them via emit_item.
1857 (dump_context::begin_scope): Reimplement, avoiding direct usage
1858 of dump_file and alt_dump_file in favor of indirectly using them
1860 (dump_context::emit_item): New member function.
1861 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1862 Set up test pretty-printer on the underlying context. Call
1863 refresh_dumps_are_enabled.
1864 (temp_dump_context::~temp_dump_context): Call
1865 refresh_dumps_are_enabled.
1866 (temp_dump_context::get_dumped_text): New member function.
1867 (selftest::verify_dumped_text): New function.
1868 (ASSERT_DUMPED_TEXT_EQ): New macro.
1869 (selftest::test_capture_of_dump_calls): Run all tests twice, with
1870 and then without optinfo enabled. Add uses of
1871 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
1872 * dumpfile.h: Update comment for the dump_* API.
1873 * optinfo-emit-json.cc
1874 (selftest::test_building_json_from_dump_calls): Update for new
1875 param for temp_dump_context ctor.
1876 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
1877 and "m_owned" field.
1878 (optinfo_item::~optinfo_item): Likewise.
1879 (optinfo::add_item): New member function.
1880 (optinfo::emit): Update comment.
1881 (optinfo::add_string): Delete.
1882 (optinfo::add_printf): Delete.
1883 (optinfo::add_printf_va): Delete.
1884 (optinfo::add_gimple_stmt): Delete.
1885 (optinfo::add_gimple_expr): Delete.
1886 (optinfo::add_tree): Delete.
1887 (optinfo::add_symtab_node): Delete.
1888 (optinfo::add_dec): Delete.
1889 * optinfo.h (class dump_context): New forward decl.
1890 (optinfo::add_item): New decl.
1891 (optinfo::add_string): Delete.
1892 (optinfo::add_printf): Delete.
1893 (optinfo::add_printf_va): Delete.
1894 (optinfo::add_gimple_stmt): Delete.
1895 (optinfo::add_gimple_expr): Delete.
1896 (optinfo::add_tree): Delete.
1897 (optinfo::add_symtab_node): Delete.
1898 (optinfo::add_dec): Delete.
1899 (optinfo::add_poly_int): Delete.
1900 (optinfo_item::optinfo_item): Remove "owned" param.
1901 (class optinfo_item): Remove field "m_owned".
1903 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1906 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
1907 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
1908 requested variable alignment.
1909 (expand_one_ssa_partition): Likewise.
1910 (expand_one_var): Likewise.
1912 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1914 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1915 to speculation_safe_value_not_needed.
1917 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1919 * targhooks.h (speculation_safe_value_not_needed): New prototype.
1920 * targhooks.c (speculation_safe_value_not_needed): New function.
1921 * target.def (have_speculation_safe_value): Update documentation.
1922 * doc/tm.texi: Regenerated.
1924 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1926 * config/aarch64/iterators.md (ALLI_TI): New iterator.
1927 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
1929 (despeculate_copy<ALLI:mode>_insn): New insn.
1930 (despeculate_copyti_insn): New insn.
1931 (despeculate_simple<ALLI:mode>): New insn
1932 (despeculate_simpleti): New insn.
1933 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
1935 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
1936 aarch64_speculation_safe_value.
1937 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
1939 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1941 * config/aarch64/aarch64-speculation.cc: New file.
1942 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
1943 before pass_reorder_blocks.
1944 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
1946 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
1947 X14 and X15 when tracking speculation.
1948 * config/aarch64/aarch64.md (register name constants): Add
1949 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
1950 (unspec): Add UNSPEC_SPECULATION_TRACKER.
1951 (speculation_barrier): New insn attribute.
1952 (cmp<mode>): Allow SP in comparisons.
1953 (speculation_tracker): New insn.
1954 (speculation_barrier): Add speculation_barrier attribute.
1955 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
1956 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
1957 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
1959 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1961 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
1962 aarch64_track_speculation is true.
1963 (tb<optab><mode>1): Likewise.
1964 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
1965 generate CB[N]Z when tracking speculation.
1966 (aarch64_split_compare_and_swap): Likewise.
1967 (aarch64_split_atomic_op): Likewise.
1969 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1971 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
1973 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1975 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
1976 (speculation_barrier): New insn.
1978 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1980 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
1981 * config/arm/arm.md (speculation_barrier): New expand.
1982 (speculation_barrier_insn): New pattern.
1984 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1986 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
1987 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
1988 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
1989 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
1991 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
1992 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
1993 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
1994 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
1995 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
1996 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
1997 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
1998 * builtins.c (expand_speculation_safe_value): New function.
1999 (expand_builtin): Call it.
2000 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
2001 * doc/extend.texi: Document __builtin_speculation_safe_value.
2002 * doc/md.texi: Document "speculation_barrier" pattern.
2003 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
2004 TARGET_HAVE_SPECULATION_SAFE_VALUE.
2005 * doc/tm.texi: Regenerated.
2006 * target.def (have_speculation_safe_value, speculation_safe_value): New
2008 * targhooks.c (default_have_speculation_safe_value): New function.
2009 (default_speculation_safe_value): New function.
2010 * targhooks.h (default_have_speculation_safe_value): Add prototype.
2011 (default_speculation_safe_value): Add prototype.
2013 2018-07-31 David Malcolm <dmalcolm@redhat.com>
2015 * dump-context.h (dump_context::dump_loc): New decl.
2016 * dumpfile.c (dump_context::dump_loc): New member function.
2017 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
2018 and dump_gimple_stmt.
2019 (dump_context::dump_gimple_expr_loc): Likewise, using
2021 (dump_context::dump_generic_expr_loc): Likewise, using
2023 (dump_context::dump_printf_loc_va): Likewise, using
2025 (dump_context::begin_scope): Explicitly using the global function
2026 "dump_loc", rather than the member function.
2028 2018-07-31 Martin Sebor <msebor@redhat.com>
2030 PR tree-optimization/86741
2031 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
2033 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
2035 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
2036 depend on whether prefetch instructions will be emitted or not.
2037 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
2038 will be emitted or not.
2039 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
2040 (TARGET_SETMEM_PFD): New macros.
2042 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2044 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
2045 (NULL_STMT_VEC_INFO): Delete.
2046 (stmt_vec_info::operator*): Likewise.
2047 (stmt_vec_info::operator gimple *): Likewise.
2048 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
2049 of NULL_STMT_VEC_INFO.
2050 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2051 (vect_reassociating_reduction_p): Likewise.
2052 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2053 (vectorizable_store): Likewise.
2054 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
2055 (vec_info::free_stmt_vec_infos): Likewise.
2057 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2059 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
2060 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
2062 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2064 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
2065 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
2066 (vec_info::free_stmt_vec_info): New private member functions.
2067 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
2068 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
2069 * tree-parloops.c (gather_scalar_reductions): Remove calls to
2070 set_stmt_vec_info_vec and free_stmt_vec_infos.
2071 * tree-vect-loop.c (_loop_vec_info): Remove call to
2072 set_stmt_vec_info_vec.
2073 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
2074 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
2075 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
2076 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
2077 (vec_info::free_stmt_vec_info): ...these new functions. Remove
2078 assignments in {vec_info::,}new_stmt_vec_info that are redundant
2079 with the clearing in the xcalloc.
2080 (stmt_vec_info_vec): Delete.
2081 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
2082 (vectorize_loops): Likewise.
2083 (vec_info::~vec_info): Remove argument from call to
2084 free_stmt_vec_infos.
2085 (vec_info::add_stmt): Remove vinfo argument from call to
2088 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2090 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
2091 rather than a gimple stmt.
2092 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
2093 information for pattern statements when passed the original
2094 statement; instead wait to be passed the pattern statement itself.
2095 Don't call set_vinfo_for_stmt here.
2096 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
2097 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
2098 stmt_vec_infos here.
2099 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
2100 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
2101 stmt_vec_infos entry.
2103 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2105 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
2106 * tree-vectorizer.c (vec_info::replace_stmt): New function.
2107 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
2108 * tree-vect-stmts.c (vectorizable_call): Likewise.
2109 (vectorizable_simd_clone_call): Likewise.
2111 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2113 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
2114 * tree-vectorizer.c (vec_info::remove_stmt): New function.
2115 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
2116 * tree-vect-loop.c (vect_transform_loop): Likewise.
2117 * tree-vect-slp.c (vect_schedule_slp): Likewise.
2118 * tree-vect-stmts.c (vect_remove_stores): Likewise.
2120 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2122 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
2123 (vect_dr_stmt): Delete.
2124 * tree-vectorizer.c (vec_info::lookup_dr): New function.
2125 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
2127 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
2128 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
2129 (vect_verify_datarefs_alignment, vect_peeling_supportable)
2130 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
2131 (vect_analyze_data_refs): Likewise.
2132 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
2134 (vect_find_same_alignment_drs): Likewise.
2135 (vect_slp_analyze_node_dependences): Update calls accordingly.
2136 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
2137 instead of DR_VECT_AUX.
2138 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
2139 of a vector data references. Use vec_info::lookup_dr instead of
2141 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
2142 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
2143 instead of DR_VECT_AUX.
2145 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2147 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
2149 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
2151 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2152 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2153 (vect_gen_prolog_loop_niters): Likewise.
2155 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2157 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
2158 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
2159 (vect_known_alignment_in_bytes, vect_dr_behavior)
2160 (vect_get_scalar_dr_size): Take references as dr_vec_infos
2161 instead of data_references. Update calls to other routines for
2162 which the same change has been made.
2163 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
2164 dr_vec_infos instead of stmt_vec_infos.
2165 (vect_analyze_data_ref_dependence): Update call accordingly.
2166 (vect_slp_analyze_data_ref_dependence)
2167 (vect_record_base_alignments): Use DR_VECT_AUX.
2168 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
2169 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
2170 (vector_alignment_reachable_p, vect_get_data_access_cost)
2171 (vect_peeling_supportable, vect_analyze_group_access_1)
2172 (vect_analyze_group_access, vect_analyze_data_ref_access)
2173 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
2174 (vect_compile_time_alias, vect_small_gap_p)
2175 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
2176 (vect_supportable_dr_alignment): Take references as dr_vec_infos
2177 instead of data_references. Update calls to other routines for
2178 which the same change has been made.
2179 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
2180 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
2181 (vect_slp_analyze_and_verify_node_alignment)
2182 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
2183 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
2184 (vect_setup_realignment): Use dr_vec_infos. Update calls after
2186 (_vect_peel_info::dr): Replace with...
2187 (_vect_peel_info::dr_info): ...this new field.
2188 (vect_peeling_hash_get_most_frequent)
2189 (vect_peeling_hash_choose_best_peeling): Update accordingly.
2190 (vect_peeling_hash_get_lowest_cost):
2191 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
2192 routines for which the same change has been made.
2193 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
2195 * tree-vect-loop-manip.c (get_misalign_in_elems)
2196 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
2198 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2199 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
2200 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
2201 (get_group_load_store_type, get_negative_load_store_type)
2202 (vect_get_data_ptr_increment, vectorizable_store)
2203 (vectorizable_load): Likewise.
2204 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
2205 Update calls to other routines for which the same change has been made.
2207 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2209 * tree-vectorizer.h (vec_info::move_dr): New member function.
2210 (dataref_aux): Rename to...
2211 (dr_vec_info): ...this and add "dr" and "stmt" fields.
2212 (_stmt_vec_info::dr_aux): Update accordingly.
2213 (_stmt_vec_info::data_ref_info): Delete.
2214 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
2215 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
2216 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
2217 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
2219 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
2220 (STMT_VINFO_DR_INFO): New macro.
2221 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
2222 (set_dr_misalignment): Update after rename of dataref_aux.
2223 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
2224 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
2225 initialization of STMT_VINFO_DATA_REF.
2226 * tree-vectorizer.c (vec_info::move_dr): New function.
2227 * tree-vect-patterns.c (vect_recog_bool_pattern)
2228 (vect_recog_mask_conversion_pattern)
2229 (vect_recog_gather_scatter_pattern): Use it.
2230 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
2231 the "dr" and "stmt" fields of dr_vec_info instead of
2232 STMT_VINFO_DATA_REF.
2234 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2236 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
2237 (is_pattern_stmt_p): Use it.
2238 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
2239 on pattern statements.
2241 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2243 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
2244 original stmt as a stmt_vec_info rather than a gimple stmt.
2245 (vect_pattern_recog_1): Take the statement directly as a
2246 stmt_vec_info, rather than via a gimple_stmt_iterator.
2247 Update call to vect_mark_pattern_stmts.
2248 (vect_pattern_recog): Update calls accordingly.
2250 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2252 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
2253 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
2254 a vect_def_type for the first argument.
2255 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
2256 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
2257 operand if it isn't defined by a vectorized statement.
2258 (vect_build_gather_load_calls): Remove the mask_dt argument and
2259 update calls to vect_get_vec_def_for_stmt_copy.
2260 (vectorizable_bswap): Likewise the dt argument.
2261 (vectorizable_call): Update calls to vectorizable_bswap and
2262 vect_get_vec_def_for_stmt_copy.
2263 (vectorizable_simd_clone_call, vectorizable_assignment)
2264 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
2265 (vectorizable_comparison): Update calls to
2266 vect_get_vec_def_for_stmt_copy.
2267 (vectorizable_store): Likewise. Remove now-unnecessary calls to
2269 (vect_get_loop_based_defs): Remove dt argument and update call
2270 to vect_get_vec_def_for_stmt_copy.
2271 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
2272 and vect_get_vec_def_for_stmt_copy.
2273 (vectorizable_load): Update calls to vect_build_gather_load_calls
2274 and vect_get_vec_def_for_stmt_copy.
2275 * tree-vect-loop.c (vect_create_epilog_for_reduction)
2276 (vectorizable_reduction, vectorizable_live_operation): Update calls
2277 to vect_get_vec_def_for_stmt_copy.
2279 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2281 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
2282 and gimple stmt arguments with a stmt_vec_info.
2283 (vect_record_base_alignments): Update calls accordingly.
2284 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
2285 and gimple stmt arguments with a stmt_vec_info.
2286 (vect_build_slp_tree_1): Remove vinfo argument and update call
2287 to vect_record_max_nunits.
2288 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
2289 and vect_record_max_nunits.
2291 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2293 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
2294 file and take a stmt_vec_info instead of a gimple stmt.
2295 (supportable_widening_operation, vect_finish_replace_stmt)
2296 (vect_finish_stmt_generation, vect_get_store_rhs)
2297 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
2298 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
2299 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
2300 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
2301 (vect_create_data_ref_ptr, bump_vector_ptr)
2302 (vect_permute_store_chain, vect_setup_realignment)
2303 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
2304 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
2305 (vectorizable_reduction, vectorizable_induction)
2306 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
2307 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
2308 than gimple stmts as arguments.
2309 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
2310 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
2311 (can_group_stmts_p, vect_check_gather_scatter)
2312 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
2313 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
2314 (vect_permute_load_chain, vect_shift_permute_load_chain)
2315 (vect_transform_grouped_load)
2316 (vect_record_grouped_load_vectors): Likewise.
2317 * tree-vect-loop.c (vect_fixup_reduc_chain)
2318 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
2319 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
2320 (vectorizable_reduction, vectorizable_induction)
2321 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
2322 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
2323 (vect_get_load_store_mask): Likewise.
2324 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
2325 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
2326 * tree-vect-stmts.c (vect_mark_relevant)
2327 (is_simple_and_all_uses_invariant)
2328 (exist_non_indexing_operands_for_use_p, process_use)
2329 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
2330 (vect_get_vec_def_for_operand, vect_get_vec_defs)
2331 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
2332 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
2333 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
2334 (get_negative_load_store_type, get_load_store_type)
2335 (vect_check_load_store_mask, vect_check_store_rhs)
2336 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
2337 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2338 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2339 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2340 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
2341 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
2342 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
2343 (supportable_widening_operation): Likewise.
2345 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2347 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
2348 a stmt_vec_info instead of a gcall.
2349 (vect_check_gather_scatter): Update call accordingly.
2350 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
2352 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
2353 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
2354 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
2355 instead of a gimple stmt.
2356 (vect_transform_loop): Update calls accordingly.
2357 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
2358 stmt_vec_infos instead of gimple stmts.
2359 (vect_analyze_slp_instance): Update use accordingly.
2360 * tree-vect-stmts.c (read_vector_array, write_vector_array)
2361 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
2362 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
2363 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
2364 (vect_gen_widened_results_half, vect_get_loop_based_defs)
2365 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
2366 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
2367 down to subroutines.
2369 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2371 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
2372 of the worklist from a vector of gimple stmts to a vector of
2374 * tree-vect-stmts.c (vect_mark_relevant, process_use)
2375 (vect_mark_stmts_to_be_vectorized): Likewise
2377 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2379 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
2380 statement before passing it to vect_analyze_stmt.
2381 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
2382 the chain of phi vector definitions. Track the exit phi via its
2384 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
2385 STMT_VINFO_REDUC_DEF.
2386 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
2387 stmt_vec_infos to handle the statement chains.
2388 (vect_get_slp_defs): Record the first statement in the node
2389 using a stmt_vec_info.
2390 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
2391 statements here and pass their stmt_vec_info down to subroutines.
2392 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
2393 down to vect_finish_stmt_generation.
2394 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
2395 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
2396 stmt_vec_infos to subroutines.
2397 (vect_remove_stores): Use stmt_vec_infos to handle the statement
2400 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2402 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
2403 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
2404 (vect_permute_store_chain, vect_setup_realignment)
2405 (vect_permute_load_chain, vect_shift_permute_load_chain)
2406 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
2407 stmts internally, and when passing values to other vectorizer routines.
2408 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2409 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
2410 (vect_analyze_loop_operations, get_initial_def_for_reduction)
2411 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
2412 (vectorizable_reduction, vectorizable_induction)
2413 (vectorizable_live_operation, vect_transform_loop_stmt)
2414 (vect_transform_loop): Likewise.
2415 * tree-vect-patterns.c (vect_reassociating_reduction_p)
2416 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
2417 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
2418 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2419 (vect_slp_analyze_node_operations_1): Likewise.
2420 * tree-vect-stmts.c (vect_mark_relevant, process_use)
2421 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
2422 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
2423 (vect_finish_stmt_generation_1, get_group_load_store_type)
2424 (get_load_store_type, vect_build_gather_load_calls)
2425 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2426 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2427 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2428 (vectorizable_store, vectorizable_load, vectorizable_condition)
2429 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
2430 (supportable_widening_operation): Likewise.
2431 (vect_get_vector_types_for_stmt): Likewise.
2432 * tree-vectorizer.h (vect_dr_behavior): Likewise.
2434 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2436 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2437 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
2438 (vect_permute_store_chain, vect_permute_load_chain)
2439 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
2440 repeated stmt_vec_info lookups.
2441 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2442 (vect_update_ivs_after_vectorizer): Likewise.
2443 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
2444 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2445 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
2446 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2447 (vect_bb_slp_scalar_cost): Likewise.
2448 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
2450 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2452 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
2453 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
2454 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
2455 to get gassigns and gcalls, rather than operating on generc gimple
2457 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
2458 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
2459 (vectorizable_load, vect_analyze_stmt): Likewise.
2460 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
2462 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2464 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
2465 return stmt_vec_infos rather than gimple stmts. Do not accept
2467 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
2469 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
2470 Update use of get_later_stmt.
2471 (vect_get_constant_vectors): Update call accordingly.
2472 (vect_schedule_slp_instance): Likewise
2473 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
2474 (vect_slp_analyze_instance_dependence): Likewise.
2475 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
2477 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2479 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
2480 (stmt_info_for_cost::stmt_info): ...this new field.
2481 (add_stmt_costs): Update accordingly.
2482 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2483 (vect_get_known_peeling_cost): Likewise.
2484 (vect_estimate_min_profitable_iters): Likewise.
2485 * tree-vect-stmts.c (record_stmt_cost): Likewise.
2487 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2489 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
2490 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2491 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
2493 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
2495 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2497 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
2498 a gimple stmt to a stmt_vec_info.
2499 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2501 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2503 * tree-vectorizer.h (vec_info::grouped_stores): Change from
2504 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2505 (_loop_vec_info::reduction_chains): Likewise.
2506 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
2508 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2510 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2512 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
2513 a gimple stmt to a stmt_vec_info.
2514 (_stmt_vec_info::next_element): Likewise.
2515 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
2516 (vect_slp_analyze_and_verify_node_alignment)
2517 (vect_analyze_group_access_1, vect_analyze_group_access)
2518 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
2519 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
2520 (vect_supportable_dr_alignment): Update accordingly.
2521 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
2522 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
2523 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
2524 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2525 * tree-vect-slp.c (vect_build_slp_tree_1)
2526 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
2527 (vect_split_slp_store_group, vect_analyze_slp_instance)
2528 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
2529 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
2530 (get_group_load_store_type, get_load_store_type)
2531 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
2532 (vect_transform_stmt, vect_remove_stores): Likewise.
2534 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2536 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
2538 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2539 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
2540 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
2541 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
2542 (vector_alignment_reachable_p, vect_get_data_access_cost)
2543 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
2544 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
2545 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
2546 (vect_analyze_group_access_1, vect_analyze_group_access)
2547 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
2548 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
2549 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
2550 result of vect_dr_stmt and use the stmt_vec_info instead of
2551 the associated gimple stmt.
2552 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2553 (vect_gen_prolog_loop_niters): Likewise.
2554 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2556 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2558 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
2559 to a vec<stmt_vec_info>.
2560 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
2561 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
2563 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
2564 to a vec<stmt_vec_info>.
2565 (bst_traits::value_type, bst_traits::value_type): Likewise.
2566 (bst_traits::hash): Update accordingly.
2567 (vect_get_and_check_slp_defs): Change the stmts parameter from
2568 a vec<gimple *> to a vec<stmt_vec_info>.
2569 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
2570 (vect_build_slp_tree): Likewise.
2571 (vect_build_slp_tree_2): Likewise. Update uses of
2572 SLP_TREE_SCALAR_STMTS.
2573 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
2574 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
2575 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
2576 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
2577 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
2578 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
2579 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
2580 (vect_get_constant_vectors, vect_get_slp_defs)
2581 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
2582 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
2583 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
2584 instead of gimple stmts.
2585 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
2586 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
2587 (vect_slp_analyze_instance_dependence): Update uses of
2588 SLP_TREE_SCALAR_STMTS.
2589 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2590 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
2591 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
2592 (get_initial_defs_for_reduction): Likewise.
2593 (vect_create_epilog_for_reduction): Likewise.
2594 (vectorize_fold_left_reduction): Likewise.
2595 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
2596 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
2597 (can_vectorize_live_stmts): Likewise.
2599 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2601 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
2602 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2603 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
2605 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
2607 (gather_scalar_reductions): Update after above interface changes.
2608 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
2609 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
2611 (vect_force_simple_reduction): Likewise.
2612 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
2613 LOOP_VINFO_REDUCTIONS.
2614 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2616 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2618 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
2619 a gimple stmt to a stmt_vec_info.
2620 * tree-vect-loop.c (vect_active_double_reduction_p)
2621 (vect_force_simple_reduction, vectorizable_reduction): Update
2624 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2626 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
2627 vec<gimple *> to a vec<stmt_vec_info>.
2628 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
2629 the reduction_phis argument from a vec<gimple *> to a
2631 (vectorizable_reduction): Likewise the phis local variable that
2632 is passed to vect_create_epilog_for_reduction. Update for new type
2633 of SLP_TREE_VEC_STMTS.
2634 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
2635 (vectorizable_live_operation): Likewise.
2636 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
2637 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
2639 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2641 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
2642 a gimple stmt to a stmt_vec_info.
2643 (vectorizable_condition, vectorizable_live_operation)
2644 (vectorizable_reduction, vectorizable_induction): Pass back the
2645 vectorized statement as a stmt_vec_info.
2646 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
2647 use of STMT_VINFO_VEC_STMT.
2648 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
2649 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
2650 as stmt_vec_infos rather than gimple stmts.
2651 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
2653 (vectorizable_live_operation): Likewise.
2654 (vectorizable_reduction, vectorizable_induction): Likewise,
2655 updating use of STMT_VINFO_VEC_STMT.
2656 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
2657 of STMT_VINFO_VEC_STMT.
2658 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
2659 (vectorizable_simd_clone_call, vectorizable_conversion)
2660 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2661 (vectorizable_store, vectorizable_load, vectorizable_condition)
2662 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
2663 from a gimple stmt to a stmt_vec_info.
2664 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
2665 pointer to a stmt_vec_info to the vectorizable_* routines.
2667 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2669 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
2670 a gimple stmt to a stmt_vec_info.
2671 (is_pattern_stmt_p): Update accordingly.
2672 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
2673 (vect_record_grouped_load_vectors): Likewise.
2674 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2675 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
2676 (vect_model_reduction_cost): Likewise.
2677 (vect_create_epilog_for_reduction): Likewise.
2678 (vectorizable_reduction, vectorizable_induction): Likewise.
2679 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2680 Return the stmt_vec_info for the pattern statement.
2681 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
2682 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
2683 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2684 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
2685 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2686 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
2687 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
2688 (free_stmt_vec_info, vect_is_simple_use): Likewise.
2690 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2692 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
2693 (vect_finish_stmt_generation): Likewise.
2694 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2695 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
2696 (vect_build_gather_load_calls): Use the return value of the above
2697 functions instead of a separate call to vinfo_for_stmt. Use narrow
2698 scopes for the input gimple stmt and wider scopes for the associated
2699 stmt_vec_info. Use vec_info::lookup_def when setting these
2700 stmt_vec_infos from an SSA_NAME definition.
2701 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2702 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2703 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2704 (vectorizable_store, vectorizable_load, vectorizable_condition)
2705 (vectorizable_comparison): Likewise.
2706 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
2707 (vectorizable_reduction): Likewise.
2709 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2711 * tree-vectorizer.h (vect_is_simple_use): Add an optional
2712 stmt_vec_info * parameter before the optional gimple **.
2713 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2714 (process_use, vect_get_vec_def_for_operand_1): Update callers.
2715 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
2716 * tree-vect-loop.c (vectorizable_reduction): Likewise.
2717 (vectorizable_live_operation): Likewise.
2718 * tree-vect-patterns.c (type_conversion_p): Likewise.
2719 (vect_look_through_possible_promotion): Likewise.
2720 (vect_recog_rotate_pattern): Likewise.
2721 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2723 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2725 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
2726 a typedef to a wrapper class.
2727 (NULL_STMT_VEC_INFO): New macro.
2728 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
2729 (stmt_vec_info::operator*): New function.
2730 (stmt_vec_info::operator gimple *): Likewise.
2731 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
2732 (add_stmt_costs): Likewise.
2733 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
2734 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2735 (vect_get_known_peeling_cost): Likewise.
2736 (vect_estimate_min_profitable_iters): Likewise.
2737 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2738 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
2739 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2740 (vectorizable_store, free_stmt_vec_infos): Likewise.
2741 (new_stmt_vec_info): Change return type of xcalloc to
2744 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2746 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
2747 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
2748 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
2749 a single_imm_use-based sequence.
2750 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
2752 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2754 * tree-vectorizer.h (vec_info::lookup_def): Declare.
2755 * tree-vectorizer.c (vec_info::lookup_def): New function.
2756 * tree-vect-patterns.c (vect_get_internal_def): Use it.
2757 (vect_widened_op_tree): Likewise.
2758 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2759 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
2760 (vectorizable_reduction): Likewise.
2761 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
2763 (vect_is_slp_reduction): Update calls accordingly. Use
2764 vec_info::lookup_def.
2765 (vect_is_simple_reduction): Likewise
2766 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
2768 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2770 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
2771 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
2772 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
2774 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
2775 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
2776 (vect_update_vf_for_slp, vect_analyze_loop_operations)
2777 (vect_is_slp_reduction, vectorizable_induction)
2778 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
2779 * tree-vect-patterns.c (vect_init_pattern_stmt):
2780 (vect_determine_min_output_precision_1, vect_determine_precisions)
2781 (vect_pattern_recog): Likewise.
2782 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
2783 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
2784 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
2785 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2786 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
2787 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
2788 info field from a loop to a loop_vec_info.
2790 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2792 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
2793 (vec_info::add_stmt): Declare.
2794 * tree-vectorizer.c (vec_info::add_stmt): New function.
2795 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
2796 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
2797 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2798 (vectorizable_induction): Likewise.
2799 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
2800 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2801 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
2802 (vectorizable_load): Likewise.
2803 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2804 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
2805 (vect_recog_gather_scatter_pattern): Likewise.
2806 (append_pattern_def_seq): Likewise. Remove a check that is
2807 performed by add_stmt itself.
2809 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2811 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
2812 which make_ssa_name was called with new_stmt before new_stmt
2815 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2817 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
2819 (vect_is_slp_reduction): ...here...
2820 (vect_is_simple_reduction): ...and here. Remove repetition of tests
2821 that are already known to be false.
2823 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2825 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
2826 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
2827 STMT_VINFO_NUM_SLP_USES when it's true.
2828 (vect_free_slp_instance): Add a final_p parameter and pass it to
2830 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
2831 (vect_analyze_slp_instance): Likewise.
2832 (vect_slp_analyze_operations): Likewise.
2833 (vect_slp_analyze_bb_1): Likewise.
2834 * tree-vectorizer.c (vec_info): Likewise.
2835 * tree-vect-loop.c (vect_transform_loop): Likewise.
2837 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2839 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2840 function is not called for second and subsequent members of
2843 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2845 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
2846 cases for nested loops from here to ...
2847 (vect_create_epilog_for_reduction): ...here. Only call
2848 vect_is_simple_use for inner-loop reductions.
2850 2018-07-31 Martin Liska <mliska@suse.cz>
2852 PR gcov-profile/85338
2853 PR gcov-profile/85350
2854 PR gcov-profile/85372
2855 * profile.c (struct location_triplet): New.
2856 (struct location_triplet_hash): Likewise.
2857 (output_location): Do not output a BB that
2858 is already recorded for a line.
2859 (branch_prob): Use streamed_locations.
2861 2018-07-31 Martin Liska <mliska@suse.cz>
2863 PR gcov-profile/85370
2864 * coverage.c (coverage_begin_function): Do not mark target
2865 clones as artificial functions.
2867 2018-07-31 Martin Liska <mliska@suse.cz>
2869 PR gcov-profile/83813
2870 PR gcov-profile/84758
2871 PR gcov-profile/85217
2872 PR gcov-profile/85332
2873 * profile.c (branch_prob): Do not record GOTO expressions
2874 for GIMPLE statements which locations are already streamed.
2876 2018-07-31 Olivier Hainque <hainque@adacore.com>
2878 * gcc.c (handle_spec_function): Accept a soft_matched_part
2879 argument, as do_spec_1. Pass it down to ...
2880 (eval_spec_function): Accept a soft_matched_part argument,
2881 and pass it down to ...
2882 (do_spec_2): Accept a soft_matched_part argument, and pass
2883 it down to do_spec_1.
2884 (do_spec_1): Pass soft_matched_part to handle_spec_function.
2885 (handle_braces): Update call to handle_spec_function.
2886 (driver::set_up_specs): Update calls to do_spec_2.
2887 (compare_debug_dump_opt_spec_function): Likewise.
2888 (compare_debug_self_opt_spec_function): Likewise.
2890 2018-07-31 Olivier Hainque <hainque@adacore.com>
2892 * common.opt (nolibc): New option.
2893 * doc/invoke.texi (Link Options): Document it.
2894 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
2895 * config/alpha/linux.h: Likewise.
2896 * config/arc/elf.h: Likewise.
2897 * config/arm/uclinux-elf.h: Likewise.
2898 * config/arm/unknown-elf.h: Likewise.
2899 * config/avr/avrlibc.h: Likewise.
2900 * config/bfin/bfin.h: Likewise.
2901 * config/bfin/linux.h: Likewise.
2902 * config/bfin/uclinux.h: Likewise.
2903 * config/darwin.h: Likewise.
2904 * config/darwin10.h: Likewise.
2905 * config/darwin12.h: Likewise.
2906 * config/gnu-user.h: Likewise.
2907 * config/lm32/uclinux-elf.h: Likewise.
2908 * config/pa/pa-hpux11.h: Likewise.
2909 * config/pa/pa64-hpux.h: Likewise.
2910 * config/sparc/sparc.h: Likewise.
2912 2018-07-31 Olivier Hainque <hainque@adacore.com>
2914 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
2915 undefined variables.
2917 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2920 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
2923 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2925 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2926 terminated string literal.
2928 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2930 PR rtl-optimization/85160
2931 * combine.c (is_just_move): New function.
2932 (try_combine): Allow combining two instructions into two if neither of
2933 the original instructions was a move.
2935 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
2938 * doc/extend.texi (Global Register Variables): Discourage use of type
2940 (Local Register Variables): Likewise.
2942 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
2944 PR tree-optimization/86506
2945 * hwint.h (ceil_log2): Resync with hwint.c implementation.
2947 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
2950 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
2951 hard_regno, make sure no insn between `from` and `to` clobbers it.
2953 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
2954 Tom de Vries <tdevries@suse.de>
2956 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
2957 (PTX_DEFAULT_RUNTIME_DIM): ... this.
2958 (nvptx_goacc_validate_dims): Set default worker and gang dims to
2959 PTX_DEFAULT_RUNTIME_DIM.
2960 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
2962 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
2964 * config/pa/pa.c (pa_output_addr_vec): Align address table.
2965 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
2966 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
2968 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
2970 * config/rs6000/constraints.md (wG constraint): Delete, no longer
2972 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
2973 predicate to reflect toc fusion has been deleted.
2974 (toc_fusion_mem_raw): Delete, no longer used.
2975 (toc_fusion_mem_wrapped): Likewise.
2976 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
2978 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
2979 Delete, no longer used.
2980 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
2981 meant to be used for toc fusion.
2982 (rs6000_debug_print_mode): Delete toc fusion debugging.
2983 (rs6000_debug_reg_global): Likewise.
2984 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
2985 fusion and secondary reload support that were never used.
2986 (rs6000_option_override_internal): Delete TOC fusion, that was only
2987 partially defined, and it did not work unless you also used the
2989 (rs6000_legitimate_address_p): Delete TOC fusion support.
2990 (rs6000_opt_masks): Likewise.
2991 (fusion_wrap_memory_address): Delete function, no longer used.
2992 (fusion_split_address); Delete TOC fusion support.
2993 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
2994 longer used with toc fusion being deleted.
2995 (TARGET_TOC_FUSION_FP): Likewise.
2996 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
2998 (toc fusion spliter): Delete TOC fusion support.
2999 (toc_fusionload_<mode>): Likewise.
3000 (toc_fusionload_di): Likewise.
3001 (fusion_gpr_load_<mode>): Delete generator function, this insn no
3002 longer needs to be named. Rename predicate to delete TOC fusion.
3003 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
3004 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
3005 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
3006 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
3007 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
3009 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
3011 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
3012 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
3013 __int128 in built-in function prototypes.
3014 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
3015 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
3017 2018-07-27 Martin Sebor <msebor@redhat.com>
3019 PR tree-optimization/86696
3020 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
3021 types, including enums.
3022 (handle_char_store): Be prepared for the above function to fail.
3024 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
3026 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
3027 when optimization level is lower than 2 or optimize for size.
3029 2018-07-26 Martin Sebor <msebor@redhat.com>
3031 PR tree-optimization/86043
3032 PR tree-optimization/86042
3033 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
3034 (get_string_cst_length): Rename...
3035 (get_min_string_length): ...to this. Add argument.
3036 (handle_char_store): Extend to handle multi-character stores by
3038 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
3039 * tree.h (initializer_zerop): Add argument.
3041 2018-07-26 Jakub Jelinek <jakub@redhat.com>
3044 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
3045 declare target to variables if they have always,{to,from,tofrom} map
3048 2018-07-26 Martin Liska <mliska@suse.cz>
3051 * lto-wrapper.c: Add linker_output as prefix
3052 for ltrans_output_file.
3054 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
3056 PR rtl-optimization/85805
3057 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
3058 value for hard registers if that was written in the same mode.
3060 2018-07-26 Martin Liska <mliska@suse.cz>
3062 PR gcov-profile/86536
3063 * gcov.c (format_gcov): Use printf format %.*f directly
3064 and do not handle special values.
3066 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
3068 * common/config/arc/arc-common.c (arc_option_optimization_table):
3069 Update default optimizations for size.
3071 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
3073 * config/arc/arc.md (movsf_insn): Add short instruction selection.
3074 * config/arc/constraints.md (CfZ): New constraint.
3075 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
3076 (subsf3_fpu): Likewise.
3077 (cmpsf_fpu): Likewise.
3078 (cmpsf_fpu_uneq): Likewise.
3080 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
3082 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
3084 (arc_is_uncached_mem_p): uncached applies to both the variable and
3087 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
3089 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
3092 2018-07-25 David Malcolm <dmalcolm@redhat.com>
3094 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
3095 field "m_scopes" from vec to auto_vec.
3097 2018-07-25 Martin Liska <mliska@suse.cz>
3099 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
3102 2018-07-25 Richard Biener <rguenther@suse.de>
3105 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
3106 special wrt context_die late.
3107 (gen_subprogram_die): Re-use DIEs in local scope.
3109 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
3111 PR tree-optimization/86644
3112 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
3114 2018-07-25 Martin Liska <mliska@suse.cz>
3117 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
3118 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
3120 2018-07-25 Martin Liska <mliska@suse.cz>
3123 * params.def: Explain ASan abbreviation and provide
3124 a documentation link.
3126 2018-07-24 Martin Sebor <msebor@redhat.com>
3128 PR tree-optimization/86622
3129 PR tree-optimization/86532
3130 * builtins.h (string_length): Declare.
3131 * builtins.c (c_strlen): Correct handling of non-constant offsets.
3132 (check_access): Be prepared for non-constant length ranges.
3133 (string_length): Make extern.
3134 * expr.c (string_constant): Only handle the minor non-constant
3135 array index. Use string_constant to compute the length of
3136 a generic string constant.
3138 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
3140 PR tree-optimization/86618
3141 * tree-vect-stmts.c (vectorizable_call): Don't take the address
3142 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
3144 2018-07-24 David Malcolm <dmalcolm@redhat.com>
3146 PR tree-optimization/86636
3147 * json.cc (json::object::set): Fix comment. Add assertions.
3148 (json::array::append): Move here from json.h. Add comment and an
3150 (json::string::string): Likewise.
3151 * json.h (json::array::append): Move to json.cc.
3152 (json::string::string): Likewise.
3153 * optinfo-emit-json.cc
3154 (optrecord_json_writer::impl_location_to_json): Assert that we
3155 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
3156 wrapper around it. Expand the location once, rather than three
3158 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
3159 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
3161 (optrecord_json_writer::optinfo_to_json): Likewise, in four
3162 places. Fix some overlong lines.
3164 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
3166 * config/aarch64/aarch64-simd.md
3167 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
3168 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
3169 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
3170 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
3171 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
3172 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
3173 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
3174 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
3175 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
3177 2018-07-24 Jakub Jelinek <jakub@redhat.com>
3180 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
3181 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
3182 and abs_d == d, do the power of two handling if profitable.
3184 2018-07-24 Richard Biener <rguenther@suse.de>
3186 * match.pd: Add BIT_FIELD_REF canonicalizations.
3188 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3191 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
3193 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3195 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
3196 terminated STRING_CST object.
3198 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3200 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
3202 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
3204 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
3205 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
3206 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
3207 the elements into a register.
3208 (rs6000_split_v4si_init_di_reg): Delete.
3209 (rs6000_split_v4si_init): Delete.
3210 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
3211 (vsx_init_v4si): Rewrite as a define_expand.
3213 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
3215 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
3216 zero_extend argument from memory): New.
3218 2018-07-22 Martin Sebor <msebor@redhat.com>
3221 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
3222 diagnosing calls with unknown arguments unless -Walloca-larger-than
3223 is restricted to less than PTRDIFF_MAX bytes.
3225 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
3227 * doc/gcov.texi (Invoking Gcov): Editorial changes.
3229 2018-07-20 David Malcolm <dmalcolm@redhat.com>
3231 * pretty-print.c (text_info::set_location): Remove redundant
3232 "line_table" parameter from call to rich_location::set_range.
3234 2018-07-20 Martin Sebor <msebor@redhat.com>
3237 * builtins.c (expand_builtin_alloca): Adjust.
3238 * calls.c (alloc_max_size): Simplify.
3239 * cgraphunit.c (cgraph_node::expand): Adjust.
3240 * common.opt (larger_than_size, warn_frame_larger_than): Remove
3242 (frame_larger_than_size): Same.
3243 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
3244 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
3246 * doc/invoke.texi (GCC Command Options): Document option arguments.
3247 Explain byte-size arguments and suffixes.
3248 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
3249 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
3250 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
3251 * doc/options.texi (UInteger): Expand.
3252 (Host_Wide_Int, ByteSize): Document new properties.
3253 * final.c (final_start_function_1): Include sizes in an error message.
3254 * function.c (frame_offset_overflow): Same.
3255 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
3256 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
3257 Diagnose unbounded alloca calls only for limits of less than
3259 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
3260 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
3262 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
3263 only for limits of less than PTRDIFF_MAX.
3264 * langhooks-def.h (lhd_handle_option): Change function argument
3266 * langhooks.c (lhd_handle_option): Same.
3267 * langhooks.h (handle_option): Same.
3268 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
3270 (var_type, var_type_struct): Same.
3271 (var_set): Handle ByteSize flag.
3272 * optc-gen.awk: Add comments to output to ease debugging. Make
3273 use of HOST_WIDE_INT where appropriate.
3274 * opts-gen-save.awk: Use %lx to format unsigned long.
3275 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
3276 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
3277 arguments. Parse bytes-size suffixes.
3278 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
3279 (enum_value_to_arg): Same.
3280 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
3281 (handle_option): Adjust.
3282 (generate_option): Change function argument to HOST_WIDE_INT.
3283 (cmdline_handle_error): Adjust.
3284 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
3285 (set_option): Change function argument to HOST_WIDE_INT.
3286 (option_enabled): Handle cl_host_wide_int.
3287 (get_option_state): Handle CLVC_SIZE.
3288 (control_warning_option): Same.
3289 * opts.c (common_handle_option): Change function argument to
3290 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
3291 OPT_Wvla_larger_than_.
3292 * opts.h (enum cl_var_type): Add an enumerator.
3293 * stor-layout.c (layout_decl): Print a more meaningful warning.
3294 * toplev.c (output_stack_usage): Adjust.
3296 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
3298 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
3299 call to inline_expand_builtin_string_cmp.
3300 (expand_builtin_strcmp): Likewise.
3301 (expand_builtin_strncmp): Likewise.
3302 (inline_string_cmp): Delete the last parameter, change char_type_node
3303 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
3305 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
3306 the inlining expansion on target where the type of the call has same or
3307 narrower precision than unsigned char.
3309 2018-07-20 David Malcolm <dmalcolm@redhat.com>
3311 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
3312 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
3313 * common.opt (fsave-optimization-record): New option.
3314 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
3315 * doc/invoke.texi (-fsave-optimization-record): New option.
3316 * dumpfile.c: Include "optinfo-emit-json.h".
3317 (struct kv_pair): Move to coretypes.h.
3318 (optgroup_options): Make non-static.
3319 (dump_context::end_scope): Call
3320 optimization_records_maybe_pop_dump_scope.
3321 * dumpfile.h (optgroup_options): New decl.
3322 * json.cc: New file.
3324 * optinfo-emit-json.cc: New file.
3325 * optinfo-emit-json.h: New file.
3326 * optinfo.cc: Include "optinfo-emit-json.h".
3327 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
3328 (optinfo_enabled_p): Check optimization_records_enabled_p.
3329 (optinfo_wants_inlining_info_p): Likewise.
3330 * optinfo.h: Update comment.
3331 * profile-count.c (profile_quality_as_string): New function.
3332 * profile-count.h (profile_quality_as_string): New decl.
3333 (profile_count::quality): New accessor.
3334 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
3335 and optinfo_emit_json_cc_tests.
3336 * selftest.h (selftest::json_cc_tests): New decl.
3337 (selftest::optinfo_emit_json_cc_tests): New decl.
3338 * toplev.c: Include "optinfo-emit-json.h".
3339 (compile_file): Call optimization_records_finish.
3340 (do_compile): Call optimization_records_start.
3341 * tree-ssa-live.c: Include optinfo.h.
3342 (remove_unused_scope_block_p): Retain inlining information if
3343 optinfo_wants_inlining_info_p returns true.
3345 2018-07-20 Richard Biener <rguenther@suse.de>
3348 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
3349 to cover -flto-partition=none.
3351 2018-07-20 Martin Liska <mliska@suse.cz>
3353 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
3354 (get_decl_source_range): Remove unused function.
3356 2018-07-20 Richard Biener <rguenther@suse.de>
3358 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
3359 (struct vn_phi_s): Likewise.
3360 (struct vn_reference_s): Likewise.
3361 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
3362 for searching the slot of an entry known to be in the hash itself.
3363 (vn_phi_hasher::equal): Likewise.
3364 (vn_reference_hasher::equal): Likewise.
3365 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
3367 (optimistic_info, current_info): Remove, keeping only valid_info.
3368 (vn_reference_lookup_1): Remove fallback lookup.
3369 (vn_reference_lookup_2): Likewise.
3370 (vn_nary_op_lookup_1): Likewise.
3371 (vn_phi_lookup): Likewise.
3372 (vn_nary_build_or_lookup_1): Make sure to not chain the built
3374 (vn_reference_insert): Adjust, chain the inserted hash element
3375 at last_inserted_ref.
3376 (vn_reference_insert_pieces): Likewise.
3377 (visit_reference_op_call): Likewise.
3378 (vn_nary_op_insert_into): Chain the inserted hash element at
3380 (vn_nary_op_insert_pieces): Adjust.
3381 (vn_nary_op_insert): Likewise.
3382 (vn_nary_op_insert_stmt): Likewise.
3383 (vn_phi_insert): Adjust, chain the inserted hash element at
3385 (process_scc): Remove clearing and copying the optimistic
3386 table. Instead remove elements inserted during an optimistic
3387 iteration from the single table we maintain.
3388 (init_scc_vn): Adjust.
3389 (free_scc_vn): Likewise.
3390 (sccvn_dom_walker::record_cond): Likewise.
3391 (sccvn_dom_walker::after_dom_children): Likewise.
3393 2018-07-19 Martin Sebor <msebor@redhat.com>
3395 PR tree-optimization/84047
3396 PR tree-optimization/83776
3397 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
3398 (check_array_bounds): Call it.
3400 2018-07-19 Martin Sebor <msebor@redhat.com>
3402 * align.h (align_flags): Use member initialization.
3404 2018-07-19 David Malcolm <dmalcolm@redhat.com>
3406 * Makefile.in (OBJS): Add optinfo.o.
3407 * coretypes.h (class symtab_node): New forward decl.
3408 (struct cgraph_node): New forward decl.
3409 (class varpool_node): New forward decl.
3410 * dump-context.h: New file.
3411 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
3413 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
3414 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
3415 (set_alt_dump_file): Likewise.
3416 (dump_context::~dump_context): New dtor.
3417 (dump_gimple_stmt): Move implementation to...
3418 (dump_context::dump_gimple_stmt): ...this new member function.
3419 Add the stmt to any pending optinfo, creating one if need be.
3420 (dump_gimple_stmt_loc): Move implementation to...
3421 (dump_context::dump_gimple_stmt_loc): ...this new member function.
3422 Start a new optinfo and add the stmt to it.
3423 (dump_gimple_expr): Move implementation to...
3424 (dump_context::dump_gimple_expr): ...this new member function.
3425 Add the stmt to any pending optinfo, creating one if need be.
3426 (dump_gimple_expr_loc): Move implementation to...
3427 (dump_context::dump_gimple_expr_loc): ...this new member function.
3428 Start a new optinfo and add the stmt to it.
3429 (dump_generic_expr): Move implementation to...
3430 (dump_context::dump_generic_expr): ...this new member function.
3431 Add the tree to any pending optinfo, creating one if need be.
3432 (dump_generic_expr_loc): Move implementation to...
3433 (dump_context::dump_generic_expr_loc): ...this new member
3434 function. Add the tree to any pending optinfo, creating one if
3436 (dump_printf): Move implementation to...
3437 (dump_context::dump_printf_va): ...this new member function. Add
3438 the text to any pending optinfo, creating one if need be.
3439 (dump_printf_loc): Move implementation to...
3440 (dump_context::dump_printf_loc_va): ...this new member function.
3441 Start a new optinfo and add the stmt to it.
3442 (dump_dec): Move implementation to...
3443 (dump_context::dump_dec): ...this new member function. Add the
3444 value to any pending optinfo, creating one if need be.
3445 (dump_context::dump_symtab_node): New member function.
3446 (dump_context::get_scope_depth): New member function.
3447 (dump_context::begin_scope): New member function.
3448 (dump_context::end_scope): New member function.
3449 (dump_context::ensure_pending_optinfo): New member function.
3450 (dump_context::begin_next_optinfo): New member function.
3451 (dump_context::end_any_optinfo): New member function.
3452 (dump_context::s_current): New global.
3453 (dump_context::s_default): New global.
3454 (dump_scope_depth): Delete global.
3455 (dumpfile_ensure_any_optinfo_are_flushed): New function.
3456 (dump_symtab_node): New function.
3457 (get_dump_scope_depth): Reimplement in terms of dump_context.
3458 (dump_begin_scope): Likewise.
3459 (dump_end_scope): Likewise.
3460 (selftest::temp_dump_context::temp_dump_context): New ctor.
3461 (selftest::temp_dump_context::~temp_dump_context): New dtor.
3462 (selftest::verify_item): New function.
3463 (ASSERT_IS_TEXT): New macro.
3464 (ASSERT_IS_TREE): New macro.
3465 (ASSERT_IS_GIMPLE): New macro.
3466 (selftest::test_capture_of_dump_calls): New test.
3467 (selftest::dumpfile_c_tests): Call it.
3468 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
3469 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
3470 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
3471 descriptive comment.
3472 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
3473 (dump_node, dump_bb): Move these unrelated decls.
3474 (class dump_manager): Add leading comment.
3475 * optinfo.cc: New file.
3476 * optinfo.h: New file.
3478 2018-07-19 Michael Collison <michael.collison@arm.com>
3479 Richard Henderson <rth@redhat.com>
3481 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
3482 (subti): Handle op1 zero.
3483 (subvti4, usub4ti4): New.
3484 (*sub<GPI>3_compare1_imm): New.
3485 (sub<GPI>3_carryinCV): New.
3486 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
3487 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
3489 2018-07-19 Michael Collison <michael.collison@arm.com>
3490 Richard Henderson <rth@redhat.com>
3492 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
3493 (addti3): Create simpler code if low part is already known to be 0.
3494 (addvti4, uaddvti4): New.
3495 (*add<GPI>3_compareC_cconly_imm): New.
3496 (*add<GPI>3_compareC_cconly): New.
3497 (*add<GPI>3_compareC_imm): New.
3498 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
3499 handle constants within this pattern..
3500 (*add<GPI>3_compareV_cconly_imm): New.
3501 (*add<GPI>3_compareV_cconly): New.
3502 (*add<GPI>3_compareV_imm): New.
3503 (add<GPI>3_compareV): New.
3504 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
3505 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
3506 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
3507 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
3508 with 'comparison' operator.
3509 (*add<GPI>3_compareV_cconly_imm): Ditto.
3510 (*add<GPI>3_compareV_cconly): Ditto.
3511 (*add<GPI>3_compareV_imm): Ditto.
3512 (add<GPI>3_compareV): Ditto.
3513 (add<mode>3_carryinC): Ditto.
3514 (*add<mode>3_carryinC_zero): Ditto.
3515 (*add<mode>3_carryinC): Ditto.
3516 (add<mode>3_carryinV): Ditto.
3517 (*add<mode>3_carryinV_zero): Ditto.
3518 (*add<mode>3_carryinV): Ditto.
3520 2018-07-19 Michael Collison <michael.collison@arm.com>
3521 Richard Henderson <rth@redhat.com>
3523 * config/aarch64/aarch64-modes.def (CC_V): New.
3524 * config/aarch64/aarch64-protos.h
3525 (aarch64_addti_scratch_regs): Declare
3526 (aarch64_subvti_scratch_regs): Declare.
3527 (aarch64_expand_subvti): Declare.
3528 (aarch64_gen_unlikely_cbranch): Declare
3529 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
3530 for signed overflow using CC_Vmode.
3531 (aarch64_get_condition_code_1): Handle CC_Vmode.
3532 (aarch64_gen_unlikely_cbranch): New function.
3533 (aarch64_addti_scratch_regs): New function.
3534 (aarch64_subvti_scratch_regs): New function.
3535 (aarch64_expand_subvti): New function.
3537 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3539 * config/aarch64/aarch64-option-extensions.def: New entry for profile
3541 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
3542 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
3545 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3548 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3549 address check not strict.
3551 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3553 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
3555 (store_pair_lanes<mode>): Likewise.
3556 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
3557 enum value 'ADDR_QUERY_LDP_STP_N'.
3558 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
3559 (aarch64_print_address_internal): Add declaration.
3560 (aarch64_print_ldpstp_address): Remove.
3561 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
3562 (aarch64_print_operand): Change printing of 'y'.
3563 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
3564 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
3565 'true' rather than '1'.
3566 * config/aarch64/constraints.md (Uml): Likewise.
3567 (Uml): Rename to Umn.
3570 2018-07-19 Richard Biener <rguenther@suse.de>
3572 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
3574 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
3575 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
3576 (vn_reference_hasher): Likewise.
3577 (struct vn_tables_s): Remove obstack and alloc-pool members.
3578 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
3579 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
3580 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
3581 (vn_reference_insert_pieces): Likewise.
3582 (alloc_vn_nary_op_noinit): Adjust.
3583 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
3584 (vn_phi_eq): Adjust.
3585 (shared_lookup_phiargs): Remove.
3586 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
3587 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
3588 (visit_reference_op_call): Likewise.
3589 (copy_nary, copy_phi, copy_reference): Remove.
3590 (process_scc): Rewind the obstack when iterating. Do not
3591 copy the elements to valid_info but just move them from one
3592 hashtable to the other.
3593 (allocate_vn_table): Adjust.
3594 (free_vn_table): Likewise.
3595 (init_scc_vn): Likewise.
3596 (free_scc_vn): Likewise.
3598 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
3601 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
3602 indirect_return as function type attribute.
3603 (ix86_attribute_table): Change indirect_return to function
3605 * doc/extend.texi: Update indirect_return attribute.
3607 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
3609 * wide-int.h (widest2_int): New.
3610 * gimple-fold.c (arith_overflowed_p): Use it.
3611 * tree.h (widest2_int_cst): New.
3612 * tree-vrp.c (wide_int_binop_overflow): Rename from
3613 vrp_int_const_binop.
3614 Rewrite to work on trees.
3615 (extract_range_from_multiplicative_op_1): Abstract code to...
3616 (wide_int_range_min_max): ...here.
3617 (wide_int_range_cross_product): ...and here.
3618 (extract_range_from_binary_expr_1): Abstract overflow code to...
3619 (wide_int_range_mult_wrapping): ...here.
3620 * tree-vrp.h (wide_int_range_cross_product): New.
3621 (wide_int_range_mult_wrapping): New.
3623 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
3624 Julia Koval <julia.koval@intel.com>
3626 * config/i386/x86-tune-costs.h (skylake_memcpy,
3627 skylake_memset): Replace rep_prefix with unrolling for size 512.
3629 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
3632 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
3633 comparision with EQ_EXPR in last stmt.
3635 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
3637 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
3638 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
3639 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
3640 previously known as "PowerPC AltiVec Built-in Functions". Move
3641 some material to new subsubsections "PowerPC AltiVec Built-in
3642 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
3644 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
3645 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
3646 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
3647 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
3649 2018-07-18 Richard Biener <rguenther@suse.de>
3651 PR tree-optimization/86557
3652 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
3655 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
3657 * config/s390/s390.c (s390_function_profiler): Generate CFI.
3659 2018-07-17 Jeff Law <law@redhat.com>
3661 * config/arm/arm.c (get_label_padding): Update for recent
3662 changes to label_to_alignment.
3664 PR tree-optimization/86010
3665 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
3667 * config/mips/mips.c (vr4130_align_insns): Update for recent
3668 changes to label_to_alignment.
3670 * config/frv/frv.c (frv_label_align): Update for recent changes
3671 to label_to_alignment.
3673 * config/nios2/nios2.c (nios2_label_align): Update for recent
3674 changes which dropped ALIGN_LABELS_LOG.
3676 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
3678 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
3679 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
3681 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
3683 * config/arc/arc.c (arc_label_align): Use align_labels instead of
3684 deprecated align_labels_log.
3686 2018-07-17 Richard Biener <rguenther@suse.de>
3689 * dwarf2out.c (init_sections_and_labels): Always generate
3690 a debug_line_str_section for early LTO debug.
3691 (dwarf2out_finish): Reset debug_line_str_hash output early.
3692 Bump counter for extra dwarf5 .debug_loc labels to not conflict
3694 (dwarf2out_early_finish): Output debug_line_str.
3696 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3698 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
3699 index register on z196 or later.
3701 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3703 * config/s390/s390.c (s390_default_align): Set default function
3705 (s390_override_options_after_change): Call s390_default align.
3706 (s390_option_override_internal): Call s390_default_align.
3707 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
3709 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3712 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
3713 to _looptemp_ clauses, other than the first two.
3715 2018-07-17 Martin Liska <mliska@suse.cz>
3717 * opts.c: Do not enable OPT_falign_* for -Os.
3719 2018-07-17 Martin Liska <mliska@suse.cz>
3721 * align.h (MAX_CODE_ALIGN): New.
3722 (MAX_CODE_ALIGN_VALUE): New.
3723 * common/config/i386/i386-common.c (ix86_handle_option):
3724 (MAX_CODE_ALIGN): Moved to align.h.
3725 * final.c (MAX_CODE_ALIGN): Likewise.
3726 * opts.c (parse_and_check_align_values):
3727 (MAX_CODE_ALIGN): Likewise.
3728 (MAX_CODE_ALIGN_VALUE): Likewise.
3730 2018-07-17 Martin Liska <mliska@suse.cz>
3732 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
3733 in order to fulfil coding style.
3734 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
3735 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3736 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3737 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
3738 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
3739 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
3740 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
3741 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
3742 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3744 2018-07-17 Martin Liska <mliska@suse.cz>
3746 * align.h: New file.
3747 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
3749 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
3750 align_flags of label_to_alignment.
3751 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
3753 * config/m68k/m68k.c: Do not use removed align_labels_value and
3755 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
3756 (LOOP_ALIGN): Likewise.
3757 (LABEL_ALIGN): Likewise.
3758 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3759 Remove not used macro.
3760 (rs6000_loop_align): Change return type to align_flags.
3761 (rs6000_loop_align_max_skip): Remove.
3762 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
3763 Change return type to align_flags.
3764 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3765 Remove not used macro.
3766 (rs6000_loop_align): Change return type to align_flags.
3767 (rs6000_loop_align_max_skip): Remove.
3768 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
3769 * config/rx/rx-protos.h (rx_align_for_label): Make it
3771 * config/rx/rx.c (rx_align_for_label): Change return type
3773 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
3775 into align_flags class.
3776 (LABEL_ALIGN): Likewise.
3777 (LOOP_ALIGN): Likewise.
3778 * config/s390/s390.c (s390_label_align): Use align_flags
3780 (s390_asm_output_function_label): Likewise.
3781 * config/sh/sh.c (sh_override_options_after_change):
3782 Use align_flags class directly without macros.
3783 (find_barrier): Likewise.
3784 (barrier_align): Likewise.
3785 (sh_loop_align): Likewise.
3786 * config/spu/spu.c (spu_option_override):
3787 Use align_flags_tuple::get_value instead of removed macros.
3788 (spu_sched_init): Likewise.
3789 * config/spu/spu.h (GTY): Likewise.
3790 * config/visium/visium.c (visium_option_override):
3791 Set "8" as default secondary alignment.
3792 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
3793 in order to guarantee secondary alignment of 8.
3794 * coretypes.h: Include align.h header file.
3795 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
3796 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
3797 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
3798 * doc/tm.texi.in: Likewise.
3799 * final.c (struct label_alignment): Remove not used structure.
3800 (LABEL_ALIGN): Change type to align_flags.
3801 (LOOP_ALIGN): Likewise.
3802 (JUMP_ALIGN): Likewise.
3803 (default_loop_align_max_skip): Remove.
3804 (default_label_align_max_skip): Likewise.
3805 (default_jump_align_max_skip): Likewise.
3806 (default_label_align_after_barrier_max_skip):
3807 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
3808 (LABEL_TO_MAX_SKIP): Remove.
3809 (label_to_alignment): Return align_flags type instead of integer.
3810 (label_to_max_skip): Remove.
3811 (align_fuzz): Use align_flags type.
3812 (compute_alignments): Use align_flags type and use align_flags::max
3813 to combine multiple alignments.
3814 (grow_label_align): Grow vec instead of C array.
3815 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
3816 (shorten_branches): Use align_flags type and use align_flags::max
3817 to combine multiple alignments.
3818 (final_scan_insn_1): Remove usage of secondary alignment that comes
3819 from label alignment, but instead use proper secondary alignment
3820 which is computed in grow_label_align.
3821 * flags.h (struct align_flags_tuple): Move to align.h.
3822 (struct align_flags): Likewise.
3823 (state_align_loops): Rename to align_loops.
3824 (state_align_jumps): Rename to align_jumps.
3825 (state_align_labels): Rename to align_labels.
3826 (state_align_functions): Rename to align_functions.
3827 (align_loops_log): Remove.
3828 (align_jumps_log): Remove.
3829 (align_labels_log): Remove.
3830 (align_functions_log): Remove.
3831 (align_loops_max_skip): Remove.
3832 (align_jumps_max_skip): Remove.
3833 (align_labels_max_skip): Remove.
3834 (align_functions_max_skip): Remove.
3835 (align_loops_value): Remove.
3836 (align_jumps_value): Remove.
3837 (align_labels_value): Remove.
3838 (align_functions_value): Remove.
3839 * output.h (label_to_alignment): Change return type to align_flags.
3840 (label_to_max_skip): Remove.
3841 * target.def: Remove loop_align_max_skip, label_align_max_skip,
3842 jump_align_max_skip macros.
3843 * targhooks.h (default_loop_align_max_skip): Remove.
3844 (default_label_align_max_skip): Likewise.
3845 (default_jump_align_max_skip): Likewise.
3846 (default_label_align_after_barrier_max_skip): Remove.
3847 * toplev.c (read_log_maxskip): Use ::normalize function.
3848 (parse_N_M): Remove not used argument and also call ::normalize.
3849 (parse_alignment_opts): Do not pass unused arguments.
3850 * varasm.c (assemble_start_function): Use directly align_functions
3851 instead of removed macros.
3852 * system.h: Do not poison removed macros.
3854 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3857 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
3858 and cond temporaries don't have reference type if iterator has
3859 pointer type. For init use &for_pre_body instead of pre_p if
3860 for_pre_body is non-empty.
3862 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3864 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
3865 double-double modes to SFmode directly directly.
3866 (trunc<mode>sf2_fprs): Delete.
3868 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3870 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
3871 for conversions between IFmode and the decimal floating point modes.
3872 (init_float128_ieee): Use the correct names for conversions between
3873 KFmode and the decimal floating point modes.
3875 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3877 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
3878 for the conversions between TDmode and IFmode.
3879 (init_float128_ieee): Use more correct names for the conversions
3880 between TDmode and KFmode.
3882 2018-07-16 Jakub Jelinek <jakub@redhat.com>
3884 PR tree-optimization/86526
3885 * builtins.c (expand_builtin_memcmp): Formatting fixes.
3886 (inline_expand_builtin_string_cmp): Likewise.
3887 (inline_string_cmp): Likewise. Use c_readstr instead of
3888 builtin_memcpy_read_str. Add unit_mode temporary.
3890 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
3893 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
3894 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
3896 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3898 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
3899 Alphabetize prototypes of built-in functions, separating out
3900 built-in functions that are listed in this section but should be
3901 described elsewhere.
3903 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
3906 * expmed.c (emit_store_flag): Do not emit setcc followed by a
3907 conditional move when trapping comparison was split to a
3908 non-trapping one (and vice versa).
3910 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3912 * config/s390/s390.c (s390_function_profiler): Generate nops
3913 instead of profiler call sequences.
3914 * config/s390/s390.opt: Add the new option.
3916 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3918 * config/s390/s390.c (s390_function_profiler): Generate
3919 __mcount_loc section.
3920 * config/s390/s390.opt: Add the new option.
3922 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3924 * common.opt: Add the new warning.
3925 * config/s390/s390.c (s390_function_profiler): Emit "brasl
3926 %r0,__fentry__" when -mfentry is specified.
3927 (s390_option_override_internal): Disallow -mfentry for 31-bit
3929 * config/s390/s390.opt: Add the new option.
3931 2018-07-16 Richard Biener <rguenther@suse.de>
3934 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
3935 for function-local FUNCTION_DECL and RESULT_DECL immediately.
3937 2018-07-16 Martin Liska <mliska@suse.cz>
3940 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
3943 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3945 * config/arc/arcHS.md: Update ARCHS scheduling rules.
3947 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3949 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
3951 * config/arc/arc-cpus.def (hs4x): New cpu.
3953 * config/arc/arc-tables.opt: Regenerate.
3954 * config/arc/arc.c (arc_sched_issue_rate): New function.
3955 (TARGET_SCHED_ISSUE_RATE): Define.
3956 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
3957 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
3959 (attr tune): Add ARCHS4x tune values.
3960 (attr tune_dspmpy): Define.
3961 (*tst): Correct instruction type.
3962 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
3963 * config/arc/arcHS4x.md: New file.
3964 * config/arc/fpu.md: Update instruction type attributes.
3965 * config/arc/t-multilib: Regenerate.
3967 2018-07-16 Tom de Vries <tdevries@suse.de>
3970 * var-tracking.c (vt_initialize): Fix pre_dec handling.
3972 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3974 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
3977 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
3979 PR tree-optimization/86514
3980 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
3981 conversion to a boolean type from a type with greater precision.
3983 2018-07-16 Tom de Vries <tdevries@suse.de>
3985 * var-tracking.c (vt_initialize): Print adjusted insn slim if
3986 dump_flags request TDF_SLIM.
3988 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
3990 * fold-const.c (int_const_binop_1): Abstract...
3991 (wide_int_binop): ...wide int code here.
3992 (poly_int_binop): ...poly int code here.
3993 Abstract the rest of int_const_binop_1 into int_const_binop.
3994 * fold-const.h (wide_int_binop): New.
3995 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
3996 Remove useless PLUS/MINUS_EXPR case.
3997 (zero_nonzero_bits_from_vr): Move wide int code...
3998 (zero_nonzero_bits_from_bounds): ...here.
3999 (extract_range_from_binary_expr_1): Move mask optimization code...
4000 (range_easy_mask_min_max): ...here.
4001 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
4002 (range_easy_mask_min_max): New.
4004 2018-07-15 Jeff Law <law@redhat.com>
4007 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
4010 2018-07-14 Jim Wilson <jimw@sifive.com>
4012 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
4014 2018-07-14 Paul Koning <ni1d@arrl.net>
4016 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
4018 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
4020 * lto-streamer-out.c (copy_function_or_variable): Dump info about
4023 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
4024 Steve Munroe <munroesj52@gmail.com>
4026 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
4027 (_mm_andnot_si128): Likewise.
4028 (_mm_or_si128): Likewise.
4029 (_mm_xor_si128): Likewise.
4031 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
4034 * builtins.c (expand_builtin_memcmp): Inline the calls first
4035 when result_eq is false.
4036 (expand_builtin_strcmp): Inline the calls first.
4037 (expand_builtin_strncmp): Likewise.
4038 (inline_string_cmp): New routine. Expand a string compare
4039 call by using a sequence of char comparison.
4040 (inline_expand_builtin_string_cmp): New routine. Inline expansion
4041 a call to str(n)cmp/memcmp.
4042 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
4044 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
4046 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
4048 * config/arm/driver-arm.c: Include arm-native.h.
4049 (host_detect_local_cpu): Use auto-generated data tables.
4050 (vendors, arm_cpu_table): Delete. Move part information to ...
4051 * config/arm/arm-cpus.in: ... here.
4052 * config/arm/parsecpu.awk (gen_native): New function.
4053 (vendor, part): New CPU fields.
4054 (END): Add support for building the native CPU detection tables.
4055 * config/arm/t-arm (arm-native.h): Add build rule.
4056 (driver-arm.o): Add dependency on arm-native.h.
4058 2018-07-13 Richard Biener <rguenther@suse.de>
4061 * match.pd (addr1 - addr2): Allow either of the operand to
4064 2018-07-13 Tom de Vries <tdevries@suse.de>
4066 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
4069 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
4071 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
4072 arrays instead of numbered variables.
4074 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
4076 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
4077 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
4079 2018-07-13 Richard Biener <rguenther@suse.de>
4082 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
4083 instead of get_context_die.
4085 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
4086 Richard Biener <rguenther@suse.de>
4089 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
4090 that the loop latch destination where phi is defined.
4092 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
4094 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
4095 (riscv_expand_epilogue): Add assertion to check interrupt mode.
4096 (riscv_set_current_function): Extract getting interrupt type to new
4098 (riscv_get_interrupt_type): New function.
4099 (riscv_merge_decl_attributes): New function, checking interrupt type is
4101 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
4103 2018-07-12 Paul Koning <ni1d@arrl.net>
4105 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
4108 2018-07-12 Paul Koning <ni1d@arrl.net>
4110 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
4111 zero reference, add doloop_end instead.
4112 * doc/md.texi (decrement_and_branch_until_zero): Remove.
4113 (Looping patterns): Remove decrement_and_branch_until_zero. Add
4114 detail for doloop_end.
4116 2018-07-12 Martin Sebor <msebor@redhat.com>
4119 * attribs.c (decl_attributes): Reject conflicting attributes before
4120 calling attribute handlers.
4122 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
4124 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
4126 (gcc::dump_manager::get_dump_file_name): likewise.
4127 (dump_begin): Likewise.
4128 * dumpfile.h (dump_begin): Update prototype.
4129 (gcc::dump_manager::get_dump_file_name,
4130 gcc::dump_manager::get_dump_file_name): Update prototype.
4132 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4134 * internal-fn.h (vectorizable_internal_fn_p): New function.
4135 * tree-vect-slp.c (compatible_calls_p): Likewise.
4136 (vect_build_slp_tree_1): Remove nops argument. Handle calls
4137 to internal functions.
4138 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
4140 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4142 * fold-const.h (inverse_conditions_p): Declare.
4143 * fold-const.c (inverse_conditions_p): New function.
4144 * match.pd: Use inverse_conditions_p. Add folds of view_converts
4145 that test the inverse condition of a conditional internal function.
4146 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
4147 * internal-fn.c (internal_fn_mask_index): Handle conditional
4149 (vectorized_internal_fn_supported_p): New function.
4150 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
4151 (any_pred_load_store): Replace with...
4152 (need_to_predicate): ...this new variable.
4153 (redundant_ssa_names): New variable.
4154 (ifcvt_can_use_mask_load_store): Move initial checks to...
4155 (ifcvt_can_predicate): ...this new function. Handle tree codes
4156 for which a conditional internal function exists.
4157 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
4158 instead of ifcvt_can_use_mask_load_store. Update after variable
4160 (predicate_load_or_store): New function, split out from
4161 predicate_mem_writes.
4162 (check_redundant_cond_expr): New function.
4163 (value_available_p): Likewise.
4164 (predicate_rhs_code): Likewise.
4165 (predicate_mem_writes): Rename to...
4166 (predicate_statements): ...this. Use predicate_load_or_store
4167 and predicate_rhs_code.
4168 (combine_blocks, tree_if_conversion): Update after above name changes.
4169 (ifcvt_local_dce): Handle redundant_ssa_names.
4170 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
4171 general conditional functions.
4172 * tree-vect-stmts.c (vectorizable_call): Likewise.
4174 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4175 Alan Hayward <alan.hayward@arm.com>
4176 David Sherwood <david.sherwood@arm.com>
4178 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
4179 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
4180 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
4181 plus and minus and convert them into IFN_COND_FMA-based sequences.
4182 (convert_mult_to_fma): Handle conditional plus and minus.
4184 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4186 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
4187 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
4188 (cond_fnms_optab): New optabs.
4189 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
4191 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
4192 * internal-fn.h (get_conditional_internal_fn): Declare.
4193 (get_unconditional_internal_fn): Likewise.
4194 * internal-fn.c (cond_ternary_direct): New macro.
4195 (expand_cond_ternary_optab_fn): Likewise.
4196 (direct_cond_ternary_optab_supported_p): Likewise.
4197 (FOR_EACH_COND_FN_PAIR): Likewise.
4198 (get_conditional_internal_fn): New function.
4199 (get_unconditional_internal_fn): Likewise.
4200 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
4201 (gimple_match_op::gimple_match_op): Add a new overload for 5
4203 (gimple_match_op::set_op): Likewise.
4204 (gimple_resimplify5): Declare.
4205 * genmatch.c (decision_tree::gen): Generate simplifications for
4207 * gimple-match-head.c (gimple_simplify): Define an overload for
4208 5 operands. Handle calls with 5 arguments in the top-level overload.
4209 (convert_conditional_op): Handle conversions from unconditional
4210 internal functions to conditional ones.
4211 (gimple_resimplify5): New function.
4212 (build_call_internal): Pass a fifth operand.
4213 (maybe_push_res_to_seq): Likewise.
4214 (try_conditional_simplification): Try converting conditional
4215 internal functions to unconditional internal functions.
4216 Handle 3-operand unconditional forms.
4217 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
4218 Define ternary equivalents of the current rules for binary conditional
4220 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
4222 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
4223 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
4224 (optab): Handle them.
4225 (SVE_COND_FP_TERNARY): New int iterator.
4226 (sve_fmla_op, sve_fmad_op): New int attributes.
4227 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
4228 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
4229 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
4231 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4233 * target.def (preferred_else_value): New target hook.
4234 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
4235 * doc/tm.texi: Regenerate.
4236 * targhooks.h (default_preferred_else_value): Declare.
4237 * targhooks.c (default_preferred_else_value): New function.
4238 * internal-fn.h (conditional_internal_fn_code): Declare.
4239 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
4240 (get_conditional_internal_fn): Use it.
4241 (conditional_internal_fn_code): New function.
4242 * gimple-match.h (gimple_match_cond): New struct.
4243 (gimple_match_op): Add a cond member function.
4244 (gimple_match_op::gimple_match_op): Update all forms to take a
4246 * genmatch.c (expr::gen_transform): Use the same condition as res_op
4247 for the suboperation, but don't specify a particular else_value.
4248 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
4249 (visit_nary_op, visit_reference_op_load): Pass
4250 gimple_match_cond::UNCOND to the gimple_match_op constructor.
4251 * gimple-match-head.c: Include tree-eh.h
4252 (convert_conditional_op): New function.
4253 (maybe_resimplify_conditional_op): Likewise.
4254 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
4255 (gimple_resimplify2): Likewise.
4256 (gimple_resimplify3): Likewise.
4257 (gimple_resimplify4): Likewise.
4258 (maybe_push_res_to_seq): Return null for conditional operations.
4259 (try_conditional_simplification): New function.
4260 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
4262 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
4264 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
4266 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
4268 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
4270 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
4272 (hash_tree): Do not hash DECL_FCONTEXT
4273 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
4274 Do not stream DECL_FCONTEXT.
4275 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
4276 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
4278 2018-07-12 Richard Biener <rguenther@suse.de>
4281 * dwarf2out.c (gen_block_die): Only output blocks when they have
4282 at least one !DECL_IGNORED_P variable.
4284 2018-07-12 Richard Biener <rguenther@suse.de>
4287 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
4288 Remove -mieee-fp handling.
4290 2018-07-12 Richard Biener <rguenther@suse.de>
4292 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
4293 left-over from last patch.
4295 2018-07-12 Jakub Jelinek <jakub@redhat.com>
4297 PR tree-optimization/86492
4298 * gimple-ssa-store-merging.c
4299 (imm_store_chain_info::coalesce_immediate_stores): Call
4300 check_no_overlap even for the merge_overlapping case. Formatting fix.
4302 2018-07-12 Richard Biener <rguenther@suse.de>
4305 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
4306 move possibly trapping operations into the conditional.
4308 2018-07-12 Richard Biener <rguenther@suse.de>
4310 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
4311 (vn_lookup_simplify_result): Remove recursion limit applied
4313 (vn_nary_build_or_lookup_1): Adjust.
4314 (try_to_simplify): Likewise.
4315 * gimple-match-head.c (gimple_resimplify1): Instead apply one
4317 (gimple_resimplify2): Likewise.
4318 (gimple_resimplify3): Likewise.
4319 (gimple_resimplify4): Likewise.
4321 2018-07-11 Jakub Jelinek <jakub@redhat.com>
4323 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
4324 Use __mmask64 type instead of __mmask8 for __M argument.
4325 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
4326 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
4328 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
4329 __mmask16 for __M argument.
4330 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
4331 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
4332 to __mmask16 instead of __mmask8.
4333 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
4334 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
4335 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
4336 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
4337 instead of __mmask16 for __U argument.
4338 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
4339 __mmask16 instead of __mmask8 for __U argument.
4340 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
4342 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
4344 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
4346 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
4348 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
4350 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
4352 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
4354 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
4356 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
4358 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
4360 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
4361 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
4362 return type as well as __M argument type and all casts from __mmask8
4364 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
4365 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
4366 return type as well as __M argument type and all casts from __mmask8
4368 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
4369 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
4370 return type as well as __M argument type and all casts from __mmask8
4372 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
4373 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
4374 return type as well as __M argument type and all casts from __mmask8
4376 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
4377 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
4380 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
4382 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
4383 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
4386 2018-07-11 Paul Koning <ni1d@arrl.net>
4388 * doc/md.texi (define_subst): Document how multiple occurrences of
4389 the same argument in the replacement pattern are handled.
4391 2018-07-11 Paul Koning <ni1d@arrl.net>
4393 * doc/extend.texi (Common Variable Attributes): Move "mode" into
4395 (Common Type Attributes): Add "mode" attribute.
4397 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
4399 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
4400 stream DECL_ORIGINAL_TYPE.
4401 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
4402 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
4403 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
4404 Do not walk original type.
4405 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
4407 (write_ts_decl_non_common_tree_pointers): Do not stream
4409 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
4410 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
4412 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
4414 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
4415 thread twice from the same starting edge.
4417 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
4419 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
4420 * gimple.c (gimple_call_nonnull_result_p): ...here...
4421 (gimple_call_nonnull_arg): ...and here.
4422 * gimple.h (gimple_call_nonnull_result_p): New.
4423 (gimple_call_nonnull_arg): New.
4425 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
4427 * config/arm/arm-cpus.in: Move information from fpu field of each
4428 cpu definition to the isa field.
4429 * config/arm/parsecpu.awk (fpu): Delete match rule.
4430 (gen_comm_data): Don't add bits from the CPU's FPU entry.
4432 2018-07-11 Richard Biener <rguenther@suse.de>
4435 * dwarf2out.c (init_sections_and_labels): Use
4436 output_asm_line_debug_info consistently.
4437 (dwarf2out_early_finish): Likewise.
4438 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
4441 2018-07-11 Richard Biener <rguenther@suse.de>
4443 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
4444 Rework father_bb setting in a way to avoid propagating constants
4445 multiple times on a loop body.
4447 2018-07-10 Mark Wielaard <mark@klomp.org>
4450 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
4452 2018-07-10 Richard Biener <rguenther@suse.de>
4454 * hash-map.h (hash_map::iterator::operator*): Return
4455 references to key and value.
4457 2018-07-10 Jakub Jelinek <jakub@redhat.com>
4460 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
4461 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
4462 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
4463 (gimplify_omp_for): For composite loops, move outer
4464 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
4465 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
4466 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
4467 TREE_LIST for both the original class iterator and the "last" helper
4468 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
4469 loop, remember has_decl_expr from outer composite loops for the
4470 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
4472 2018-07-09 Martin Sebor <msebor@redhat.com>
4476 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
4477 accessing implicitly initialized array elements.
4478 * expr.c (string_constant): Handle string initializers of
4479 character arrays within aggregates.
4480 * gimple-fold.c (fold_array_ctor_reference): Add argument.
4481 Store element offset. As a special case, handle zero size.
4482 (fold_nonarray_ctor_reference): Same.
4483 (fold_ctor_reference): Add argument. Store subobject offset.
4484 * gimple-fold.h (fold_ctor_reference): Add argument.
4486 2018-07-09 Paul Koning <ni1d@arrl.net>
4488 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
4489 (pdp11_insn_cost): New function.
4490 (pdp11_md_asm_adjust): New function.
4491 (TARGET_INVALID_WITHIN_DOLOOP): Define.
4492 (pdp11_rtx_costs): Update to match machine better.
4493 (output_addr_const_pdp11): Correct format mismatch warnings.
4494 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
4495 * config/pdp11/pdp11.md: General change to add base_cost and/or
4496 length attributes for use by new pdp11_insn_cost function.
4497 (MIN_BRANCH): Correct definition.
4499 (doloop_end): Use standard pattern name for looping pattern.
4500 (doloop_end_nocc): New.
4501 (movsf): Add another constraint alternative.
4502 (zero_extendqihi2): Add constraint alternatives for not in place
4504 (zero_extendhisi2): Remove.
4505 (shift patterns): Add CC handling variants.
4509 (define_peephole2): New peephole to recognize mov that sets CC for
4512 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4514 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
4515 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
4518 2018-07-09 Jakub Jelinek <jakub@redhat.com>
4521 * real.c (real_nextafter): Return true if result is denormal.
4523 2018-07-09 Martin Liska <mliska@suse.cz>
4525 * common.opt: Add back wrongly removed attribute.
4527 2018-07-09 Richard Biener <rguenther@suse.de>
4530 * dwarf2out.c (gen_block_die): For an early generated DIE
4531 always output high/low PC attributes.
4533 2018-07-09 Tom de Vries <tdevries@suse.de>
4535 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
4536 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
4537 onto VAR_DECL with abstract origin.
4539 2018-07-07 Jim Wilson <jimw@sifive.com>
4541 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
4543 2018-07-07 Tom de Vries <tdevries@suse.de>
4545 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
4547 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
4549 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
4551 (combine_bound): Use wide-int overflow calculation instead of
4553 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
4555 * fold-const.c (int_const_binop_2): Same.
4556 (extract_muldiv_1): Same.
4557 (fold_div_compare): Same.
4558 (fold_abs_const): Same.
4560 * poly-int.h (add): Same.
4564 * predict.c (predict_iv_comparison): Same.
4565 * profile-count.c (slow_safe_scale_64bit): Same.
4566 * simplify-rtx.c (simplify_const_binary_operation): Same.
4567 * tree-chrec.c (tree_fold_binomial): Same.
4568 * tree-data-ref.c (split_constant_offset_1): Same.
4569 * tree-if-conv.c (idx_within_array_bound): Same.
4570 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
4571 * tree-ssa-phiopt.c (minmax_replacement): Same.
4572 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
4573 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
4574 * vr-values.c (vr_values::adjust_range_with_scev): Same.
4575 * wide-int.cc (wi::add_large): Same.
4576 (wi::mul_internal): Same.
4577 (wi::sub_large): Same.
4578 (wi::divmod_internal): Same.
4579 * wide-int.h: Change overflow type to overflow_type for neg, add,
4580 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
4581 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
4582 mul_internal, divmod_internal.
4583 (overflow_type): New enum.
4584 (accumulate_overflow): New.
4586 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
4588 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
4589 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
4591 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
4593 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
4594 argument is checked for zero before entering loop, avoid checking again.
4596 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
4598 * gimplify.h (generic_expr_could_trap_p): Set as global function.
4599 * gimplify.h (generic_expr_could_trap_p): Likwise.
4600 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
4602 2018-07-06 Jakub Jelinek <jakub@redhat.com>
4604 PR tree-optimization/86401
4605 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
4606 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
4607 (fold_bit_and_mask): ... here. New helper function for match.pd.
4608 * fold-const.h (fold_bit_and_mask): Declare.
4609 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
4611 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
4614 * target.def (translate_mode_attribute): New hook.
4615 * targhooks.h (default_translate_mode_attribute): Declare.
4616 * targhooks.c (default_translate_mode_attribute): New function.
4617 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
4618 * doc/tm.texi: Regenerate.
4619 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
4620 (rs6000_translate_mode_attribute): New function.
4622 2018-07-06 Paul Koning <ni1d@arrl.net>
4624 * doc/md.texi (define_split): Document DONE and FAIL.
4625 (define_peephole2): Ditto.
4627 2018-07-05 Jeff Law <law@redhat.com>
4629 PR tree-optimization/86010
4630 * tree-ssa-dse.c (compute_trims): More aggressively trim at
4631 both the head and tail of mem* and str* calls.
4633 2018-07-05 Jim Wilson <jimw@sifive.com>
4635 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
4637 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
4639 * config/aarch64/aarch64-simd.md: correct flags text for
4640 MIN_EXPR replacement.
4642 2018-07-05 James Clarke <jrtc27@jrtc27.com>
4644 * configure: Regenerated.
4646 2018-07-05 Carl Love <cel@us.ibm.com>
4648 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
4649 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
4650 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
4651 VSX_BUILTIN_DOUBLEL_V4SF.
4653 2018-07-05 Martin Sebor <msebor@redhat.com>
4656 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
4657 than its domain to compute its the upper bound of a char array.
4659 2018-07-05 Nathan Sidwell <nathan@acm.org>
4661 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
4662 * doc/cpp.texi: Update comment.
4663 * doc/tm.texi: Rebuilt.
4664 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
4665 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
4666 * doc/extend.texi (Backwards Compatibility): Clarify it is system
4667 headers affected by extern "C".
4668 * system.h: Poison NO_IMPLICIT_EXTERN_C.
4669 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
4670 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
4671 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
4672 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
4673 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
4674 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
4675 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
4676 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
4677 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
4678 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
4679 config/sparc/sp64-elf.h, config/spu/spu.h,
4680 config/stormy16/stormy16.h, config/v850/v850.h,
4681 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
4682 define NO_IMPLICIT_EXTERN_C.
4683 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
4685 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4688 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4689 instead of GET_MODE_SIZE when comparing Units.
4691 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4694 * rtlanal.c (set_noop_p): Constrain on mode change,
4695 include hard-reg-set.h
4697 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4699 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
4701 2018-07-05 Jakub Jelinek <jakub@redhat.com>
4704 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4707 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4709 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4711 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4714 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4716 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4718 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
4721 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
4723 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
4725 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
4726 neg_*_op* variables.
4728 2018-07-04 Martin Liska <mliska@suse.cz>
4730 * tree-switch-conversion.c: Define
4731 max_ratio_for_speed and max_ratio_for_size constants.
4733 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4734 Martin Liska <mliska@suse.cz>
4739 * common.opt: Rename align options with 'str_' prefix.
4740 * common/config/i386/i386-common.c (set_malign_value): New
4742 (ix86_handle_option): Use it to set -falign-* options/
4743 * config/aarch64/aarch64-protos.h (struct tune_params): Change
4744 type from int to string.
4745 * config/aarch64/aarch64.c: Update default values from int
4747 * config/alpha/alpha.c (alpha_override_options_after_change):
4749 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
4750 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4751 max skip conditionally.
4752 * config/i386/freebsd.h (SUBALIGN_LOG): New.
4753 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4754 max skip conditionally.
4755 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4756 max skip conditionally.
4757 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
4758 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4759 max skip conditionally.
4760 * config/i386/i386.c (struct ptt): Change type from int to
4762 (ix86_default_align): Set default values.
4763 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
4764 max skip conditionally.
4765 * config/i386/iamcu.h (SUBALIGN_LOG): New.
4766 (ASM_OUTPUT_MAX_SKIP_ALIGN):
4767 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
4768 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4769 max skip conditionally.
4770 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
4771 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
4772 * config/i386/x86-64.h (SUBALIGN_LOG): New.
4773 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4774 max skip conditionally.
4775 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4776 * config/ia64/ia64.c (ia64_option_override): Set default values
4777 for alignment options.
4778 * config/m68k/m68k.c: Handle new str_align_* options.
4779 * config/mips/mips.c (mips_set_compression_mode): Change
4781 (mips_option_override): Set default values for options.
4782 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4784 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4786 * config/rx/rx.c (rx_option_override): Likewise.
4787 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
4788 (LABEL_ALIGN): Use align_labels_log.
4789 (LOOP_ALIGN): Use align_loops_align.
4790 * config/s390/s390.c (s390_asm_output_function_label): Use new
4792 * config/sh/sh.c (sh_override_options_after_change):
4793 Change type of constants.
4794 * config/spu/spu.c (spu_sched_init): Likewise.
4795 * config/sparc/sparc.c (sparc_option_override): Set default
4797 * config/visium/visium.c (visium_option_override): Likewise.
4798 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
4799 emit p2align format with last argument if it's not needed.
4800 * doc/invoke.texi: Document extended format of -falign-*.
4801 * final.c: Use align_labels alignment.
4802 * flags.h (struct target_flag_state): Change type to use
4804 (struct align_flags_tuple): New.
4805 (struct align_flags): Likewise.
4806 (align_loops_log): Redefine macro to use new types.
4807 (align_loops_max_skip): Redefine macro to use new types.
4808 (align_jumps_log): Redefine macro to use new types.
4809 (align_jumps_max_skip): Redefine macro to use new types.
4810 (align_labels_log): Redefine macro to use new types.
4811 (align_labels_max_skip): Redefine macro to use new types.
4812 (align_functions_log): Redefine macro to use new types.
4813 (align_loops): Redefine macro to use new types.
4814 (align_jumps): Redefine macro to use new types.
4815 (align_labels): Redefine macro to use new types.
4816 (align_functions): Redefine macro to use new types.
4817 (align_functions_max_skip): Redefine macro to use new types.
4818 (align_loops_value): New macro.
4819 (align_jumps_value): New macro.
4820 (align_labels_value): New macro.
4821 (align_functions_value): New macro.
4822 * function.c (invoke_set_current_function_hook): Propagate
4823 alignment values from flags to global variables default in
4825 * ipa-icf.c (sem_function::equals_wpa): Use
4826 cl_optimization_option_eq instead of memcmp.
4827 * lto-streamer.h (cl_optimization_stream_out): Support streaming
4829 (cl_optimization_stream_in): Likewise.
4830 * optc-save-gen.awk: Support strings in cl_optimization.
4831 * opth-gen.awk: Likewise.
4832 * opts.c (finish_options): Remove error checking of invalid
4834 (MAX_CODE_ALIGN): Remove.
4835 (MAX_CODE_ALIGN_VALUE): Likewise.
4836 (parse_and_check_align_values): New function.
4837 (check_alignment_argument): Likewise.
4838 (common_handle_option): Use check_alignment_argument.
4839 * opts.h (parse_and_check_align_values): Declare.
4840 * toplev.c (init_alignments): Remove.
4841 (read_log_maxskip): New.
4842 (parse_N_M): Likewise.
4843 (parse_alignment_opts): Likewise.
4844 (backend_init_target): Remove usage of init_alignments.
4845 * toplev.h (parse_alignment_opts): Declare.
4846 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
4848 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
4849 * tree.c (cl_option_hasher::equal): New.
4850 * varasm.c: Use new global macros.
4852 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4854 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
4855 Use a simpler align directive also if MAXSKIP = ALIGN-1.
4856 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4857 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4858 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4859 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4860 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
4861 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
4862 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
4863 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4864 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4865 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4866 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4868 2018-07-04 Martin Liska <mliska@suse.cz>
4869 Jonathan Wakely <jwakely@redhat.com>
4871 * coverage.c: Use correct type.
4872 * doc/invoke.texi: Language correction.
4874 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
4877 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
4878 ENDBRANCH for non-tail call which may return via indirect branch.
4879 * doc/extend.texi: Document indirect_return attribute.
4881 2018-07-03 Martin Sebor <msebor@redhat.com>
4883 PR tree-optimization/86274
4884 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
4886 (format_floating): Correct handling of infinities and NaNs.
4888 2018-07-03 Martin Sebor <msebor@redhat.com>
4890 * print-tree.c (print_real_cst): New function.
4891 (print_node_brief): Call it.
4892 (print_node): Ditto.
4894 2018-07-03 Jeff Law <law@redhat.com>
4896 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
4897 into a single pattern.
4899 * config/h8300/h8300.md (ors code_iterator): New.
4900 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
4901 a single pattern and single splitter.
4902 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
4903 (iorqi3_1, xorqi3_1): Likewise.
4904 (iorqi3, xorqi3 expanders): Similarly.
4906 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
4907 (movmd_internal) into a single pattern using the P mode iterator.
4908 (movmd splitters): Similarly.
4909 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
4910 (movsd splitters): Similarly.
4912 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
4913 ADDB, ADDW and ADDL into a single ADD attribute which selects the
4914 right table based on the size of the operand.
4915 * config/h8300/h8300.md (length_table): Corresponding changes. All
4916 references to "addb", "addw" and "addl" changed to "add".
4917 (btst patterns): Merge two variants into a single pattern.
4918 (tstqi, tsthi): Likewise.
4919 (addhi3_incdec, addsi3_incdec): Likewise.
4920 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
4921 (mulhi3, mulsi3): Likewise.
4922 (udivhi3, udivsi3): Likewise.
4923 (divhi3, divsi3): Likewise.
4924 (andorqi3, andorhi3, andorsi3): Likewise.
4926 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
4929 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
4930 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
4932 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4934 PR tree-optimization/85694
4935 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
4936 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
4938 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
4939 (<u>avg<mode>3_ceil): New patterns.
4941 2018-07-03 David Malcolm <dmalcolm@redhat.com>
4943 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
4944 scan-tree-dump directive.
4945 * gcc.dg/vect/slp-perm-2.c: Likewise.
4946 * gcc.dg/vect/slp-perm-3.c: Likewise.
4947 * gcc.dg/vect/slp-perm-5.c: Likewise.
4948 * gcc.dg/vect/slp-perm-6.c: Likewise.
4949 * gcc.dg/vect/slp-perm-7.c: Likewise.
4950 * gcc.dg/vect/slp-perm-8.c: Likewise.
4952 2018-07-03 Marek Polacek <polacek@redhat.com>
4955 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
4957 2018-07-03 Richard Biener <rguenther@suse.de>
4960 * tree-ssa-structalias.c (find_func_clobbers): Properly
4961 handle indirect calls.
4963 2018-07-03 Jeff Law <law@redhat.com>
4965 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
4966 (shifts): New code iterator.
4967 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
4968 expander. Fix HImode handling on H8/SX.
4969 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
4970 (subqi3, subhi3, subsi3 expanders): Likewise.
4971 (andqi3, andhi3, andsi3 expanders): Likewise.
4972 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
4973 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
4974 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
4975 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
4976 (zero_extendqihi2, zero_extendqisi2): Likewise.
4977 (extendqihi2, extendqisi2): Likewise.
4978 (rotlqi3, rotlhi3, rotlsi3): Likewise.
4979 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
4980 (rotlqi3_1, rotlhi3_1): Likewise.
4981 (logicalhi3_sn, logicalsi3_sn): Likewise.
4982 (logicalhi3, logicalsi3): Likewise.
4984 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4986 * tree-vect-patterns.c (vect_recog_rotate_pattern)
4987 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4988 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
4989 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
4990 type to append_pattern_def_seq instead of creating a stmt_vec_info
4992 (build_mask_conversion): Likewise. Remove vinfo argument.
4993 (vect_add_conversion_to_patterm): Likewise, renaming to...
4994 (vect_add_conversion_to_pattern): ...this.
4995 (vect_recog_mask_conversion_pattern): Update call to
4996 build_mask_conversion. Pass the vector type to
4997 append_pattern_def_seq here too.
4998 (vect_recog_gather_scatter_pattern): Update call to
4999 vect_add_conversion_to_pattern.
5001 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5003 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
5004 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5005 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
5006 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
5007 STMT_VINFO_PATTERN_DEF_SEQ to null here.
5008 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
5009 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
5010 append_pattern_def_seq instead of new_pattern_def_seq.
5011 (vect_recog_divmod_pattern): Do both of the above.
5012 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
5015 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5017 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
5018 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
5019 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
5020 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
5021 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
5022 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
5023 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
5024 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
5025 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
5026 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
5027 parameter with a single stmt_vec_info.
5028 (vect_recog_func_ptr): Likewise.
5029 (vect_recog_gather_scatter_pattern): Likewise, folding in...
5030 (vect_try_gather_scatter_pattern): ...this.
5031 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
5032 the stmt_vec_info of the statement to be matched. Don't clear
5033 STMT_VINFO_RELATED_STMT.
5034 (vect_pattern_recog): Update call accordingly.
5036 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5038 PR tree-optimization/85694
5039 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
5040 (uavgM3_ceil): Document new optabs.
5041 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
5042 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
5044 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
5045 (savg_ceil_optab): New optabs.
5046 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
5047 (vect_vect_recog_func_ptrs): Add it.
5048 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
5049 constant directly from the associated lhs.
5051 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5053 * tree-vect-patterns.c (vect_split_statement): New function.
5054 (vect_convert_input): Use it to try to split an existing cast.
5056 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5058 * poly-int.h (print_hex): New function.
5059 * dumpfile.h (dump_dec, dump_hex): Declare.
5060 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
5061 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
5062 min_input_precision, operation_precision and operation_sign.
5063 * tree-vect-patterns.c (vect_get_range_info): New function.
5064 (vect_same_loop_or_bb_p, vect_single_imm_use)
5065 (vect_operation_fits_smaller_type): Delete.
5066 (vect_look_through_possible_promotion): Add an optional
5067 single_use_p parameter.
5068 (vect_recog_over_widening_pattern): Rewrite to use new
5069 stmt_vec_info infomration. Handle one operation at a time.
5070 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
5071 (vect_truncatable_operation_p, vect_set_operation_type)
5072 (vect_set_min_input_precision): New functions.
5073 (vect_determine_min_output_precision_1): Likewise.
5074 (vect_determine_min_output_precision): Likewise.
5075 (vect_determine_precisions_from_range): Likewise.
5076 (vect_determine_precisions_from_users): Likewise.
5077 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
5078 (vect_vect_recog_func_ptrs): Put over_widening first.
5080 (vect_pattern_recog): Call vect_determine_precisions.
5082 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
5084 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
5085 statements that have been replaced by further pattern statements.
5086 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
5088 2018-07-03 Richard Biener <rguenther@suse.de>
5090 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
5091 always set *dt. Dump vectype in vectype overload.
5092 * dumpfile.h (dump_gimple_expr): New function.
5093 (dump_gimple_expr_loc): Likewise.
5094 * dumpfile.c (dump_gimple_expr): New function.
5095 (dump_gimple_expr_loc): Likewise.
5097 2018-07-02 Jeff Law <law@redhat.com>
5099 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
5100 the H8/300, H8/300H and H8/S variants into a single pattern.
5101 (movhi_h8300, movqi_h8300hs): Similarly.
5102 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
5103 (QHI mode iterator): New.
5105 * config/h8300/h8300.md: Remove trailing whitespace.
5107 2018-07-02 Jim Wilson <jimw@sifive.com>
5109 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
5110 instead of emit_insn for interrupt returns.
5111 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
5112 (riscv_sret, riscv_uret): Likewise.
5114 2018-07-02 David Malcolm <dmalcolm@redhat.com>
5116 * pretty-print.c (selftest::test_pp_format): Move save and restore
5117 of quotes to class auto_fix_quotes, and add an instance.
5118 * selftest.c: Include "intl.h".
5119 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
5120 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
5121 * selftest.h (selftest::auto_fix_quotes): New class.
5123 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
5125 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
5126 (aarch64_sve_prepare_conditional_op): Remove.
5127 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
5128 Allow aarch64_simd_reg_or_zero as select operand; remove
5129 the aarch64_sve_prepare_conditional_op call.
5130 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
5131 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
5132 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
5133 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
5134 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
5135 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
5136 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
5137 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
5138 and a splitters to match all of the *_any patterns.
5139 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
5141 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
5142 (SVE_COND_FP_BINARY_REV): Remove.
5143 (sve_int_op_rev, sve_fp_op_rev): New.
5144 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
5145 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
5146 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
5147 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
5148 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
5149 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
5150 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
5151 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
5152 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
5154 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
5155 Remove match_dup 1 from the inner unspec.
5156 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
5158 * config/aarch64/aarch64.md (movprfx): New attr.
5159 (length): Default movprfx to 8.
5160 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
5161 (*madd<SVE_I>, *msub<SVE_I): Likewise.
5162 (*<su>mul<SVE_I>3_highpart): Likewise.
5163 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
5164 (*v<ASHIFT><SVE_I>3): Likewise.
5165 (*<su><MAXMIN><SVE_I>3): Likewise.
5166 (*<su><MAXMIN><SVE_F>3): Likewise.
5167 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
5168 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
5169 (*div<SVE_F>4): Likewise.
5171 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
5173 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
5176 2018-07-02 Richard Biener <rguenther@suse.de>
5178 PR tree-optimization/86363
5179 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
5180 memset argument refers to a non-variable address.
5182 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
5184 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
5185 {PLUS,MINUS}_EXPR code to...
5186 (adjust_symbolic_bound): ...here,
5187 (combine_bound): ...here,
5188 (set_value_range_with_overflow): ...and here.
5190 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
5192 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
5194 (extract_range_from_abs_expr): ...here.
5196 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
5198 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
5199 -fno-omit-frame-pointer when not optimizing.
5201 2018-07-02 Martin Liska <mliska@suse.cz>
5204 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
5205 (propagate_nothrow): Likewise.
5207 2018-07-02 Martin Liska <mliska@suse.cz>
5210 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
5212 2018-07-02 David Malcolm <dmalcolm@redhat.com>
5214 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
5215 function in r262149, changing "loc" param from source_location to
5216 const dump_location_t &.
5217 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
5218 declaration, as above.
5220 2018-07-01 Paul Koning <ni1d@arrl.net>
5222 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
5223 -munit-asm, -mgnu-asm, -mdec-asm.
5224 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
5225 (pdp11_output_labelref): New.
5226 (pdp11_output_def): New.
5227 (pdp11_output_addr_vec_elt): New.
5228 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
5229 %# and %@ format codes.
5230 (pdp11_option_override): New.
5231 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
5232 (pdp11_output_ident): New.
5233 (pdp11_asm_named_section): New.
5234 (pdp11_asm_init_sections): New.
5235 (pdp11_file_start): New.
5236 (pdp11_file_end): New.
5237 (output_ascii): Use .ascii/.asciz for -mdec-asm.
5238 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
5239 %o, like %c but octal.
5240 (pdp11_option_override): New.
5241 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
5243 (DATA_SECTION_ASM_OP): Ditto.
5244 (READONLY_DATA_SECTION_ASM_OP): New.
5245 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
5246 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
5247 (ASM_OUTPUT_LABELREF): Ditto.
5248 (ASM_OUTPUT_DEF): Ditto.
5249 (ASM_OUTPUT_EXTERNAL): New.
5250 (ASM_OUTPUT_SOURCE_FILENAME): New.
5251 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
5252 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
5253 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
5254 %# and %@ format codes.
5255 * config/pdp11/pdp11.opt (mgnu-asm): New.
5256 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
5257 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
5258 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
5260 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
5262 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
5263 dereferencing path[] beyond its length.
5265 (debug_all_paths): New.
5266 (rewire_first_differing_edge): New.
5267 (adjust_paths_after_duplication): New.
5268 (duplicate_thread_path): Call adjust_paths_after_duplication.
5270 (thread_through_all_blocks): Add new argument to
5271 duplicate_thread_path.
5273 2018-06-30 Jim Wilson <jimw@sifive.com>
5275 * config/riscv/predicates.md (p2m1_shift_operand): New.
5276 (high_mask_shift_operand): New.
5277 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
5278 pattern using p2m1_shift_operand.
5279 (lshsi3_zero_extend_3+2): New combiner pattern using
5280 high_mask_shift_operand.
5282 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5284 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
5286 (vect_recog_rotate_pattern): ...here.
5287 (vect_convert_input): Try to insert casts of invariants in the
5289 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
5290 preheader to be empty.
5292 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5294 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
5295 vector type. If given, install it in the new statement's
5297 (vect_element_precision): New function.
5298 (vect_unpromoted_value): New struct.
5299 (vect_unpromoted_value::vect_unpromoted_value): New function.
5300 (vect_unpromoted_value::set_op): Likewise.
5301 (vect_look_through_possible_promotion): Likewise.
5302 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
5303 (vect_widened_op_tree, vect_convert_input): Likewise.
5304 (vect_convert_inputs, vect_convert_output): Likewise.
5305 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
5306 to handle the optional cast of the multiplication result and
5307 vect_widened_op_tree to detect the widened multiplication itself.
5308 Do not require the input and output of promotion casts to have
5309 the same sign, but base the signedness of the operation on the
5310 input rather than the result. If the pattern includes two
5311 promotions, check that those promotions have the same sign.
5312 Do not restrict the MULT_EXPR handling to a double-width result;
5313 handle quadruple-width results and wider. Use vect_convert_inputs
5314 to convert the inputs to the common type.
5315 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
5316 to handle the optional cast of the ABS result. Also allow a sign
5317 change or a sign extension between the ABS and MINUS.
5318 Use vect_widened_op_tree to detect the widened subtraction and use
5319 vect_convert_inputs to convert the inputs to the common type.
5320 (vect_handle_widen_op_by_const): Delete.
5321 (vect_recog_widen_op_pattern): New function.
5322 (vect_recog_widen_mult_pattern): Use it.
5323 (vect_recog_widen_shift_pattern): Likewise.
5324 (vect_recog_widen_sum_pattern): Use
5325 vect_look_through_possible_promotion to handle the promoted
5328 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5330 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
5331 the containing gimple_seq *.
5332 * gimple-iterator.h (gsi_for_stmt): Declare it.
5333 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
5334 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
5335 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
5336 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
5337 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
5339 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
5341 (vect_mark_pattern_stmts): ...here. Handle cases in which the
5342 statement being replaced is part of an existing pattern
5343 definition sequence, inserting the new pattern statements before
5345 (vect_pattern_recog_1): Don't return a bool. If the statement
5346 is already part of a pattern, instead apply pattern matching
5347 to the pattern definition statements. Don't clear the
5348 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
5349 (vect_pattern_recog): Don't break after the first match;
5350 continue processing the pattern definition statements instead.
5351 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
5353 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5355 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5356 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5357 (vect_recog_widen_sum_pattern): Use it.
5359 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5361 * tree-vect-loop.c (vectorizable_reduction): Assert that the
5362 phi is not a pattern statement and has not been replaced by
5363 a pattern statement.
5364 * tree-vect-patterns.c (type_conversion_p): Don't check
5365 STMT_VINFO_IN_PATTERN_P.
5366 (vect_recog_vector_vector_shift_pattern): Likewise.
5367 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
5368 the pattern statement rather than the original statement; check
5369 directly for a WIDEN_MULT_EXPR here.
5370 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
5371 vect_is_simple_use to return the pattern statement rather
5372 than the original statement; use is_pattern_stmt_p to check
5373 for such a pattern statement.
5374 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
5375 to return the pattern statement rather than the original statement;
5376 don't do the same transformation here.
5377 (vect_is_simple_use): If the defining statement has been replaced
5378 by a pattern statement, return the pattern statement instead.
5379 Remove the corresponding (local) transformation from the vectype
5382 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5384 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
5385 end and default to null.
5386 * tree-vect-loop.c (vect_create_epilog_for_reduction)
5387 (vectorizable_reduction): Update calls accordingly, dropping the
5388 gimple ** argument if the passed-back statement isn't needed.
5389 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
5390 (vect_recog_rotate_pattern): Likewise.
5391 (vect_recog_mask_conversion_pattern): Likewise.
5392 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5393 (vect_mask_constant_operand_p): Likewise.
5394 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
5395 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
5396 (get_group_load_store_type, get_load_store_type): Likewise.
5397 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
5398 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
5399 (vectorizable_conversion, vectorizable_assignment): Likewise.
5400 (vectorizable_shift, vectorizable_operation): Likewise.
5401 (vectorizable_store, vect_is_simple_cond): Likewise.
5402 (vectorizable_condition, vectorizable_comparison): Likewise.
5403 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
5404 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
5405 and move it to the end. Cope with null def_stmt_outs.
5407 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
5409 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
5411 2018-06-29 Jeff Law <law@redhat.com>
5413 * config/v850/v850.c (v850_legitimate_address_p): Handle large
5414 displacements for TARGET_V850E2V3 and newer.
5415 (TARGET_LRA_P): Remove. Defaults to LRA now.
5416 * config/v850/v850.md (sign23byte_load): Remove.
5417 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
5418 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
5420 2018-06-29 Martin Liska <mliska@suse.cz>
5423 * coverage.c (coverage_init): Mangle full path name.
5424 * doc/invoke.texi: Document the change.
5425 * gcov-io.c (mangle_path): New.
5426 * gcov-io.h (mangle_path): Likewise.
5427 * gcov.c (mangle_name): Use mangle_path for path mangling.
5429 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5431 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
5432 if starting source register is not even.
5434 2018-06-29 Martin Liska <mliska@suse.cz>
5436 PR tree-optimization/86263
5437 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
5438 Make edge redirection.
5440 2018-06-29 David Malcolm <dmalcolm@redhat.com>
5442 * dumpfile.c (dump_loc): Add indentation based on scope depth.
5443 (dump_scope_depth): New variable.
5444 (get_dump_scope_depth): New function.
5445 (dump_begin_scope): New function.
5446 (dump_end_scope): New function.
5447 * dumpfile.h (get_dump_scope_depth): New declaration.
5448 (dump_begin_scope): New declaration.
5449 (dump_end_scope): New declaration.
5450 (class auto_dump_scope): New class.
5451 (AUTO_DUMP_SCOPE): New macro.
5452 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
5455 2018-06-29 Richard Biener <rguenther@suse.de>
5457 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
5458 compute_all_dependences succeeds.
5459 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
5460 exceed --param loop-max-datarefs-for-datadeps.
5462 2018-06-29 Jakub Jelinek <jakub@redhat.com>
5464 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
5466 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
5469 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
5470 alternative 0 in preferred_for_speed attribute.
5472 2018-06-28 Paul Koning <ni1d@arrl.net>
5474 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
5475 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
5476 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
5477 * config/pdp11/pdp11.md: Correct "length" attribute calculation
5478 for shift insn patterns.
5480 2018-06-28 David Malcolm <dmalcolm@redhat.com>
5482 * cgraph.c (cgraph_node::get_body): Replace assignments to
5483 "dump_file" with calls to set_dump_file.
5484 * dumpfile.c (alt_dump_file): Make static, and group with...
5485 (alt_flags): ...this definition.
5486 (dumps_are_enabled): New variable.
5487 (refresh_dumps_are_enabled): New function.
5488 (set_dump_file): New function.
5489 (set_alt_dump_file): New function.
5490 (gcc::dump_manager::dump_start): Replace assignments to
5491 "dump_file" and "alt_dump_file" with calls to set_dump_file and
5493 (gcc::dump_manager::dump_finish): Likewise.
5494 * dumpfile.h (alt_dump_file): Delete decl.
5495 (dumps_are_enabled): New variable decl.
5496 (set_dump_file): New function decl.
5497 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
5499 * tree-nested.c (lower_nested_functions): Replace assignments to
5500 "dump_file" with calls to set_dump_file.
5502 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
5504 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
5505 goto_locus of each outgoing edge of each basic block.
5507 2018-06-28 Richard Biener <rguenther@suse.de>
5509 * dwarf2out.c (decl_scope_table): Remove.
5510 (push_decl_scope): Likewise.
5511 (pop_decl_scope): Likewise.
5512 (gen_type_die_for_member): Do not call push/pop_decl_scope.
5513 (gen_struct_or_union_type_die): Likewise.
5514 (gen_tagged_type_die): Likewise.
5515 (dwarf2out_init): Do not initialize decl_scope_table.
5516 (dwarf2out_c_finalize): Do not free it.
5518 2018-06-28 Richard Biener <rguenther@suse.de>
5520 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
5521 deciding whether to not re-use a DIE.
5523 2018-06-28 Richard Biener <rguenther@suse.de>
5525 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
5526 DW_AT_abstract_origin attribute.
5528 2018-06-28 Martin Liska <mliska@suse.cz>
5530 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
5531 Use newly introduced constants.
5532 * tree-switch-conversion.h (struct jump_table_cluster):
5533 Define max_ratio_for_size and max_ratio_for_speed.
5535 2018-06-28 Martin Liska <mliska@suse.cz>
5537 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5538 Add new checking assert to catch invalid state.
5539 (jump_table_cluster::can_be_handled): Handle single case
5541 (jump_table_cluster::is_beneficial): Bail out for such case.
5542 (bit_test_cluster::find_bit_tests):
5543 Add new checking assert to catch invalid state.
5544 (bit_test_cluster::can_be_handled): Handle single case
5546 (bit_test_cluster::is_beneficial): Bail out for such case.
5547 (switch_decision_tree::analyze_switch_statement):
5550 2018-06-28 Martin Liska <mliska@suse.cz>
5552 * common.opt: Introduce -completion option.
5553 * gcc.c (driver_handle_option): Handle it.
5554 (driver::main): Print completions if completion
5556 * opt-suggestions.c (option_proposer::get_completions):
5558 (option_proposer::suggest_completion): Likewise.
5559 (option_proposer::find_param_completions): Likewise.
5560 (verify_autocompletions): Likewise.
5561 (test_completion_valid_options): Likewise.
5562 (test_completion_valid_params): Likewise.
5563 (in_completion_p): Likewise.
5564 (empty_completion_p): Likewise.
5565 (test_completion_partial_match): Likewise.
5566 (test_completion_garbage): Likewise.
5567 (opt_proposer_c_tests): Likewise.
5568 * opt-suggestions.h: Declare new functions.
5569 * opts.c (common_handle_option): Handle OPT__completion_.
5570 * selftest-run-tests.c (selftest::run_tests): Add
5571 opt_proposer_c_tests.
5572 * selftest.c (assert_str_startswith): New.
5573 * selftest.h (assert_str_startswith): Likewise.
5574 (opt_proposer_c_tests): New.
5575 (ASSERT_STR_STARTSWITH): Likewise.
5577 2018-06-28 Martin Liska <mliska@suse.cz>
5579 * Makefile.in: Add opt-suggestions.o.
5580 * gcc-main.c: Include opt-suggestions.h.
5581 * gcc.c (driver::driver): Likewise.
5582 (driver::~driver): Remove m_option_suggestions.
5583 (driver::build_option_suggestions): Moved to option_proposer.
5584 (driver::suggest_option): Likewise.
5585 (driver::handle_unrecognized_options): Use option_proposer.
5586 * gcc.h (class driver): Add new memver m_option_proposer.
5587 * opt-suggestions.c: New file.
5588 * opt-suggestions.h: New file.
5590 2018-06-28 Martin Liska <mliska@suse.cz>
5592 * vec.h (class auto_string_vec): New (moved from auto_argvec).
5593 (auto_string_vec::~auto_string_vec): Likewise.
5595 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
5597 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
5598 prevent_decl_creation_for_types fields up and add reset_location field.
5599 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
5600 statement if id->reset_location is true.
5601 (copy_edges_for_bb): Do not set goto_locus on the new edges if
5602 id->reset_location is true.
5603 (copy_phis_for_bb): Force input_location on the arguments if
5604 id->reset_location is true.
5605 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
5606 is set on the function to be inlined.
5608 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
5610 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
5612 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
5614 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
5615 registers for Pmode.
5616 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
5617 hard registers for the clobbered pseudo.
5619 2018-06-27 Paul Koning <ni1d@arrl.net>
5621 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
5622 mutually exclusive options.
5623 * config/pdp11/constraints.md (h): New constraint.
5624 (O): Update definition to match shift code generation.
5625 (D): New constraint.
5626 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
5628 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
5630 (output_jump): Change arguments.
5631 (pdp11_fixed_cc_regs): New function.
5632 (pdp11_cc_mode): Ditto.
5633 (pdp11_expand_shift): Ditto.
5634 (pdp11_assemble_shift): Ditto.
5635 (pdp11_small_shift): Ditto.
5636 (pdp11_branch_cost): Remove.
5637 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
5639 (pdp11_register_move_cost): Update for CC registers.
5640 (pdp11_rtx_costs): Add case for LSHIFTRT.
5641 (pdp11_output_jump): Add CCNZ mode conditional branches.
5642 (notice_update_cc_on_set): Remove.
5643 (pdp11_cc_mode): New function.
5644 (simple_memory_operand): Correct pre/post decrement case.
5645 (no_side_effect_operand): New function.
5646 (pdp11_regno_reg_class): Add CC_REGS class.
5647 (pdp11_fixed_cc_regs): New function.
5648 (pdp11_small_shift): New function.
5649 (pdp11_expand_shift): New function to expand shift insns.
5650 (pdp11_assemble_shift): New function to output shifts.
5651 (pdp11_branch_cost): Remove.
5652 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
5653 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
5654 (WCHAR_TYPE): Ditto.
5655 (PTRDIFF_TYPE): Ditto.
5656 (ADJUST_INSN_LENGTH): New macro.
5657 (FIXED_REGISTERS): Add CC registers.
5658 (CALL_USED_REGISTERS): Ditto.
5660 (REG_CLASS_NAMES): Ditto.
5661 (REG_CLASS_CONTENTS): Ditto.
5662 (SELECT_CC_MODE): Use new function.
5663 (TARGET_FLAGS_REGNUM): New macro.
5664 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
5665 (cc0_reg_rtx): Remove.
5666 (CC_STATUS_MDEP): Remove.
5667 (CC_STATUS_MDEFP_INIT): Remove.
5668 (CC_IN_FPU): Remove.
5669 (NOTICE_UPDATE_CC): Remove.
5670 (REGISTER_NAMES): Add CC registers.
5671 (BRANCH_COST): Change to constant 1.
5672 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
5674 * config/pdp11/pdp11.opt (mbcopy): Remove.
5675 (mbcopy-builtin): Remove.
5676 (mbranch-cheap): Remove.
5677 (mbranch-expensive): Remove.
5678 * config/pdp11/predicates.md (expand_shift_operand): Update to
5679 match shift code generation.
5680 (ccnz_operator): New predicate.
5681 * doc/invoke.texi (PDP-11 Options): Remove deleted options
5682 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
5683 Remove non-existent option -mabshi, -mno-abshi. Document mutually
5685 * doc/md.texi (PDP-11): Document new D and h constraints. Update
5686 description of O constraint.
5688 2018-06-27 Jeff Law <law@redhat.com>
5689 Austin Law <austinklaw@gmail.com>
5691 * config/v850/v850.md (addsi3_set_flags): New pattern.
5692 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
5693 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
5694 (zero_extendhisi2_v850_set_flags): Likewise.
5695 (zero_extendqisi2_v850_set_flags): Likewise.
5696 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
5697 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
5698 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
5700 * config/v850/v850-protos.h (notice_update_cc): Remove.
5701 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
5702 (v850_print_operand): Handle 'D' and "d".
5703 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
5704 Add handling of arithmetic/logical operations compared against zero.
5705 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
5706 Do not look at v850_compare_op, instead get mode from last argument.
5707 (v850_gen_compare): Remove
5708 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
5709 after reload for prologue insns.
5710 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
5712 (construct_save_jarl): Likewise.
5713 (TARGET_FLAGS_REGNUM): Define.
5714 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
5715 (NOTICE_UPDATE_CC): Remove.
5716 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
5717 than cc0. Conditionalize on reload_completed.
5718 (cmpsi_insn, setfcc_insn): Likewise.
5719 (tst1 splitter): Turn into define_and_split which sets the flags
5721 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
5722 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
5723 (cstoresf4, cstoredf4): Clobber the flags.
5724 (cmpsi, cmpsf, cmpdf): Remove expanders.
5725 (setf_insn): Remove pattern.
5726 (addsi3): Turn into define_and_split which clobbers the flags after
5727 reload and a suitable pattern (addsi3_clobber_flags) for use after
5729 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
5730 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
5731 (ashrsi3, ashrsi3_v850e2): Likewise.
5732 (bins): Clobber the flags.
5733 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
5734 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
5735 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
5736 (call_value_internal_short, call_value_internal_long): Likewise.
5737 (callt_save_interrupt, callt_return_interrupt): Likewise.
5738 (save_interrupt, return_interrupt): Likewise.
5739 (callt_save_all_interrupt, save_all_interrupt): Likewise.
5740 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
5741 (restore_all_interrupt, _restore_all_interrupt): Likewise.
5742 (All FP comparisons): Only allow after reload has completed.
5744 (divh, divhu): Tweak output template.
5745 (branch_z_normal, branch_z_invert): Remove
5746 (branch_nz_normal, branch_nz_invert): Likewise.
5747 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
5749 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
5750 * config/v850/v850.c (notice_update_cc): Remove.
5751 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
5752 (CC_NO_CARRY): Likewise.
5753 (NOTICE_UPDATE_CC): Define to nothing.
5754 * config/v850/v850.md: Remove block comment on cc0 handling
5755 Remove "cc" attribute from all patterns. Remove cc_status handling
5756 from all patterns. Minor formatting fixes.
5758 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5760 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
5761 (cortex-a76.cortex-a55): Likewise.
5762 * config/aarch64/aarch64-tune.md: Regenerate.
5763 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
5764 cortex-a76.cortex-a55.
5766 2018-06-27 Jeff Law <law@redhat.com>
5768 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
5769 (MULTILIB_DIRNAMES): Similarly.
5771 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
5773 * gimple.h (gimple_return_retbnd): Delete.
5774 (gimple_return_set_retbnd): Likewise.
5775 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
5776 gimple_return_set_retbnd.
5777 * gimple-pretty-print.c (dump_gimple_return): Remove call to
5778 gimple_return_retbnd and adjust.
5779 * tree-inline.h (struct copy_body_data): Remove retbnd field.
5780 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
5781 Explicitly return NULL in a couple more cases. Move assertion
5782 on debug statements and remove unreachable code.
5783 (reset_debug_binding): Do not test id->retbnd.
5784 (expand_call_inline): Do not set it.
5786 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5788 * configure.ac: Add --disable-gcov option.
5789 * configure: Regenerate.
5790 * Makefile.in: Honour @enable_gcov@.
5791 * doc/install.texi: Document --disable-gcov.
5793 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5795 * config/arm/arm-cpus.in (cortex-a76): New entry.
5796 (cortex-a76.cortex-a55): Likewise.
5797 * config/arm/arm-tables.opt: Regenerate.
5798 * config/arm/arm-tune.md: Likewise.
5799 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
5800 * doc/invoke.texi (ARM Options): Document cortex-a76 and
5801 cortex-a76.cortex-a55.
5803 2018-06-27 Tamar Christina <tamar.christina@arm.com>
5806 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
5808 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
5810 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
5812 (EPILOGUE_USES): Likewise.
5814 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
5816 * tree-inline.c (remap_location): New function extracted from...
5817 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
5818 (copy_phis_for_bb): ...here. Call remap_location.
5819 (copy_cfg_body): Adjust call to copy_edges_for_bb.
5821 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
5823 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
5824 unaligned vsx for 16B memset.
5826 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
5829 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
5830 ieee128_float_type_node to long_double_type_node unless
5831 TARGET_LONG_DOUBLE_128 is set.
5833 2018-06-26 David Malcolm <dmalcolm@redhat.com>
5835 * cfgloop.c (get_loop_location): Convert return type from
5836 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
5837 by implicit construction from rtx_insn *, and using
5838 dump_user_location_t::from_function_decl for the fallback case.
5839 * cfgloop.h (get_loop_location): Convert return type from
5840 location_t to dump_user_location_t.
5841 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
5842 dump_printf_loc to pass in a dump_location_t rather than a
5843 location_t, via the gimple stmt.
5844 * coverage.c (get_coverage_counts): Update calls to
5845 dump_printf_loc to pass in dump_location_t rather than a
5847 * doc/optinfo.texi (Dump types): Convert example of
5848 dump_printf_loc from taking "locus" to taking "insn". Update
5849 description of the "_loc" calls to cover dump_location_t.
5850 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
5852 (dump_user_location_t::dump_user_location_t): New constructors,
5853 from gimple *stmt and rtx_insn *.
5854 (dump_user_location_t::from_function_decl): New function.
5855 (dump_loc): Make static.
5856 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
5857 const dump_location_t &.
5858 (dump_generic_expr_loc): Delete.
5859 (dump_printf_loc): Convert param "loc" from location_t to
5860 const dump_location_t &.
5861 (selftest::test_impl_location): New function.
5862 (selftest::dumpfile_c_tests): New function.
5863 * dumpfile.h: Include "profile-count.h".
5864 (class dump_user_location_t): New class.
5865 (struct dump_impl_location_t): New struct.
5866 (class dump_location_t): New class.
5867 (dump_printf_loc): Convert 2nd param from source_location to
5868 const dump_location_t &.
5869 (dump_generic_expr_loc): Delete.
5870 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
5871 const dump_location_t &.
5872 * gimple-fold.c (fold_gimple_assign): Update call to
5873 dump_printf_loc to pass in a dump_location_t rather than a
5874 location_t, via the gimple stmt.
5875 (gimple_fold_call): Likewise.
5876 * gimple-loop-interchange.cc
5877 (loop_cand::analyze_iloop_reduction_var): Update for change to
5878 check_reduction_path.
5879 (tree_loop_interchange::interchange): Update for change to
5881 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
5882 change in return-type of find_loop_location.
5883 (graphite_regenerate_ast_isl): Likewise.
5884 * graphite-optimize-isl.c (optimize_isl): Likewise.
5885 * graphite.c (graphite_transform_loops): Likewise.
5886 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
5887 pass in a dump_location_t rather than a location_t, via the
5889 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5890 * ipa.c (walk_polymorphic_call_targets): Likewise.
5891 * loop-unroll.c (report_unroll): Convert "locus" param from
5892 location_t to dump_location_t.
5893 (decide_unrolling): Update for change to get_loop_location's
5895 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
5896 location_t to dump_user_location_t.
5897 (grid_find_single_omp_among_assignments_1): Updates calls to
5898 dump_printf_loc to pass in a dump_location_t rather than a
5899 location_t, via the gimple stmt.
5900 (grid_parallel_clauses_gridifiable): Convert "tloc" from
5901 location_t to dump_location_t. Updates calls to dump_printf_loc
5902 to pass in a dump_location_t rather than a location_t, via the
5904 (grid_inner_loop_gridifiable_p): Likewise.
5905 (grid_dist_follows_simple_pattern): Likewise.
5906 (grid_gfor_follows_tiling_pattern): Likewise.
5907 (grid_target_follows_gridifiable_pattern): Likewise.
5908 (grid_attempt_target_gridification): Convert initialization
5909 of local "grid" from memset to zero-initialization; FIXME: does
5910 this require C++11? Update call to dump_printf_loc to pass in a
5911 optinfo_location rather than a location_t, via the gimple stmt.
5912 * profile.c (read_profile_edge_counts): Updates call to
5913 dump_printf_loc to pass in a dump_location_t rather than a
5915 (compute_branch_probabilities): Likewise.
5916 * selftest-run-tests.c (selftest::run_tests): Call
5918 * selftest.h (dumpfile_c_tests): New decl.
5919 * tree-loop-distribution.c (pass_loop_distribution::execute):
5920 Update for change in return type of find_loop_location.
5921 * tree-parloops.c (parallelize_loops): Likewise.
5922 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
5923 "locus" from location_t to dump_user_location_t.
5924 (canonicalize_loop_induction_variables): Likewise.
5925 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
5926 for change in return type of find_loop_location.
5927 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
5928 to dump_printf_loc to pass in a dump_location_t rather than a
5929 location_t, via the stmt.
5930 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
5932 * tree-vect-loop-manip.c (find_loop_location): Convert return
5933 type from source_location to dump_user_location_t.
5934 (vect_do_peeling): Update for above change.
5935 (vect_loop_versioning): Update for change in type of
5937 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
5938 from location_t to dump_user_location_t.
5939 (vect_estimate_min_profitable_iters): Update for change in type
5941 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
5942 location_t to dump_location_t.
5943 (vect_slp_bb): Update for change in type of vect_location.
5944 * tree-vectorizer.c (vect_location): Convert from source_location
5945 to dump_user_location_t.
5946 (try_vectorize_loop_1): Update for change in vect_location's type.
5947 (vectorize_loops): Likewise.
5948 (increase_alignment): Likewise.
5949 * tree-vectorizer.h (vect_location): Convert from source_location
5950 to dump_user_location_t.
5951 (find_loop_location): Convert return type from source_location to
5952 dump_user_location_t.
5953 (check_reduction_path): Convert 1st param from location_t to
5954 dump_user_location_t.
5955 * value-prof.c (check_counter): Update call to dump_printf_loc to
5956 pass in a dump_user_location_t rather than a location_t; update
5957 call to error_at for change in type of "locus".
5958 (check_ic_target): Update call to dump_printf_loc to
5959 pass in a dump_user_location_t rather than a location_t, via the
5962 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
5964 * config/s390/s390.h (enum processor_flags): Do not use
5965 default tune parameter when -march was specified.
5967 2018-06-26 Jakub Jelinek <jakub@redhat.com>
5970 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
5971 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
5974 2018-06-26 Richard Biener <rguenther@suse.de>
5976 PR tree-optimization/86287
5978 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
5979 (vect_analyze_loop): Initialize n_stmts.
5981 2018-06-26 Richard Biener <rguenther@suse.de>
5984 * fold-const.c (fold_convertible_p): Pointer extension
5987 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
5990 * dwarf2out.c (loc_list_has_views): Adjust comments.
5991 (dw_loc_list): Split single cross-partition range with
5994 2018-06-25 Jeff Law <law@redhat.com>
5996 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
5997 on -mbig-switch by default.
5999 * config/v850/predicates.md (const_float_1_operand): Fix match_code
6001 (const_float_0_operand): Remove unused predicate.
6002 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
6003 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
6004 (recipsf2): New expander. Original pattern now called
6006 (recipdf2, recipdf2_insn): Similarly.
6007 (rsqrtsf2, rsqrtsf2_insn): Similarly
6008 (rsqrtdf2, rsqrtdf2_insn): Similarly
6010 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
6012 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
6013 Simplify logic for FreeBSD (twice).
6015 2018-06-25 Martin Sebor <msebor@redhat.com>
6017 PR tree-optimization/86204
6018 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
6019 a strnlen result if it's less than the length of the string.
6021 2018-06-25 Martin Sebor <msebor@redhat.com>
6023 PR tree-optimization/85700
6024 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
6025 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
6026 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
6028 2018-06-25 Martin Sebor <msebor@redhat.com>
6030 * doc/extend.texi (Zero-length arrays): Update and clarify.
6032 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
6034 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
6035 added IEEE/IBM long double multilib support on PowerPC little
6036 endian Linux systems.
6037 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
6038 (MULTILIB_DEFAULTS): Likewise.
6039 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6041 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
6042 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
6043 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
6045 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
6048 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
6049 (REORDER_45): Likewise.
6051 2018-06-25 Jeff Law <law@redhat.com>
6053 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
6054 dividend to 32 bits. Adjust length.
6055 (udivmodhi4): Cleanup output template. Fix length.
6057 2018-06-25 Carl Love <cel@us.ibm.com>
6059 * config/rs6000/vsx.md: Change word selector to prefered location.
6061 2018-06-25 Richard Biener <rguenther@suse.de>
6063 PR tree-optimization/86304
6064 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
6065 epilogue-if-converted loops as well.
6067 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
6069 * lto-section-out.c (lto_begin_section): Do not print section
6070 name for noaddr and unnumbered dumps.
6072 2018-06-25 Richard Biener <rguenther@suse.de>
6074 * tree-vectorizer.h (struct vec_info_shared): New structure
6075 with parts split out from struct vec_info and loop_nest from
6076 struct _loop_vec_info.
6077 (struct vec_info): Adjust accordingly.
6078 (struct _loop_vec_info): Likewise.
6079 (LOOP_VINFO_LOOP_NEST): Adjust.
6080 (LOOP_VINFO_DATAREFS): Likewise.
6081 (LOOP_VINFO_DDRS): Likewise.
6082 (struct _bb_vec_info): Likewise.
6083 (BB_VINFO_DATAREFS): Likewise.
6084 (BB_VINFO_DDRS): Likewise.
6085 (struct _stmt_vec_info): Add dr_aux member.
6086 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
6087 (DR_MISALIGNMENT_UNINITIALIZED): New.
6088 (set_dr_misalignment): Adjust.
6089 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
6090 (vect_analyze_loop): Adjust prototype.
6091 (vect_analyze_loop_form): Likewise.
6092 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6093 Compute dependences lazily.
6094 (vect_record_base_alignments): Use shared datarefs/ddrs.
6095 (vect_verify_datarefs_alignment): Likewise.
6096 (vect_analyze_data_refs_alignment): Likewise.
6097 (vect_analyze_data_ref_accesses): Likewise.
6098 (vect_analyze_data_refs): Likewise.
6099 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
6100 constructor parameter for shared part.
6101 (vect_analyze_loop_form): Pass in shared part and adjust.
6102 (vect_analyze_loop_2): Pass in storage for the number of
6103 stmts. Move loop nest finding to the caller. Compute
6105 (vect_analyze_loop): Pass in shared part.
6106 (vect_transform_loop): Verify shared datarefs are unchanged.
6107 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
6108 constructor parameter for shared part.
6109 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
6110 (vect_slp_bb): Verify shared datarefs are unchanged before
6112 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
6114 (new_stmt_vec_info): Initialize DR_AUX misalignment to
6115 DR_MISALIGNMENT_UNINITIALIZED.
6116 * tree-vectorizer.c (vec_info::vec_info): Add constructor
6117 parameter for shared part.
6118 (vec_info::~vec_info): Adjust.
6119 (vec_info_shared::vec_info_shared): New.
6120 (vec_info_shared::~vec_info_shared): Likewise.
6121 (vec_info_shared::save_datarefs): Likewise.
6122 (vec_info_shared::check_datarefs): Likewise.
6123 (try_vectorize_loop_1): Construct shared part live for analyses
6124 of a single loop for multiple vector sizes.
6125 * tree-parloops.c (gather_scalar_reductions): Adjust.
6127 2018-06-25 Richard Biener <rguenther@suse.de>
6129 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
6130 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
6131 (vect_analyze_data_refs): Remove similar code from here and
6132 simplify accordingly.
6134 2018-06-25 Richard Biener <rguenther@suse.de>
6136 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
6137 for reverse storage order accesses rather than asserting
6138 they cannot happen here.
6140 2018-06-25 Tom de Vries <tdevries@suse.de>
6143 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
6144 Use data16 instead of .byte for insn prefix.
6146 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
6149 * parser.c (make_char_string_pack): Pass this literal chars
6150 through cpp_interpret_string.
6151 (cp_parser_userdef_numeric_literal): Check the result of
6152 make_char_string_pack.
6154 2018-06-24 Maya Rashish <coypu@sdf.org>
6156 * ginclude/stddef.h: Simplify conditions around avoiding
6157 re-definition of __size_t.
6159 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6161 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
6162 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
6164 2018-06-22 Maya Rashish <coypu@sdf.org>
6166 * doc/invoke.texi (mno-fancy-math-387): Update for changes
6167 made to OpenBSD and NetBSD through the years.
6169 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6171 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
6172 behavior of vec_pack (vector double, vector double) to match
6173 behavior of vec_float2 (vector double, vector double).
6175 2018-06-22 Olivier Hainque <hainque@adacore.com>
6177 * gimplify.c (gimplify_function_tree): Prevent creation
6178 of a trampoline for the address of the current function
6179 passed to entry/exit instrumentation hooks.
6181 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
6184 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
6187 2018-06-22 Martin Liska <mliska@suse.cz>
6189 PR tree-optimization/86263
6190 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6191 Bail out if is_enabled is false.
6192 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
6194 (jump_table_cluster::is_enabled): New function.
6196 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6198 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6199 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
6200 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
6201 (lto_input_ts_binfo_tree_pointers): Likewise.
6202 * tree-streamer-out.c (streamer_write_tree_bitfields,
6203 write_ts_binfo_tree_pointers): Likewise.
6204 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
6206 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6208 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
6210 2018-06-22 Martin Liska <mliska@suse.cz>
6212 * symbol-summary.h (get): Make it pure and inline move
6213 functionality from ::get function.
6214 (get): Remove and inline into ::get and ::get_create.
6215 (get_create): Move code from ::get function.
6217 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6220 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
6221 -x assembler-with-cpp.
6223 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6225 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
6226 _FILE_OFFSET_BITS=64 for C++.
6228 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
6230 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
6231 conversion insn that shows up when pr85657-3.c is compiled using
6232 IEEE 128-bit long double.
6233 (neg<mode>2_internal): Use the correct mode to check whether the
6234 mode is IBM extended.
6235 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
6236 multiply and divide external functions from being created more
6239 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6241 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
6243 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
6244 the edge can be forwarded.
6245 (cfg_layout_merge_blocks): Likewise.
6247 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6249 * except.c (finish_eh_generation): Commit edge insertions only after
6250 the EH edges have been redirected from post-landing to landing pads.
6252 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6254 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
6255 create_tmp_var_for to create the FRAME decl.
6256 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
6258 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6260 * tree-inline.c (copy_edges_for_bb): Minor tweak.
6261 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
6262 debug statement when resetting its value.
6263 (expand_call_inline): Copy the locus of the call onto the assignment
6264 of the return value, if any. Use local variable in more cases.
6266 2018-06-21 Martin Liska <mliska@suse.cz>
6268 * ipa-pure-const.c (propagate_nothrow): Use
6269 funct_state_summaries->get.
6270 (dump_malloc_lattice): Likewise.
6271 (propagate_malloc): Likewise.
6273 2018-06-21 Richard Biener <rguenther@suse.de>
6275 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
6276 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
6277 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
6279 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
6280 BLOCK_ABSTRACT_ORIGIN unconditionally.
6282 2018-06-21 David Malcolm <dmalcolm@redhat.com>
6284 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
6286 * ipa-reference.c (ipa_reference_c_finalize): Delete
6287 ipa_ref_opt_sum_summaries and set it to NULL.
6289 2018-06-21 Tom de Vries <tdevries@suse.de>
6291 PR tree-optimization/85859
6292 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
6293 test with comment from bb_no_side_effects_p.
6295 2018-06-21 Richard Biener <rguenther@suse.de>
6297 PR tree-optimization/86232
6298 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
6299 max for constant niter.
6301 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
6303 * config/aarch64/aarch64-simd.md
6304 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
6306 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
6308 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
6309 Make opernads of the unspec commutative.
6311 2018-06-21 Richard Biener <rguenther@suse.de>
6313 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
6314 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
6315 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
6316 (vect_analyze_data_ref_dependence): Re-order checks to deal with
6318 (vect_record_base_alignments): Do not record base alignment
6319 for gathers or scatters.
6320 (vect_compute_data_ref_alignment): Drop return value that is always
6321 true. Bail out early for gathers or scatters.
6322 (vect_enhance_data_refs_alignment): Bail out early for gathers
6324 (vect_find_same_alignment_drs): Likewise.
6325 (vect_analyze_data_refs_alignment): Remove dead code.
6326 (vect_slp_analyze_and_verify_node_alignment): Likewise.
6327 (vect_analyze_data_refs): For possible gathers or scatters do
6328 not create an alternate DR, just check their possible validity
6329 and mark them. Adjust DECL_NONALIASED handling to not rely
6331 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
6332 update inits of gathers or scatters.
6333 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
6334 Also copy gather/scatter flag to pattern vinfo.
6336 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
6338 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
6339 behavior of vec_packsu (vector unsigned long long, vector unsigned
6340 long long) to match behavior of vec_packs with same signature.
6342 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
6343 Thomas Schwinge <thomas@codesourcery.com>
6344 Cesar Philippidis <cesar@codesourcery.com>
6346 * gimplify.c (gimplify_scan_omp_clauses): Add support for
6347 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
6348 (gimplify_adjust_omp_clauses): Likewise.
6349 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
6350 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
6351 (gimplify_omp_target_update): Update handling of acc update and
6353 * omp-low.c (install_var_field): Remove unused parameter
6354 base_pointers_restrict.
6355 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
6356 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
6358 (omp_target_base_pointers_restrict_p): Delete.
6359 (scan_omp_target): Update call to scan_sharing_clauses.
6360 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
6362 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
6363 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
6364 (convert_local_omp_clauses): Likewise.
6365 * tree-pretty-print.c (dump_omp_clause): Likewise.
6366 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
6368 (omp_clause_code_name): Likewise.
6370 2018-06-20 Jakub Jelinek <jakub@redhat.com>
6373 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
6376 PR tree-optimization/86231
6377 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
6378 anti-range don't overwrite *vr0min before using it to compute *vr0max.
6380 2018-06-20 Tom de Vries <tdevries@suse.de>
6382 PR tree-optimization/86097
6383 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
6384 iv type if signedness of iv type is not the same as that of *nit.
6386 2018-06-20 Jakub Jelinek <jakub@redhat.com>
6388 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
6389 EDGE_EH edges, verify they are all EDGE_EH.
6391 2018-06-20 Maya Rashish <coypu@sdf.org>
6393 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
6395 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6397 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
6398 * config/aarch64/aarch64.c (xgene1_tunings): Add
6399 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
6400 (aarch64_mode_valid_for_sched_fusion_p):
6401 Allow 16-byte modes.
6402 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
6403 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
6405 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
6407 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
6408 * config/aarch64/iterators.md (VQ2): New mode iterator.
6410 2018-06-20 Martin Liska <mliska@suse.cz>
6412 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
6413 Change default ratio from 10 to 8.
6415 2018-06-20 Martin Liska <mliska@suse.cz>
6417 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6419 (bit_test_cluster::find_bit_tests): Likewise.
6420 (switch_decision_tree::analyze_switch_statement): Find clusters.
6421 * tree-switch-conversion.h (struct jump_table_cluster): Document
6424 2018-06-20 Martin Liska <mliska@suse.cz>
6426 * tree-switch-conversion.c (switch_conversion::collect):
6427 Record m_uniq property.
6428 (switch_conversion::expand): Bail out for special conditions.
6429 (group_cluster::~group_cluster): New.
6430 (group_cluster::group_cluster): Likewise.
6431 (group_cluster::dump): Likewise.
6432 (jump_table_cluster::emit): New.
6433 (switch_decision_tree::fix_phi_operands_for_edges): New.
6434 (struct case_node): Remove struct.
6435 (jump_table_cluster::can_be_handled): New.
6436 (case_values_threshold): Moved to header.
6437 (reset_out_edges_aux): Likewise.
6438 (jump_table_cluster::is_beneficial): New.
6439 (bit_test_cluster::can_be_handled): Likewise.
6440 (add_case_node): Remove.
6441 (bit_test_cluster::is_beneficial): New.
6442 (case_bit_test::cmp): New.
6443 (bit_test_cluster::emit): New.
6444 (expand_switch_as_decision_tree_p): Remove.
6445 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
6446 (fix_phi_operands_for_edge): Likewise.
6447 (switch_decision_tree::analyze_switch_statement): New.
6448 (compute_cases_per_edge): Move ...
6449 (switch_decision_tree::compute_cases_per_edge): ... here.
6450 (try_switch_expansion): Likewise.
6451 (switch_decision_tree::try_switch_expansion): Likewise.
6452 (record_phi_operand_mapping): Likewise.
6453 (switch_decision_tree::record_phi_operand_mapping): Likewise.
6454 (emit_case_decision_tree): Likewise.
6455 (switch_decision_tree::emit): Likewise.
6456 (balance_case_nodes): Likewise.
6457 (switch_decision_tree::balance_case_nodes): Likewise.
6458 (dump_case_nodes): Likewise.
6459 (switch_decision_tree::dump_case_nodes): Likewise.
6460 (emit_jump): Likewise.
6461 (switch_decision_tree::emit_jump): Likewise.
6462 (emit_cmp_and_jump_insns): Likewise.
6463 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
6464 (emit_case_nodes): Likewise.
6465 (switch_decision_tree::emit_case_nodes): Likewise.
6466 (conditional_probability): Remove.
6467 * tree-switch-conversion.h (enum cluster_type): New.
6469 (struct cluster): Likewise.
6470 (cluster::cluster): Likewise.
6471 (struct simple_cluster): Likewise.
6472 (simple_cluster::simple_cluster): Likewise.
6473 (struct group_cluster): Likewise.
6474 (struct jump_table_cluster): Likewise.
6475 (struct bit_test_cluster): Likewise.
6476 (struct min_cluster_item): Likewise.
6477 (struct case_tree_node): Likewise.
6478 (case_tree_node::case_tree_node): Likewise.
6479 (jump_table_cluster::case_values_threshold): Likewise.
6480 (struct case_bit_test): Likewise.
6481 (struct switch_decision_tree): Likewise.
6482 (struct switch_conversion): Likewise.
6483 (switch_decision_tree::reset_out_edges_aux): Likewise.
6485 2018-06-20 Martin Liska <mliska@suse.cz>
6487 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
6488 (hoist_edge_and_branch_if_true): Likewise.
6489 (expand_switch_using_bit_tests_p): Likewise.
6490 (struct case_bit_test): Likewise.
6491 (case_bit_test_cmp): Likewise.
6492 (emit_case_bit_tests): Likewise.
6493 (switch_conversion::switch_conversion): New class.
6494 (struct switch_conv_info): Remove old struct.
6495 (collect_switch_conv_info): More to ...
6496 (switch_conversion::collect): ... this.
6497 (check_range): Likewise.
6498 (switch_conversion::check_range): Likewise.
6499 (check_all_empty_except_final): Likewise.
6500 (switch_conversion::check_all_empty_except_final): Likewise.
6501 (check_final_bb): Likewise.
6502 (switch_conversion::check_final_bb): Likewise.
6503 (create_temp_arrays): Likewise.
6504 (switch_conversion::create_temp_arrays): Likewise.
6505 (free_temp_arrays): Likewise.
6506 (gather_default_values): Likewise.
6507 (switch_conversion::gather_default_values): Likewise.
6508 (build_constructors): Likewise.
6509 (switch_conversion::build_constructors): Likewise.
6510 (constructor_contains_same_values_p): Likewise.
6511 (switch_conversion::contains_same_values_p): Likewise.
6512 (array_value_type): Likewise.
6513 (switch_conversion::array_value_type): Likewise.
6514 (build_one_array): Likewise.
6515 (switch_conversion::build_one_array): Likewise.
6516 (build_arrays): Likewise.
6517 (switch_conversion::build_arrays): Likewise.
6518 (gen_def_assigns): Likewise.
6519 (switch_conversion::gen_def_assigns): Likewise.
6520 (prune_bbs): Likewise.
6521 (switch_conversion::prune_bbs): Likewise.
6522 (fix_phi_nodes): Likewise.
6523 (switch_conversion::fix_phi_nodes): Likewise.
6524 (gen_inbound_check): Likewise.
6525 (switch_conversion::gen_inbound_check): Likewise.
6526 (process_switch): Use the newly created class.
6527 (switch_conversion::expand): New.
6528 (switch_conversion::~switch_conversion): New.
6529 * tree-switch-conversion.h: New file.
6531 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6533 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
6534 tree-vect-patterns.c.
6535 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
6536 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
6537 (vect_recog_sad_pattern): Likewise.
6538 (vect_recog_widen_sum_pattern): Likewise.
6539 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
6540 (vect_recog_widen_shift_pattern): Remove the type_in argument.
6541 (vect_recog_rotate_pattern): Likewise.
6542 (vect_recog_mult_pattern): Likewise.
6543 (vect_recog_vector_vector_shift_pattern): Likewise.
6544 (vect_recog_divmod_pattern): Likewise.
6545 (vect_recog_mixed_size_cond_pattern): Likewise.
6546 (vect_recog_bool_pattern): Likewise.
6547 (vect_recog_mask_conversion_pattern): Likewise.
6548 (vect_try_gather_scatter_pattern): Likewise.
6549 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
6550 (vect_recog_over_widening_pattern): Likewise.
6551 (vect_recog_gather_scatter_pattern): Likewise.
6552 (vect_recog_func_ptr): Move from tree-vectorizer.h
6553 (vect_vect_recog_func_ptrs): Move further down the file.
6554 (vect_recog_func): Likewise. Remove the third argument.
6555 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
6556 (vect_pattern_recog_1): Expect the pattern function to do any
6557 necessary target tests. Also expect it to provide a vector type.
6558 Remove the type_in handling.
6560 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6562 * tree-vect-patterns.c (vect_pattern_detected): New function.
6563 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
6564 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
6565 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
6566 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
6567 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6568 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6569 (vect_recog_mask_conversion_pattern)
6570 (vect_try_gather_scatter_pattern): Likewise.
6572 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6574 * tree-vect-patterns.c (vect_get_internal_def): New function.
6575 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6576 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
6577 (search_type_for_mask_1): Use it.
6579 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6581 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6582 redundant WIDEN_SUM_EXPR handling.
6583 (vect_recog_sad_pattern): Likewise.
6585 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6587 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6588 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
6589 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
6590 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
6591 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
6593 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6595 * tree-vect-stmts.c (vectorizable_call): Make sure that we
6596 use the stmt_vec_info of the original bb statement for the
6597 new zero assignment, even if the call is part of a pattern.
6599 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6601 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
6602 that the sequence is attached to the original statement rather
6603 than the pattern statement.
6604 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
6605 PATTERN_DEF_SEQ from the original statement rather than
6606 the main pattern statement.
6607 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
6608 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
6609 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
6611 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6613 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
6614 definition statements before the early exit for statements that aren't
6616 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
6618 (vect_transform_loop): ...here. Process pattern definition
6619 statements without first checking whether the main pattern
6620 statement is live or relevant.
6622 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6624 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
6625 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
6627 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
6629 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
6630 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
6631 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
6632 (expand_block_compare): Change select_block_compare_mode call.
6633 (expand_strncmp_align_check): Use new functions, fix comment.
6634 (emit_final_str_compare_gpr): New function.
6635 (expand_strn_compare): Refactor and clean up code.
6636 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
6638 2018-06-19 Tony Reix <tony.reix@atos.com>
6639 Damien Bergamini <damien.bergamini@atos.com>
6640 David Edelsohn <dje.gcc@gmail.com>
6642 * collect2.c (static_obj): New variable.
6643 (static_libs): New variable.
6644 (is_in_list): Uncomment declaration.
6645 (main): Track AIX libraries linked statically.
6646 (is_in_list): Uncomment definition.
6647 (scan_prog_file): Don't add AIX shared libraries initializer
6648 to constructor list if linking statically.
6650 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
6652 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
6654 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
6656 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
6658 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
6661 2018-06-19 Martin Liska <mliska@suse.cz>
6663 * config/i386/i386.c (ix86_can_inline_p): Do not use
6664 ipa_fn_summaries::get_create.
6665 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
6667 (devirtualization_time_bonus): Likewise.
6668 (ipcp_propagate_stage): Likewise.
6669 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
6670 (edge_set_predicate): Likewise.
6671 (evaluate_conditions_for_known_args): Likewise.
6672 (evaluate_properties_for_edge): Likewise.
6673 (ipa_call_summary::reset): Tranform to ...
6674 (ipa_call_summary::~ipa_call_summary): ... this.
6675 (ipa_fn_summary::reset): Transform to ...
6676 (ipa_fn_summary::~ipa_fn_summary): ... this.
6677 (ipa_fn_summary_t::remove): Rename to ...
6678 (ipa_fn_summary_t::remove_callees): ... this.
6679 (ipa_fn_summary_t::duplicate): Use placement new
6680 instead of memory copy.
6681 (ipa_call_summary_t::duplicate): Likewise.
6682 (ipa_call_summary_t::remove): Remove.
6683 (dump_ipa_call_summary): Change get_create to get.
6684 (ipa_dump_fn_summary): Dump only when summary exists.
6685 (analyze_function_body): Use symbol_summary::get instead
6687 (compute_fn_summary): Likewise.
6688 (estimate_edge_devirt_benefit): Likewise.
6689 (estimate_edge_size_and_time): Likewise.
6690 (inline_update_callee_summaries): Likewise.
6691 (remap_edge_change_prob): Likewise.
6692 (remap_edge_summaries): Likewise.
6693 (ipa_merge_fn_summary_after_inlining): Likewise.
6694 (write_ipa_call_summary): Likewise.
6695 (ipa_fn_summary_write): Likewise.
6696 (ipa_free_fn_summary): Likewise.
6697 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
6698 (struct ipa_call_summary): Likewise.
6699 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
6701 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
6702 (estimate_size_after_inlining): Likewise.
6703 (estimate_growth): Likewise.
6704 (growth_likely_positive): Likewise.
6705 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
6706 (inline_call): Likewise.
6707 * ipa-inline.c (caller_growth_limits): Likewise.
6708 (can_inline_edge_p): Likewise.
6709 (can_inline_edge_by_limits_p): Likewise.
6710 (compute_uninlined_call_time): Likewise.
6711 (compute_inlined_call_time): Likewise.
6712 (want_inline_small_function_p): Likewise.
6713 (edge_badness): Likewise.
6714 (update_caller_keys): Likewise.
6715 (update_callee_keys): Likewise.
6716 (inline_small_functions): Likewise.
6717 (inline_to_all_callers_1): Likewise.
6718 (dump_overall_stats): Likewise.
6719 (early_inline_small_functions): Likewise.
6720 (early_inliner): Likewise.
6721 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
6722 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6723 * ipa-pure-const.c (malloc_candidate_p): Likewise.
6724 * ipa-split.c (execute_split_functions): Likewise.
6725 * symbol-summary.h: Likewise.
6726 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
6728 2018-06-19 Richard Biener <rguenther@suse.de>
6730 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
6731 (vectorize_loops): ... here. Fix dbgcnt handling.
6732 (try_vectorize_loop): Wrap try_vectorize_loop_1.
6734 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
6737 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
6738 ieee128 argument takes up only one (vector) register, not two (floating
6741 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6743 * gimplify.c (gimplify_init_constructor): Really never clear for an
6744 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
6746 2018-06-19 Richard Biener <rguenther@suse.de>
6748 PR tree-optimization/86179
6749 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
6750 after failed recognition.
6752 2018-06-18 Martin Sebor <msebor@redhat.com>
6755 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
6756 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
6757 Handle integer subtraction.
6758 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
6759 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
6761 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6763 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
6764 param from rtx to rtx_insn *.
6765 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
6767 (frv_ifcvt_modify_insn): Likwise.
6768 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
6769 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
6770 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
6771 as_a <rtx_insn *> cast to local "unprotected_region" once
6772 it's been established that it's not NULL or pc_rtx.
6773 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
6774 param "sethi" from rtx to rtx_insn *.
6775 (nds32_group_float_insns): Likewise for param "insn".
6776 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
6778 (vax_output_int_subtract): Likewise.
6779 * config/vax/vax.c (vax_output_int_add): Likewise for param
6781 (vax_output_int_subtract): Likewise.
6782 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
6783 (emit_pattern_after): Likewise for param "after".
6784 (emit_insn_after): Likewise.
6785 (emit_jump_insn_after): Likewise.
6786 (emit_call_insn_after): Likewise.
6787 (emit_debug_insn_after): Likewise.
6788 (emit_pattern_before): Likewise for param "before".
6789 (emit_insn_before): Likewise.
6790 (emit_jump_insn_before): Likewise.
6791 * final.c (get_insn_template): Likewise for param "insn", removing
6793 * output.h (get_insn_template): Likewise for 2nd param.
6794 * rtl.h (emit_insn_before): Likewise.
6795 (emit_jump_insn_before): Likewise.
6796 (emit_debug_insn_before_noloc): Likewise.
6797 (emit_insn_after): Likewise.
6798 (emit_jump_insn_after): Likewise.
6799 (emit_call_insn_after): Likewise.
6800 (emit_debug_insn_after): Likewise.
6801 (set_insn_deleted): Likewise for param.
6803 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
6806 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
6807 floating point modes, so that IFmode is numerically greater than
6808 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
6809 to declare the ordering. This prevents IFmode from being
6810 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
6811 machine. Include rs6000-modes.h to share the fractional values
6812 between genmodes* and the rest of the compiler.
6816 * config/rs6000/rs6000-modes.h: New file.
6817 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
6818 meaning of rs6000_long_double_size so that 126..128 selects an
6819 appropriate 128-bit floating point type.
6820 (rs6000_option_override_internal): Likewise.
6821 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
6822 (TARGET_LONG_DOUBLE_128): Change the meaning of
6823 rs6000_long_double_size so that 126..128 selects an appropriate
6824 128-bit floating point type.
6825 (LONG_DOUBLE_TYPE_SIZE): Update comment.
6826 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
6827 source and destination to match the standard usage.
6828 (truncifkf2): Likewise.
6829 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
6830 ISA 2.07 to use an explicit clobber, instead of passing in a
6832 (copysign<mode>3_soft): Likewise.
6834 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6836 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6837 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
6838 (vect_slp_analyze_instance_dependence): Likewise.
6839 (vect_enhance_data_refs_alignment): Likewise.
6840 (vect_analyze_data_refs_alignment): Likewise.
6841 (vect_slp_analyze_and_verify_instance_alignment
6842 (vect_analyze_data_ref_accesses): Likewise.
6843 (vect_prune_runtime_alias_test_list): Likewise.
6844 (vect_analyze_data_refs): Likewise.
6845 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
6846 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6847 (vect_analyze_scalar_cycles_1): Likewise.
6848 (vect_get_loop_niters): Likewise.
6849 (vect_analyze_loop_form_1): Likewise.
6850 (vect_update_vf_for_slp): Likewise.
6851 (vect_analyze_loop_operations): Likewise.
6852 (vect_analyze_loop): Likewise.
6853 (vectorizable_induction): Likewise.
6854 (vect_transform_loop): Likewise.
6855 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
6856 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6857 (vect_make_slp_decision): Likewise.
6858 (vect_detect_hybrid_slp): Likewise.
6859 (vect_slp_analyze_operations): Likewise.
6860 (vect_slp_bb): Likewise.
6861 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
6862 (vectorizable_bswap): Likewise.
6863 (vectorizable_call): Likewise.
6864 (vectorizable_simd_clone_call): Likewise.
6865 (vectorizable_conversion): Likewise.
6866 (vectorizable_assignment): Likewise.
6867 (vectorizable_shift): Likewise.
6868 (vectorizable_operation): Likewise.
6869 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
6871 2018-06-18 Martin Sebor <msebor@redhat.com>
6873 PR tree-optimization/81384
6874 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
6875 * builtins.c (expand_builtin_strnlen): New function.
6876 (expand_builtin): Call it.
6877 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
6878 * builtins.def (BUILT_IN_STRNLEN): New.
6879 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
6880 Warn for bounds in excess of maximum object size.
6881 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
6882 single-value ranges. Handle strnlen.
6883 (handle_builtin_strlen): Handle strnlen.
6884 (strlen_check_and_optimize_stmt): Same.
6885 * doc/extend.texi (Other Builtins): Document strnlen.
6887 2018-06-18 Maya Rashish <coypu@sdf.org>
6889 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
6890 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
6891 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
6893 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
6895 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
6897 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6899 * tree.c (escaped_string::escape): Replace cast to char * by
6900 const_cast<char *> (unescaped).
6902 2018-06-18 Nick Clifton <nickc@redhat.com>
6905 * tree.c (escaped_string): New class. Converts an unescaped
6906 string into its escaped equivalent.
6907 (warn_deprecated_use): Use the new class to convert the
6908 deprecation message, if present.
6909 (test_escaped_strings): New self test.
6910 (test_c_tests): Add test_escaped_strings.
6911 * doc/extend.texi (deprecated): Add a note that the
6912 deprecation message is affected by the -fmessage-length
6913 option, and that control characters will be escaped.
6914 (#pragma GCC error): Document this pragma.
6915 (#pragma GCC warning): Likewise.
6916 * doc/invoke.texi (-fmessage-length): Document this option's
6917 effect on the #warning and #error preprocessor directives and
6918 the deprecated attribute.
6920 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
6922 * tree.c (decl_value_expr_lookup): Revert latest change.
6923 (decl_value_expr_insert): Likewise.
6925 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
6927 * gimplify.c (nonlocal_vlas): Delete.
6928 (nonlocal_vla_vars): Likewise.
6929 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
6931 (gimplify_body): Do not create and destroy nonlocal_vlas.
6932 * tree-nested.c: Include diagnostic.h.
6933 (use_pointer_in_frame): Tweak.
6934 (lookup_field_for_decl): Add assertion and declare the transformation.
6935 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
6936 internal error when the reference is in a wrong context. Do not
6937 create a debug decl by default.
6938 (note_nonlocal_block_vlas): Delete.
6939 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
6940 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
6941 create a debug decl by default.
6942 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
6943 call is in a wrong context.
6944 (fixup_vla_decls): New function.
6945 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
6946 debug variables were created.
6947 * tree.c (decl_value_expr_lookup): Add checking assertion.
6948 (decl_value_expr_insert): Likewise.
6950 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6953 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
6954 * tree-scalar-evolution.c (interpret_expr): Likewise.
6955 (expression_expensive_p): Likewise.
6956 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
6957 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
6958 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
6959 (ssa_defined_by_minus_one_stmt_p): New.
6961 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6964 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
6965 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
6966 * dojump.c (do_jump): Likewise.
6967 * expr.c (expand_expr_real_2): Check operand type's sign.
6968 * fold-const.c (const_unop): Handle ABSU_EXPR.
6969 (fold_abs_const): Likewise.
6970 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
6971 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
6972 (strip_sign_op_1): Likesise.
6973 * match.pd: Add new pattern to generate ABSU_EXPR.
6974 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
6975 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6976 * tree-eh.c (operation_could_trap_helper_p): Likewise.
6977 * tree-inline.c (estimate_operator_cost): Likewise.
6978 * tree-pretty-print.c (dump_generic_node): Likewise.
6979 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
6980 * tree.def (ABSU_EXPR): New.
6982 2018-06-16 Jakub Jelinek <jakub@redhat.com>
6985 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
6986 documented as preserved for backward compatibility only.
6987 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
6989 PR rtl-optimization/86108
6990 * bb-reorder.c (create_forwarder_block): Renamed to ...
6991 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
6992 jump from new landing pad to the second part.
6993 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
6996 2018-06-15 Jakub Jelinek <jakub@redhat.com>
6999 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
7000 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
7001 Only call store_expr for halves if the mode is the same.
7004 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
7005 Fix up comment formatting.
7007 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
7009 * typed-splay-tree.h (typed_splay_tree::remove): New function.
7010 (typed_splay_tree::closure,
7011 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
7012 (typed_splay_tree::typed_splay_tree,
7013 typed_splay_tree::operator =): Declared private.
7014 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
7015 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
7016 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
7017 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
7018 typed_splay_tree::splay_tree_splay,
7019 typed_splay_tree::splay_tree_foreach_helper,
7020 typed_splay_tree::splay_tree_insert,
7021 typed_splay_tree::splay_tree_remove,
7022 typed_splay_tree::splay_tree_lookup,
7023 typed_splay_tree::splay_tree_predecessor,
7024 typed_splay_tree::splay_tree_successor,
7025 typed_splay_tree::splay_tree_min,
7026 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
7027 (typed_splay_tree::root, typed_splay_tree::comp,
7028 typed_splay_tree::delete_key,
7029 typed_splay_tree::delete_value): New data members.
7030 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
7031 typed_splay_tree::remove.
7033 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
7035 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
7036 -mginv and -mno-ginv to the assembler.
7037 * config/mips/mips.opt (-mcrc): New option.
7039 * doc/invoke.text (-mcrc): Document.
7042 2018-06-15 Nick Clifton <nickc@redhat.com>
7045 * tree.c (escaped_string): New class. Converts an unescaped
7046 string into its escaped equivalent.
7047 (warn_deprecated_use): Use the new class to convert the
7048 deprecation message, if present.
7049 (test_escaped_strings): New self test.
7050 (test_c_tests): Add test_escaped_strings.
7051 * doc/extend.texi (deprecated): Add a note that the
7052 deprecation message is affected by the -fmessage-length
7053 option, and that control characters will be escaped.
7054 (#pragma GCC error): Document this pragma.
7055 (#pragma GCC warning): Likewise.
7056 * doc/invoke.texi (-fmessage-length): Document this option's
7057 effect on the #warning and #error preprocessor directives and
7058 the deprecated attribute.
7060 2018-06-15 Richard Biener <rguenther@suse.de>
7062 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
7063 here, also noting vector size used.
7064 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
7065 size used in MSG_OPTIMIZED_LOCATIONS dump.
7066 (pass_slp_vectorize::execute): Adjust.
7068 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
7071 * config/arc/arc.c (arc_return_address_register): Fix
7074 2018-06-15 Richard Biener <rguenther@suse.de>
7077 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
7078 leave useless conversion stripping to force_gimple_operand_gsi.
7079 (gimplify_build2): Likewise.
7080 (gimplify_build1): Likewise.
7082 2018-06-15 Richard Biener <rguenther@suse.de>
7085 * tree-cfg.c (move_stmt_op): unshare invariant addresses
7086 before adjusting their block.
7088 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
7090 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
7091 multilibs for *-*-rtems*.
7092 * config/riscv/t-rtems: New file.
7094 2018-06-14 Jakub Jelinek <jakub@redhat.com>
7097 * match.pd ((A +- CST1) +- CST2): Punt if last resort
7098 unsigned_type_for returns NULL.
7101 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
7102 subregs of multi-word pseudos unless the float mode has word size.
7104 2018-06-14 Richard Biener <rguenther@suse.de>
7107 * tree-vect-generic.c (build_word_mode_vector_type): Remove
7108 duplicate and harmful type_hash_canon.
7109 * tree.c (type_hash_canon): Assert we didn't find ourselves.
7111 2018-06-14 Richard Biener <rguenther@suse.de>
7114 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
7117 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
7119 * config/rtems.h (STDINT_LONG32): Define.
7121 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
7122 Prachi Godbole <prachi.godbole@imgtec.com>
7124 * config/mips/mips-cpus.def: Define P6600.
7125 * config/mips/mips-tables.opt: Regenerate.
7126 * config/mips/mips.c (mips_ucbranch_type): New enum.
7127 (mips_rtx_cost_data): Add support for P6600.
7128 (mips_issue_rate): Likewise.
7129 (mips_multipass_dfa_lookahead): Likewise.
7130 (mips_avoid_hazard): Likewise.
7131 (mips_reorg_process_insns): Likewise.
7132 (mips_classify_branch_p6600): New function.
7133 * config/mips/mips.h (TUNE_P6600): New define.
7134 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
7135 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
7136 * config/mips/mips.md: Include p6600.md.
7137 (processor): Add p6600.
7138 * config/mips/p6600.md: New file.
7139 * doc/invoke.texi: Add p6600 to supported architectures.
7141 2018-06-13 Martin Sebor <msebor@redhat.com>
7143 PR tree-optimization/86114
7144 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
7146 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
7148 2018-06-13 Richard Biener <rguenther@suse.de>
7150 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
7151 Properly set vector type of the intermediate stmt.
7152 * tree-vect-stmts.c (vectorizable_operation): The destination
7153 var always has vectype_out type.
7155 2018-06-13 Jeff Law <law@redhat.com>
7157 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
7158 integer 0 for argument to print_rtl_with_bb.
7159 (rl78_reorg): Likewise.
7161 2018-06-13 David Malcolm <dmalcolm@redhat.com>
7163 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
7164 from rtx to rtx_insn *.
7165 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
7167 (add_sched_insns_for_speculation): Likewise for local "target",
7168 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
7169 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
7170 from rtx_insn ** to rtx_code_label **.
7171 (reorg_emit_nops): Likewise.
7172 (c6x_reorg): Likewise for local "call_labels".
7173 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
7175 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
7176 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
7177 the loops over LABEL_REFS.
7178 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
7180 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
7181 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
7182 (split_branches): Strengthen local "olabel" from rtx to
7183 rtx_insn *, adding a safe_as_a cast.
7184 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
7186 (add_insn_after): Likewise for first two params.
7187 (add_insn_before): Likewise.
7188 (remove_insn): Likewise for param.
7189 (emit_pattern_before_noloc): Likewise for second and third params.
7190 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
7191 (emit_call_insn_before_noloc): Likewise.
7192 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
7194 (emit_barrier_before): Likewise.
7195 (emit_label_before): Strengthen "label" param from "rtx" to
7196 "rtx_code_label *". Strengthen "before" param from "rtx" to
7198 (emit_insn_after_1): Strengthen "after" param from "rtx" to
7200 (emit_pattern_after_noloc): Likewise.
7201 (emit_insn_after_noloc): Likewise.
7202 (emit_jump_insn_after_noloc): Likewise.
7203 (emit_call_insn_after_noloc): Likewise.
7204 (emit_debug_insn_after_noloc): Likewise.
7205 (emit_barrier_after): Likewise.
7206 (emit_label_after): Likewise for both params.
7207 (emit_pattern_after_setloc): Likewise for "after" param. Convert
7208 "loc" param from "int" to "location_t".
7209 (emit_insn_after_setloc): Likewise.
7210 (emit_jump_insn_after_setloc): Likewise.
7211 (emit_call_insn_after_setloc): Likewise.
7212 (emit_debug_insn_after_setloc): Likewise.
7213 (emit_pattern_before_setloc): Likewise for "before" param. Convert
7214 "loc" param from "int" to "location_t".
7215 (emit_pattern_before): Convert NULL_RTX to NULL.
7216 (emit_insn_before_setloc): Convert "loc" param from "int" to
7218 (emit_jump_insn_before_setloc): Likewise.
7219 (emit_call_insn_before_setloc): Likewise.
7220 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
7221 rtx_insn *. Convert "loc" param from "int" to "location_t".
7222 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
7223 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
7224 Convert 3rd param from "int" to "location_t".
7225 (emit_barrier_before, emit_barrier_after, next_real_insn):
7226 Strengthen param from rtx to rtx_insn *.
7227 (emit_label_before): Strengthen 1st param from "rtx" to
7228 "rtx_code_label *". Strengthen 2nd param from "rtx" to
7230 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
7231 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
7232 Strengthen 2nd param from "rtx" to "rtx_insn *".
7233 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
7234 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
7235 Likewise. Convert 3rd param from "int" to "location_t".
7236 (emit_label_after): Strengthen 1st param from "rtx" to
7238 (next_real_insn, remove_insn): Strengthen param from "rtx" to
7240 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
7241 from "rtx" to "rtx_insn *".
7243 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
7245 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
7246 bodies streamed in with -Q.
7247 * dumpfile.c (dump_files): Add lto-stream-out dump file.
7248 * dumpfile.h (tree_dump_index): Add lto_stream_out.
7249 * gimple-streamer-out.c: Include gimple-pretty-print.h
7250 (output_bb): Dump stmts streamed.
7251 * lto-section-out.c: Include print-tree.h
7252 (lto_begin_section): Dump sections created.
7253 (lto_output_decl_index): Dump decl encoded.
7254 * lto-streamer-out.c: Include print-tree.h
7255 (create_output_block): Dump output block created.
7256 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
7257 (output_function): Dump function output.
7258 (output_constructor): Dump constructor streamed.
7259 (write_global_stream): Output indexes encoded.
7260 (produce_asm_for_decls): Dump streams encoded.
7261 * lto-streamer.c (streamer_dump_file): New global var.
7262 * lto-streamer.h (streamer_dump_file): Declare.
7263 * passes.c (ipa_write_summaries): Initialize streamer dump.
7264 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
7267 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
7270 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
7271 offsets for register save directives. Emit a second batch of save
7272 directives, if need be, when the function accesses prior frames.
7274 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7276 * config/arc/fpu.md (fmasf4): Force operand to register.
7277 (fnmasf4): Likewise.
7279 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7281 * config/arc/arc-protos.h (arc_pad_return): Remove.
7282 * config/arc/arc.c (machine_function): Remove force_short_suffix
7284 (arc_print_operand): Adjust printing of '&'.
7285 (arc_verify_short): Remove conditional printing of short suffix.
7286 (arc_final_prescan_insn): Remove reference to size_reason.
7287 (pad_return): New function.
7288 (arc_reorg): Call pad_return.
7289 (arc_pad_return): Remove.
7290 (arc_init_machine_status): Remove reference to force_short_suffix.
7291 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
7292 (attr length): When attribute iscompact is true force to 2
7293 regardless; in the case of maybe check if we want to force the
7294 instruction to have 4 bytes length.
7295 (nopv): Change it to generate 4 byte long nop as well.
7296 (blockage): New pattern.
7297 (simple_return): Remove call to arc_pad_return.
7298 (p_return_i): Likewise.
7300 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7302 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
7304 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7306 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
7309 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7311 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
7312 for ARC700 and ARCv2.
7314 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
7317 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
7318 operands[2] instead of operands[1].
7321 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
7323 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
7324 case, check whether the outer register overlaps an unallocatable
7325 register, not just whether it fits the required class.
7327 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
7329 * poly-int.h (can_div_trunc_p): Add new overload in which all values
7331 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
7332 (memrefs_conflict_p): Likewise.
7333 (init_alias_analysis): Likewise.
7334 * cfgexpand.c (expand_debug_expr): Likewise.
7335 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
7336 * cse.c (fold_rtx): Likewise.
7337 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
7338 * expr.c (emit_block_move_hints): Likewise.
7339 (clear_storage_hints, push_block, emit_push_insn): Likewise.
7340 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
7341 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
7342 (emit_group_store): Likewise.
7343 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
7344 to read the PRE/POST_MODIFY increment.
7345 * calls.c (store_one_arg): Use strip_offset.
7346 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
7348 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
7350 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
7351 (simplify_binary_operation_1): Extend CONST_INT handling to
7353 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
7354 than a HOST_WIDE_INT.
7355 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
7357 (adjust_mems, add_stores): Update accodingly.
7358 (vt_canonicalize_addr): Track polynomial offsets.
7359 (emit_note_insn_var_location): Likewise.
7360 (vt_add_function_parameter): Likewise.
7361 (vt_initialize): Likewise.
7363 2018-06-12 Jeff Law <law@redhat.com>
7365 * config.gcc (alpha*-*-freebsd*): Remove.
7366 * config/alpha/freebsd.h: Remove.
7368 2018-06-12 David Malcolm <dmalcolm@redhat.com>
7371 * spellcheck-tree.c (levenshtein_distance): Rename to...
7372 (get_edit_distance): ...this, and update for underlying renaming.
7373 * spellcheck-tree.h (levenshtein_distance): Rename to...
7374 (get_edit_distance): ...this.
7375 * spellcheck.c (levenshtein_distance): Rename to...
7376 (get_edit_distance): ...this. Convert from Levenshtein distance
7377 to Damerau-Levenshtein distance by supporting transpositions of
7378 adjacent characters. Rename "v1" to "v_next" and "v0" to
7380 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
7381 (selftest::test_edit_distance_unit_test_oneway): ...this, and
7382 update for underlying renaming.
7383 (selftest::levenshtein_distance_unit_test): Rename to...
7384 (selftest::test_get_edit_distance_unit): ...this, and update for
7385 underlying renaming.
7386 (selftest::test_find_closest_string): Add example from PR 69968
7387 where transposition helps
7388 (selftest::test_metric_conditions): Update for renaming.
7389 (selftest::test_metric_conditions): Likewise.
7390 (selftest::spellcheck_c_tests): Likewise.
7391 * spellcheck.h (levenshtein_distance): Rename both overloads to...
7392 (get_edit_distance): ...this.
7393 (best_match::consider): Update for renaming.
7395 2018-06-12 Martin Sebor <msebor@redhat.com>
7397 PR tree-optimization/85259
7398 * builtins.c (compute_objsize): Handle constant offsets.
7399 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
7400 true iff a warning has been issued.
7401 * gimple.h (gimple_nonartificial_location): New function.
7402 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
7403 gimple_nonartificial_location and handle -Wno-system-headers.
7404 (handle_builtin_stxncpy): Same.
7406 2018-06-12 Martin Sebor <msebor@redhat.com>
7409 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
7411 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
7413 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
7414 BUILTIN_VEC_XST entries for pointer to double and long long.
7416 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
7419 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
7421 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
7424 2018-06-12 Martin Liska <mliska@suse.cz>
7426 * doc/options.texi: Document IntegerRange.
7428 2018-06-12 Martin Liska <mliska@suse.cz>
7430 * config/i386/i386.opt: Make MPX-related options as Deprecated.
7431 * opt-functions.awk: Handle Deprecated flag.
7432 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
7434 (read_cmdline_option): Report warning for a deprecated option.
7435 * opts.h (struct cl_option): Add new field cl_deprecated.
7436 (CL_ERR_DEPRECATED): New.
7438 2018-06-12 Martin Liska <mliska@suse.cz>
7440 * doc/options.texi: Document Deprecated option flag.
7442 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7444 * config/arc/arc-arch.h (arc_extras): New enum.
7445 (arc_cpu_t):Add field extra.
7446 (arc_cpu_types): Consider the extras.
7447 * config/arc/arc-cpus.def: Add extras info.
7448 * config/arc/arc-opts.h (processor_type): Consider extra field.
7449 * config/arc/arc.c (arc_override_options): Handle extra field.
7451 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7453 * config/arc/arc-arch.h: Update ARC_OPTX macro.
7454 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
7456 * config/arc/arc.c (arc_init): Update pic warning.
7457 (irq_range): Update irq range parsing warnings.
7458 (arc_override_options): Update various warning messages.
7459 (arc_handle_aux_attribute): Likewise.
7461 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
7463 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
7465 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7467 * doc/sourcebuild.texi: Document usage of line number 0 in verify
7468 compiler messages directives.
7470 2018-06-12 Matthew Fortune <mfortune@gmail.com>
7472 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
7473 * config/mips/mips-tables.opt: Regenerate.
7474 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
7476 * doc/invoke.texi: Document -march=i6500.
7478 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
7480 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
7481 (i6400_gpmul): Add cpu_unit.
7482 (i6400_gpdiv): Likewise.
7483 (i6400_msa_add_d): Update reservations.
7484 (i6400_msa_int_add) Likewise.
7485 (i6400_msa_short_logic3) Likewise.
7486 (i6400_msa_short_logic2) Likewise.
7487 (i6400_msa_short_logic) Likewise.
7488 (i6400_msa_move) Likewise.
7489 (i6400_msa_cmp) Likewise.
7490 (i6400_msa_short_float2) Likewise.
7491 (i6400_msa_div_d) Likewise.
7492 (i6400_msa_long_logic1) Likewise.
7493 (i6400_msa_long_logic2) Likewise.
7494 (i6400_msa_mult) Likewise.
7495 (i6400_msa_long_float2) Likewise.
7496 (i6400_msa_long_float4) Likewise.
7497 (i6400_msa_long_float5) Likewise.
7498 (i6400_msa_long_float8) Likewise.
7499 (i6400_fpu_fadd): Include frint type.
7500 (i6400_fpu_store): New define_insn_reservation.
7501 (i6400_fpu_load): Likewise.
7502 (i6400_fpu_move): Likewise.
7503 (i6400_fpu_fcmp): Likewise.
7504 (i6400_fpu_fmadd): Likewise.
7505 (i6400_int_mult): Include imul3nc type and update reservation.
7506 (i6400_int_div): Include idiv3 type and update reservation.
7507 (i6400_int_load): Update to check type not move_type.
7508 (i6400_int_store): Likewise.
7509 (i6400_int_prefetch): Set zero latency.
7511 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
7513 * gcc.c: Document new %@{...} sequence.
7514 (LINK_COMMAND_SPEC): Use it for the -L switches.
7515 (cpp_unique_options): Use it for the -I switches.
7516 (at_file_argbuf): New global variable.
7517 (in_at_file): Likewise.
7518 (alloc_args): Create at_file_argbuf.
7519 (clear_args): Truncate at_file_argbuf.
7520 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
7521 (open_at_file): New function.
7522 (close_at_file): Likewise.
7523 (create_at_file): Delete.
7524 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
7527 (validate_switches_from_spec): Deal with %@{...} sequence.
7528 (validate_switches): Likewise.
7529 (driver::finalize): Call clear_args.
7531 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
7533 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
7535 2018-06-11 Martin Sebor <msebor@redhat.com>
7537 * doc/invoke.texi (-Wall): List -Wc++17-compat.
7538 (Wno-class-memaccess): Add @opindex.
7539 (Wno-templates, Wno-multiple-inheritance): Same.
7540 (Wno-virtual-inheritance, Wno-namespaces): Same.
7541 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
7542 (Wno-format-overflow, Wno-format-truncation): Same.
7543 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
7544 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
7545 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
7546 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
7547 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
7548 (Wno-misspelled-isr): Same.
7550 2018-06-11 Martin Sebor <msebor@redhat.com>
7552 * PR tree-optimization/86083
7553 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
7555 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
7557 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
7559 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
7562 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
7563 on the correct operand.
7564 (*movdi_internal64): Ditto.
7566 2018-06-11 Martin Liska <mliska@suse.cz>
7568 PR tree-optimization/86089
7569 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
7571 2018-06-11 Julia Koval <julia.koval@intel.com>
7573 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
7574 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
7575 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
7577 2018-06-11 Olivier Hainque <hainque@adacore.com>
7579 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
7580 for Ada with strict dwarf2.
7582 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
7585 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
7588 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
7590 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
7592 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
7594 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
7597 2018-06-08 Martin Liska <mliska@suse.cz>
7599 * tree-cfg.h (debug_function): Fix argument type to match
7602 2018-06-08 Martin Liska <mliska@suse.cz>
7604 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
7605 Remove usage of MPX-related (and removed) fields.
7606 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
7608 2018-06-08 David Malcolm <dmalcolm@redhat.com>
7610 * cfg.c (debug): Use TDF_NONE rather than 0.
7611 * cfghooks.c (debug): Likewise.
7612 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
7613 (struct dump_option_value_info): Convert to...
7614 (struct kv_pair): ...this template type.
7615 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
7617 (optinfo_verbosity_options): Likewise.
7618 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
7620 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
7621 than int for "optgroup_flags" param.
7622 (dump_generic_expr_loc): Use dump_flags_t rather than int for
7624 (dump_dec): Likewise.
7625 (dump_finish): Use TDF_NONE rather than 0.
7626 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
7627 rather than int for "optgroup_flags" param. Use TDF_NONE rather
7628 than 0. Update for change to option_ptr.
7629 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
7630 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
7631 0. Update for changes to optinfo_verbosity_options and
7633 (opt_info_switch_p): Convert optgroup_flags from int to
7635 (dump_basic_block): Use dump_flags_t rather than int
7636 for "dump_kind" param.
7637 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
7638 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
7639 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
7640 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
7641 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
7642 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
7643 TDF_NONE): Convert from macros to...
7644 (enum dump_flag): ...this new enum.
7645 (dump_flags_t): Update to use enum.
7646 (operator|, operator&, operator~, operator|=, operator&=):
7647 Implement for dump_flags_t.
7648 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
7649 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
7650 Convert from macros to...
7651 (enum optgroup_flag): ...this new enum.
7652 (optgroup_flags_t): New typedef.
7653 (operator|, operator|=): Implement for optgroup_flags_t.
7654 (struct dump_file_info): Convert field "alt_flags" to
7655 dump_flags_t. Convert field "optgroup_flags" to
7657 (dump_basic_block): Use dump_flags_t rather than int for param.
7658 (dump_generic_expr_loc): Likewise.
7659 (dump_dec): Likewise.
7660 (dump_register): Convert param "optgroup_flags" to
7662 (opt_info_enable_passes): Likewise.
7663 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
7665 * gimple-pretty-print.c (debug): Likewise.
7666 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
7667 (merged_store_group::apply_stores): Likewise.
7668 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
7669 * gimple.c (verify_gimple_pp): Likewise.
7670 * graphite-poly.c (print_pbb_body): Likewise.
7671 * passes.c (pass_manager::register_one_dump_file): Convert
7672 local "optgroup_flags" to optgroup_flags_t.
7673 * print-tree.c (print_node): Use TDF_NONE rather than 0.
7675 (debug_body): Likewise.
7676 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
7677 to optgroup_flags_t.
7678 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
7680 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
7681 (convert_mult_to_fma): Likewise.
7682 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
7683 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
7684 * tree-vect-data-refs.c (dump_lower_bound): Convert param
7685 "dump_kind" to dump_flags_t.
7687 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7689 * config/rs6000/rs6000.c (min, max): Delete.
7691 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7693 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
7694 -mabi=spe and -mabi=no-spe.
7696 2018-06-08 Martin Liska <mliska@suse.cz>
7698 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
7699 where we expect an existing summary.
7701 2018-06-08 Martin Liska <mliska@suse.cz>
7703 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
7704 * ipa-inline.h (estimate_edge_growth): Likewise.
7706 2018-06-08 Martin Liska <mliska@suse.cz>
7708 * cgraph.c (function_version_hasher::hash): Use
7709 cgraph_node::get_uid ().
7710 (function_version_hasher::equal):
7711 * cgraph.h (cgraph_node::get_uid): New method.
7712 * ipa-inline.c (update_caller_keys): Use
7713 cgraph_node::get_uid ().
7714 (update_callee_keys): Likewise.
7715 * ipa-utils.c (searchc): Likewise.
7716 (ipa_reduced_postorder): Likewise.
7717 * lto-cgraph.c (input_node): Likewise.
7718 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
7719 * symbol-summary.h (symtab_insertion): Likewise.
7720 (symtab_removal): Likewise.
7721 (symtab_duplication): Likewise.
7722 * tree-pretty-print.c (dump_function_header): Likewise.
7723 * tree-sra.c (convert_callers_for_node): Likewise.
7725 2018-06-08 Martin Liska <mliska@suse.cz>
7727 * cgraph.c (symbol_table::create_edge): Always assign a new
7729 (symbol_table::free_edge): Do not recycle numbers.
7730 * cgraph.h (cgraph_edge::get): New method.
7731 * symbol-summary.h (symtab_removal): Use it.
7732 (symtab_duplication): Likewise.
7733 (call_summary::hashable_uid): Remove.
7735 2018-06-08 Martin Liska <mliska@suse.cz>
7737 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
7738 (initialize_growth_caches): Remove.
7739 (free_growth_caches): Likewise.
7740 (do_estimate_edge_time): Use edge_growth_cache.
7741 (do_estimate_edge_size): Likewise.
7742 (do_estimate_edge_hints): Likewise.
7743 * ipa-inline.c (reset_edge_caches): Likewise.
7744 (recursive_inlining): Likewise.
7745 (inline_small_functions): Likewise.
7746 * ipa-inline.h (initialize_growth_caches): Remove.
7747 (estimate_edge_size): Likewise.
7748 (estimate_edge_time): Likewise.
7749 (estimate_edge_hints): Likewise.
7750 (reset_edge_growth_cache): Likewise.
7751 * symbol-summary.h (call_summary::remove): New method.
7753 2018-06-08 Martin Liska <mliska@suse.cz>
7755 * ipa-cp.c (class edge_clone_summary): New summary.
7756 (grow_edge_clone_vectors): Remove.
7757 (ipcp_edge_duplication_hook): Remove.
7758 (class edge_clone_summary_t): New call_summary class.
7759 (ipcp_edge_removal_hook): Remove.
7760 (edge_clone_summary_t::duplicate): New function.
7761 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
7762 (create_specialized_node): Likewise.
7763 (ipcp_driver): Initialize edge_clone_summaries and do not
7766 2018-06-08 Martin Liska <mliska@suse.cz>
7768 * symbol-summary.h (get): New function.
7769 (call_summary::m_initialize_when_cloning): New class member.
7771 2018-06-08 Martin Liska <mliska@suse.cz>
7773 * cgraph.c (cgraph_node::remove): Do not recycle uid.
7774 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
7775 (symbol_table::allocate_cgraph_symbol): Do not set uid.
7776 * passes.c (uid_hash_t): Record removed_nodes by their uids.
7777 (remove_cgraph_node_from_order): Use the removed_nodes set.
7778 (do_per_function_toporder): Likwise.
7779 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
7780 instead of summary_uid.
7781 (symtab_removal): Likewise.
7782 (symtab_duplication): Likewise.
7784 2018-06-08 Martin Liska <mliska@suse.cz>
7786 * ipa-cp.c (ipcp_store_bits_results): Use
7787 ipcp_transformation_sum.
7788 (ipcp_store_vr_results): Likewise.
7789 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
7791 (ipcp_transformation_initialize): ... this.
7792 (ipa_set_node_agg_value_chain):
7793 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
7794 (write_ipcp_transformation_info): Likewise.
7795 (read_ipcp_transformation_info): Likewise.
7796 (ipcp_update_bits): Likewise.
7797 (ipcp_update_vr): Likewise.
7798 (ipcp_transform_function): Likewise.
7799 * ipa-prop.h: Rename ipcp_transformation_summary to
7800 ipcp_transformation.
7801 (class ipcp_transformation_t): New function summary.
7802 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
7803 (ipa_get_agg_replacements_for_node): Likewise.
7805 2018-06-08 Martin Liska <mliska@suse.cz>
7807 * ipa-pure-const.c (struct funct_state_d): Do it class instead
7809 (class funct_state_summary_t): New function_summary class.
7810 (has_function_state): Remove.
7811 (get_function_state): Likewise.
7812 (set_function_state): Likewise.
7813 (add_new_function): Likewise.
7814 (funct_state_summary_t::insert): New function.
7815 (duplicate_node_data): Remove.
7816 (remove_node_data): Remove.
7817 (funct_state_summary_t::duplicate): New function.
7818 (register_hooks): Create new funct_state_summaries.
7819 (pure_const_generate_summary): Use it.
7820 (pure_const_write_summary): Likewise.
7821 (pure_const_read_summary): Likewise.
7822 (propagate_pure_const): Likewise.
7823 (propagate_nothrow): Likewise.
7824 (dump_malloc_lattice): Likewise.
7825 (propagate_malloc): Likewise.
7826 (execute): Do not register hooks, just remove summary
7828 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
7831 2018-06-08 Martin Liska <mliska@suse.cz>
7833 * ipa-reference.c (remove_node_data): Remove.
7834 (duplicate_node_data): Likewise.
7835 (class ipa_ref_var_info_summary_t): New class.
7836 (class ipa_ref_opt_summary_t): Likewise.
7837 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
7838 (get_reference_optimization_summary): Use
7839 ipa_ref_opt_sum_summaries.
7840 (set_reference_vars_info): Remove.
7841 (set_reference_optimization_summary): Likewise.
7842 (ipa_init): Create summaries.
7843 (init_function_info): Use function summary.
7844 (ipa_ref_opt_summary_t::duplicate): New function.
7845 (ipa_ref_opt_summary_t::remove): New function.
7846 (get_read_write_all_from_node): Fix GNU coding style.
7847 (propagate): Use function summary.
7848 (write_node_summary_p): Fix GNU coding style.
7849 (stream_out_bitmap): Likewise.
7850 (ipa_reference_read_optimization_summary): Use function summary.
7851 (ipa_reference_c_finalize): Do not release hooks.
7853 2018-06-08 Martin Liska <mliska@suse.cz>
7855 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
7856 (analyze_function_body): Extract multiple calls of get_create.
7857 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7858 * ipa-inline.c (recursive_inlining): Use ::get method.
7859 * ipa-inline.h (estimate_edge_growth): Likewise.
7861 2018-06-08 Martin Liska <mliska@suse.cz>
7863 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
7865 (hsa_function_summary::hsa_function_summary): Use the new enum
7867 (hsa_gpu_implementation_p): Use hsa_summaries::get.
7868 * hsa-gen.c (hsa_get_host_function): Likewise.
7869 (get_brig_function_name): Likewise.
7870 * ipa-hsa.c (process_hsa_functions): Likewise.
7871 (ipa_hsa_write_summary): Likewise.
7872 * symbol-summary.h (symtab_duplication): Use ::get function/
7873 (get): New function.
7875 2018-06-08 Martin Liska <mliska@suse.cz>
7877 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
7879 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
7880 (hsa_register_kernel): Likewise.
7881 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
7882 * hsa-gen.c (hsa_get_host_function): Likewise.
7883 (get_brig_function_name): Likewise.
7884 (generate_hsa): Likewise.
7885 (pass_gen_hsail::execute): Likewise.
7886 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
7887 (devirtualization_time_bonus): Likewise.
7888 (ipcp_propagate_stage): Likewise.
7889 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7890 (edge_set_predicate): Likewise.
7891 (evaluate_conditions_for_known_args): Likewise.
7892 (evaluate_properties_for_edge): Likewise.
7893 (ipa_fn_summary::reset): Likewise.
7894 (ipa_fn_summary_t::duplicate): Likewise.
7895 (dump_ipa_call_summary): Likewise.
7896 (ipa_dump_fn_summary): Likewise.
7897 (analyze_function_body): Likewise.
7898 (compute_fn_summary): Likewise.
7899 (estimate_edge_devirt_benefit): Likewise.
7900 (estimate_edge_size_and_time): Likewise.
7901 (estimate_calls_size_and_time): Likewise.
7902 (estimate_node_size_and_time): Likewise.
7903 (inline_update_callee_summaries): Likewise.
7904 (remap_edge_change_prob): Likewise.
7905 (remap_edge_summaries): Likewise.
7906 (ipa_merge_fn_summary_after_inlining): Likewise.
7907 (ipa_update_overall_fn_summary): Likewise.
7908 (read_ipa_call_summary): Likewise.
7909 (inline_read_section): Likewise.
7910 (write_ipa_call_summary): Likewise.
7911 (ipa_fn_summary_write): Likewise.
7912 (ipa_free_fn_summary): Likewise.
7913 * ipa-hsa.c (process_hsa_functions): Likewise.
7914 (ipa_hsa_write_summary): Likewise.
7915 (ipa_hsa_read_section): Likewise.
7916 * ipa-icf.c (sem_function::merge): Likewise.
7917 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7918 (do_estimate_edge_time): Likewise.
7919 (estimate_size_after_inlining): Likewise.
7920 (estimate_growth): Likewise.
7921 (growth_likely_positive): Likewise.
7922 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7923 (inline_call): Likewise.
7924 * ipa-inline.c (caller_growth_limits): Likewise.
7925 (can_inline_edge_p): Likewise.
7926 (can_inline_edge_by_limits_p): Likewise.
7927 (compute_uninlined_call_time): Likewise.
7928 (compute_inlined_call_time): Likewise.
7929 (want_inline_small_function_p): Likewise.
7930 (edge_badness): Likewise.
7931 (update_caller_keys): Likewise.
7932 (update_callee_keys): Likewise.
7933 (recursive_inlining): Likewise.
7934 (inline_small_functions): Likewise.
7935 (inline_to_all_callers_1): Likewise.
7936 (dump_overall_stats): Likewise.
7937 (early_inline_small_functions): Likewise.
7938 (early_inliner): Likewise.
7939 * ipa-inline.h (estimate_edge_growth): Likewise.
7940 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7941 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7942 * ipa-prop.h (IPA_NODE_REF): Likewise.
7943 (IPA_EDGE_REF): Likewise.
7944 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7945 (propagate_malloc): Likewise.
7946 * ipa-split.c (execute_split_functions): Likewise.
7947 * symbol-summary.h: Rename get to get_create.
7949 (get_create): Likewise.
7950 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7952 2018-06-08 Martin Liska <mliska@suse.cz>
7954 * symbol-summary.h (release): Move definition out of class
7956 (symtab_removal): Likewise.
7957 (symtab_duplication): Likewise.
7959 2018-06-08 Martin Liska <mliska@suse.cz>
7961 * symbol-summary.h (function_summary): Move constructor
7962 implementation out of class declaration.
7963 (release): Likewise.
7964 (symtab_insertion): Likewise.
7965 (symtab_removal): Likewise.
7966 (symtab_duplication): Likewise.
7969 2018-06-08 Martin Liska <mliska@suse.cz>
7971 * Makefile.in: Remove support for MPX (macros, related functions,
7972 fields in cgraph_node, ...).
7973 * builtin-types.def (BT_BND): Likewise.
7974 (BT_FN_BND_CONST_PTR): Likewise.
7975 (BT_FN_CONST_PTR_BND): Likewise.
7976 (BT_FN_VOID_PTR_BND): Likewise.
7977 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
7978 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
7979 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
7980 (expand_builtin_mempcpy_with_bounds): Likewise.
7981 (expand_builtin_memset_with_bounds): Likewise.
7982 (expand_builtin_memset_args): Likewise.
7983 (std_expand_builtin_va_start): Likewise.
7984 (expand_builtin): Likewise.
7985 (expand_builtin_with_bounds): Likewise.
7986 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
7987 (DEF_LIB_BUILTIN_CHKP): Likewise.
7988 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
7989 (DEF_CHKP_BUILTIN): Likewise.
7990 (BUILT_IN_MEMCPY): Likewise.
7991 (BUILT_IN_MEMMOVE): Likewise.
7992 (BUILT_IN_MEMPCPY): Likewise.
7993 (BUILT_IN_MEMSET): Likewise.
7994 (BUILT_IN_STPCPY): Likewise.
7995 (BUILT_IN_STRCAT): Likewise.
7996 (BUILT_IN_STRCHR): Likewise.
7997 (BUILT_IN_STRCPY): Likewise.
7998 (BUILT_IN_STRLEN): Likewise.
7999 (BUILT_IN_MEMCPY_CHK): Likewise.
8000 (BUILT_IN_MEMMOVE_CHK): Likewise.
8001 (BUILT_IN_MEMPCPY_CHK): Likewise.
8002 (BUILT_IN_MEMSET_CHK): Likewise.
8003 (BUILT_IN_STPCPY_CHK): Likewise.
8004 (BUILT_IN_STRCAT_CHK): Likewise.
8005 (BUILT_IN_STRCPY_CHK): Likewise.
8006 * calls.c (store_bounds): Likewise.
8007 (emit_call_1): Likewise.
8008 (special_function_p): Likewise.
8009 (maybe_warn_nonstring_arg): Likewise.
8010 (initialize_argument_information): Likewise.
8011 (finalize_must_preallocate): Likewise.
8012 (compute_argument_addresses): Likewise.
8013 (expand_call): Likewise.
8014 * cfgexpand.c (expand_call_stmt): Likewise.
8015 (expand_return): Likewise.
8016 (expand_gimple_stmt_1): Likewise.
8017 (pass_expand::execute): Likewise.
8018 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
8019 (cgraph_node::remove): Likewise.
8020 (cgraph_node::dump): Likewise.
8021 (cgraph_node::verify_node): Likewise.
8022 * cgraph.h (chkp_function_instrumented_p): Likewise.
8023 (symtab_node::get_alias_target): Likewise.
8024 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
8025 (cgraph_local_p): Likewise.
8026 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
8027 (cgraph_edge::rebuild_references): Likewise.
8028 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
8029 (walk_polymorphic_call_targets): Likewise.
8030 (cgraph_node::expand_thunk): Likewise.
8031 (symbol_table::output_weakrefs): Likewise.
8032 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
8033 (ix86_handle_option): Likewise.
8034 * config/i386/constraints.md: Likewise.
8035 * config/i386/i386-builtin-types.def (BND): Likewise.
8039 * config/i386/i386-builtin.def (BDESC_END): Likewise.
8040 (BDESC_FIRST): Likewise.
8042 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
8043 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
8044 * config/i386/i386.c (enum reg_class): Likewise.
8045 (ix86_target_string): Likewise.
8046 (ix86_option_override_internal): Likewise.
8047 (ix86_conditional_register_usage): Likewise.
8048 (ix86_valid_target_attribute_inner_p): Likewise.
8049 (ix86_set_indirect_branch_type): Likewise.
8050 (ix86_set_current_function): Likewise.
8051 (ix86_function_arg_regno_p): Likewise.
8052 (init_cumulative_args): Likewise.
8053 (ix86_function_arg_advance): Likewise.
8054 (ix86_function_arg): Likewise.
8055 (ix86_pass_by_reference): Likewise.
8056 (ix86_function_value_regno_p): Likewise.
8057 (ix86_function_value_1): Likewise.
8058 (ix86_function_value_bounds): Likewise.
8059 (ix86_return_in_memory): Likewise.
8060 (ix86_setup_incoming_vararg_bounds): Likewise.
8061 (ix86_va_start): Likewise.
8062 (indirect_thunk_need_prefix): Likewise.
8063 (print_reg): Likewise.
8064 (ix86_print_operand): Likewise.
8065 (ix86_expand_call): Likewise.
8066 (ix86_output_function_return): Likewise.
8067 (reg_encoded_number): Likewise.
8068 (BDESC_VERIFYS): Likewise.
8069 (ix86_init_mpx_builtins): Likewise.
8070 (ix86_init_builtins): Likewise.
8071 (ix86_emit_cmove): Likewise.
8072 (ix86_emit_move_max): Likewise.
8073 (ix86_expand_builtin): Likewise.
8074 (ix86_builtin_mpx_function): Likewise.
8075 (ix86_get_arg_address_for_bt): Likewise.
8076 (ix86_load_bounds): Likewise.
8077 (ix86_store_bounds): Likewise.
8078 (ix86_load_returned_bounds): Likewise.
8079 (ix86_store_returned_bounds): Likewise.
8080 (ix86_class_likely_spilled_p): Likewise.
8081 (ix86_hard_regno_mode_ok): Likewise.
8082 (x86_order_regs_for_local_alloc): Likewise.
8083 (ix86_mitigate_rop): Likewise.
8084 (ix86_bnd_prefixed_insn_p): Likewise.
8085 (ix86_mpx_bound_mode): Likewise.
8086 (ix86_make_bounds_constant): Likewise.
8087 (ix86_initialize_bounds): Likewise.
8088 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
8089 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
8090 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
8091 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
8092 (TARGET_CHKP_BOUND_MODE): Likewise.
8093 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
8094 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
8095 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
8096 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
8097 * config/i386/i386.h (TARGET_MPX): Likewise.
8098 (TARGET_MPX_P): Likewise.
8099 (VALID_BND_REG_MODE): Likewise.
8100 (FIRST_BND_REG): Likewise.
8101 (LAST_BND_REG): Likewise.
8102 (enum reg_class): Likewise.
8103 (BND_REG_P): Likewise.
8104 (BND_REGNO_P): Likewise.
8105 (BNDmode): Likewise.
8106 (ADJUST_INSN_LENGTH): Likewise.
8107 * config/i386/i386.md: Likewise.
8108 * config/i386/i386.opt: Likewise.
8109 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
8110 (defined): Likewise.
8111 (LINK_MPX): Likewise.
8112 (MPX_SPEC): Likewise.
8113 (LIBMPX_SPEC): Likewise.
8114 (LIBMPXWRAPPERS_SPEC): Likewise.
8115 (CHKP_SPEC): Likewise.
8116 * config/i386/predicates.md: Likewise.
8117 * dbxout.c (dbxout_type): Likewise.
8118 * doc/extend.texi: Likewise.
8119 * doc/invoke.texi: Likewise.
8120 * doc/md.texi: Likewise.
8121 * doc/tm.texi: Likewise.
8122 * doc/tm.texi.in: Likewise.
8123 * dwarf2out.c (is_base_type): Likewise.
8124 (gen_formal_types_die): Likewise.
8125 (gen_subprogram_die): Likewise.
8126 (gen_type_die_with_usage): Likewise.
8127 (gen_decl_die): Likewise.
8128 (dwarf2out_late_global_decl): Likewise.
8129 * expr.c (expand_assignment): Likewise.
8130 (emit_storent_insn): Likewise.
8131 (store_expr_with_bounds): Likewise.
8132 (store_expr): Likewise.
8133 (expand_expr_real_1): Likewise.
8134 * expr.h (store_expr_with_bounds): Likewise.
8135 * function.c (use_register_for_decl): Likewise.
8136 (struct bounds_parm_data): Likewise.
8137 (assign_parms_augmented_arg_list): Likewise.
8138 (assign_parm_find_entry_rtl): Likewise.
8139 (assign_parm_is_stack_parm): Likewise.
8140 (assign_parm_load_bounds): Likewise.
8141 (assign_bounds): Likewise.
8142 (assign_parms): Likewise.
8143 (expand_function_start): Likewise.
8144 * gcc.c (CHKP_SPEC): Likewise.
8145 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
8146 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
8147 (wrestrict_dom_walker::check_call): Likewise.
8148 * gimple.c (gimple_build_call_from_tree): Likewise.
8149 * gimple.h (enum gf_mask): Likewise.
8150 (gimple_call_with_bounds_p): Likewise.
8151 (gimple_call_set_with_bounds): Likewise.
8152 * gimplify.c (gimplify_init_constructor): Likewise.
8153 * ipa-cp.c (initialize_node_lattices): Likewise.
8154 (propagate_constants_across_call): Likewise.
8155 (find_more_scalar_values_for_callers_subset): Likewise.
8156 * ipa-hsa.c (process_hsa_functions): Likewise.
8157 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
8158 * ipa-icf.c (sem_function::merge): Likewise.
8159 * ipa-inline.c (early_inliner): Likewise.
8160 * ipa-pure-const.c (warn_function_noreturn): Likewise.
8161 (warn_function_cold): Likewise.
8162 (propagate_pure_const): Likewise.
8163 * ipa-ref.h (enum GTY): Likewise.
8164 * ipa-split.c (find_retbnd): Likewise.
8165 (consider_split): Likewise.
8166 (split_function): Likewise.
8167 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8168 * ipa.c (walk_polymorphic_call_targets): Likewise.
8169 (symbol_table::remove_unreachable_nodes): Likewise.
8170 (process_references): Likewise.
8171 (cgraph_build_static_cdtor_1): Likewise.
8172 * lto-cgraph.c (lto_output_node): Likewise.
8173 (output_refs): Likewise.
8174 (compute_ltrans_boundary): Likewise.
8175 (input_overwrite_node): Likewise.
8176 (input_node): Likewise.
8177 (input_cgraph_1): Likewise.
8178 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
8179 * passes.c (pass_manager::execute_early_local_passes): Likewise.
8180 (class pass_chkp_instrumentation_passes): Likewise.
8181 (make_pass_chkp_instrumentation_passes): Likewise.
8182 * passes.def: Likewise.
8183 * rtl.h (struct GTY): Likewise.
8184 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
8185 * stor-layout.c (layout_type): Likewise.
8186 * symtab.c: Likewise.
8187 * target.def: Likewise.
8188 * targhooks.c (default_chkp_bound_type): Likewise.
8189 (default_chkp_bound_mode): Likewise.
8190 (default_builtin_chkp_function): Likewise.
8191 (default_chkp_function_value_bounds): Likewise.
8192 (default_chkp_make_bounds_constant): Likewise.
8193 (default_chkp_initialize_bounds): Likewise.
8194 * targhooks.h (default_chkp_bound_type): Likewise.
8195 (default_chkp_bound_mode): Likewise.
8196 (default_builtin_chkp_function): Likewise.
8197 (default_chkp_function_value_bounds): Likewise.
8198 (default_chkp_make_bounds_constant): Likewise.
8199 (default_chkp_initialize_bounds): Likewise.
8200 * toplev.c (compile_file): Likewise.
8201 (process_options): Likewise.
8202 * tree-core.h (DEF_BUILTIN): Likewise.
8203 (DEF_BUILTIN_CHKP): Likewise.
8204 * tree-inline.c (declare_return_variable): Likewise.
8205 (remap_gimple_stmt): Likewise.
8206 (copy_bb): Likewise.
8207 (initialize_inlined_parameters): Likewise.
8208 (expand_call_inline): Likewise.
8209 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
8210 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
8211 (make_pass_ipa_chkp_produce_thunks): Likewise.
8212 (make_pass_chkp): Likewise.
8213 (make_pass_chkp_opt): Likewise.
8214 (make_pass_chkp_instrumentation_passes): Likewise.
8215 * tree-pretty-print.c (dump_generic_node): Likewise.
8216 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
8217 * tree-ssa-dce.c (propagate_necessity): Likewise.
8218 (eliminate_unnecessary_stmts): Likewise.
8219 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
8220 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
8221 * tree-ssa-sccvn.h: Likewise.
8222 * tree-ssa-strlen.c (get_string_length): Likewise.
8223 (valid_builtin_call): Likewise.
8224 (adjust_last_stmt): Likewise.
8225 (handle_builtin_strchr): Likewise.
8226 (handle_builtin_strcpy): Likewise.
8227 (handle_builtin_stxncpy): Likewise.
8228 (handle_builtin_memcpy): Likewise.
8229 (handle_builtin_strcat): Likewise.
8230 (strlen_check_and_optimize_stmt): Likewise.
8231 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
8232 * tree-streamer-in.c: Likewise.
8233 * tree-streamer.c (record_common_node): Likewise.
8234 * tree.c (tree_code_size): Likewise.
8235 (wide_int_to_tree_1): Likewise.
8236 (type_contains_placeholder_1): Likewise.
8237 (build_common_tree_nodes): Likewise.
8238 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
8239 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
8240 (POINTER_BOUNDS_P): Likewise.
8241 (BOUNDED_TYPE_P): Likewise.
8242 (BOUNDED_P): Likewise.
8243 (CALL_WITH_BOUNDS_P): Likewise.
8244 (pointer_bounds_type_node): Likewise.
8245 * value-prof.c (gimple_ic): Likewise.
8246 * var-tracking.c (vt_add_function_parameters): Likewise.
8247 * varasm.c (make_decl_rtl): Likewise.
8248 (assemble_start_function): Likewise.
8249 (output_constant): Likewise.
8250 (maybe_assemble_visibility): Likewise.
8251 * varpool.c (ctor_for_folding): Likewise.
8252 * chkp-builtins.def: Remove.
8253 * ipa-chkp.c: Remove.
8254 * ipa-chkp.h: Remove.
8255 * rtl-chkp.c: Remove.
8256 * rtl-chkp.h: Remove.
8257 * tree-chkp-opt.c: Remove.
8258 * tree-chkp.c: Remove.
8259 * tree-chkp.h: Remove.
8261 2018-06-07 Carl Love <cel@us.ibm.com>
8263 * config/rs6000/vsx.md (vextract_fp_from_shorth,
8264 vextract_fp_from_shortl): Add BE support.
8266 2018-06-07 Paul Koning <ni1d@arrl.net>
8268 * compare-elim.c (try_merge_compare): Don't merge compare if
8269 address contains a side effect.
8270 (try_eliminate_compare): Likewise.
8272 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
8274 * config.gcc: Support "tremont".
8275 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
8276 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8278 * config/i386/i386.c (m_TREMONT): Define.
8279 (processor_target_table): Add "tremont".
8280 (PTA_TREMONT): Define.
8281 (ix86_lea_outperforms): Add TARGET_TREMONT.
8282 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
8283 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
8284 and M_INTEL_GOLDMONT_PLUS.
8285 (fold_builtin_cpu): Add "tremont".
8286 (ix86_add_stmt_cost): Add TARGET_TREMONT.
8287 (ix86_option_override_internal): Add "tremont".
8288 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
8289 (processor_type): Add PROCESSOR_TREMONT.
8290 * config/i386/x86-tune.def: Add m_TREMONT.
8291 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
8293 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8295 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
8296 symbol defined for msp430i* devices to be lower case.
8298 2018-06-07 Richard Biener <rguenther@suse.de>
8300 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
8301 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
8302 Properly wrap signed arithmetic if overflow wraps.
8304 2018-06-07 Jakub Jelinek <jakub@redhat.com>
8306 PR tree-optimization/69615
8307 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
8308 of a cast from a same precision integral SSA_NAME in a bb dominated
8309 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
8310 cast to utype if rhs2 has already a compatible type.
8312 2018-06-07 Richard Biener <rguenther@suse.de>
8314 PR tree-optimization/85935
8315 * graphite-scop-detection.c (find_params_in_bb): Analyze
8316 condition operands with respect to the correct loop. Assert
8317 the analysis doesn't fail.
8319 2018-06-04 Carl Love <cel@us.ibm.com>
8321 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
8322 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
8323 as it is slightly cheaper.
8324 (first_match_or_eos_index_<mode>):
8325 Calculate index using natural element order.
8326 (first_match_index_<mode>):
8327 Calculate index using natural element order.
8328 (first_match_or_eos_index_<mode>):
8329 Calculate index using natural order.
8330 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
8331 for BE and LE modes.
8332 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
8333 P9V_BUILTIN_VCLZLSBB_V16QI.
8334 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
8337 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
8339 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
8340 indentation and line wrap for many prototypes. Add missing
8341 @smallexample directives around block of prototypes for vec_xl and
8344 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
8346 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
8347 track if we pass or return IEEE 128-bit floating point.
8348 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
8349 C++ mangling that is compatible with GCC 8.1.
8350 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
8351 (init_cumulative_args): Note if we pass or return IEEE 128-bit
8352 floating point types.
8353 (rs6000_function_arg_advance_1): Likewise.
8354 (rs6000_mangle_type): Optionally generate mangled names that match
8355 what GCC 8.1 generated for IEEE 128-bit floating point types.
8356 (rs6000_globalize_decl_name): If we have an external function that
8357 passes or returns IEEE 128-bit types, generate a weak reference
8358 from the mangled name used in GCC 8.1 to the current mangled
8360 (rs6000_init_builtins): Make __ibm128 use the long double type if
8361 long double is IBM extended double. Make __float128 use the long
8362 double type if long double is IEEE 128-bit.
8365 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
8366 macro for __ibm128 built-in functions.
8367 (PACK_IF): Add __ibm128 pack/unpack functions.
8368 (UNPACK_IF): Likewise.
8369 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
8370 enable long double built-in functions if long double is IEEE
8371 128-bit floating point.
8372 (rs6000_invalid_builtin): Update long double built-in function
8374 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
8375 functions, adjust the built-in function to use the long double
8376 built-in function if __ibm128 and long double are the same type.
8377 * doc/extend.texi (PowerPC builtins): Update documention for
8378 __builtin_{,un}pack_longdouble. Add documentation for
8379 __builtin_{,un}pack_ibm128.
8381 2018-06-06 Jim Wilson <jimw@sifive.com>
8383 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
8384 (struct machine_function): New field interrupt_mode.
8385 (riscv_handle_type_attribute): New function. Add forward declaration.
8386 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
8387 (riscv_expand_epilogue): Check interrupt_mode field.
8388 (riscv_set_current_function): Check interrupt attribute args and
8389 set interrupt_mode field.
8390 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
8391 (riscv_sret, riscv_uret): New.
8392 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
8393 new arguments to interrupt attribute.
8395 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
8398 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
8399 Don't handle -mcpu=power8 if -mpower9-vector is also used.
8401 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
8403 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
8404 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
8405 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8406 several redundant entries.
8408 2018-06-06 David Malcolm <dmalcolm@redhat.com>
8410 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
8411 type from "rtx" to "rtx_insn *".
8412 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
8413 for local "call_insn", removing cast.
8414 (ix86_expand_call): Likewise, introducing a "call_insn" local.
8416 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
8418 PR tree-optimization/86066
8419 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
8420 for BIT_INSERT_EXPR stores.
8422 2018-06-06 Richard Biener <rguenther@suse.de>
8424 PR tree-optimization/86062
8425 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
8426 component refs ontop
8427 of to be offsetted base.
8429 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8431 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
8432 to be static and remove check on interrupt attribute name.
8434 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
8436 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8437 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
8439 2018-06-05 Steve Ellcey <sellcey@cavium.com>
8442 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
8444 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
8445 Remove second argument, change how error is called.
8446 (aarch64_layout_arg): Remove second argument from
8447 aarch64_err_no_fpadvsimd call.
8448 (aarch64_init_cumulative_args): Ditto.
8449 (aarch64_gimplify_va_arg_expr): Ditto.
8450 * config/aarch64/aarch64.md (mov<mode>): Ditto.
8452 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
8454 * config/i386/i386.md (simple_return_indirect_internal): New expander.
8455 (*simple_return_indirect_internal<mode>): Rename from
8456 simple_return_indirect_internal. Use W mode iterator.
8457 (rstorssp): New expander.
8458 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
8459 (clrssbsy): New expander.
8460 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
8462 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
8464 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
8466 (cmse_check_pointed_object): Likewise.
8468 2018-06-05 Martin Liska <mliska@suse.cz>
8470 PR gcov-profile/47618
8471 * doc/invoke.texi: Document how -fprofile-dir format
8474 2018-06-05 Richard Biener <rguenther@suse.de>
8476 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
8477 removal pretend DOM info isn't available so we do not update
8478 it and only remove edges, not dominated blocks. Actually free
8479 DOM info in case we removed something. Remove unreachable blocks.
8480 (mfb_keep_latches): Work with either DOM info or marked backedges.
8481 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
8482 first. Mark backedges if DOM info isn't available.
8483 (Re-)compute DOM info after cleanup_control_flow_pre.
8485 2018-06-05 Richard Biener <rguenther@suse.de>
8487 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
8488 (locus_discrim_hasher::hash): Adjust.
8489 (locus_discrim_hasher::equal): Likewise.
8490 (next_discriminator_for_locus): Work on line directly.
8491 (same_line_p): Pass in expanded locus1 as well.
8492 (assign_discriminators): Avoid redundant location expansions.
8494 2018-06-05 Richard Biener <rguenther@suse.de>
8496 PR tree-optimization/86046
8497 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
8498 if required after clearing TREE_ADDRESSABLE.
8500 2018-06-05 Richard Biener <rguenther@suse.de>
8502 PR tree-optimization/86047
8503 * tree-ssa-loop.c (for_each_index): Glob handling of all
8504 decls and constants and really handle all of them.
8506 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8509 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
8510 qualifier_void_pointer and qualifier_const_void_pointer.
8511 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
8512 (arm_init_builtins): Handle the above.
8513 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
8514 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
8515 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
8518 2018-06-05 Martin Liska <mliska@suse.cz>
8520 * auto-profile.c (read_autofdo_file): Do not use
8521 gcov_ctr_summary struct.
8522 (afdo_callsite_hot_enough_for_early_inline): Likewise.
8523 * coverage.c (struct counts_entry): Likewise.
8524 (read_counts_file): Read just single summary entry.
8525 (get_coverage_counts): Use gcov_summary struct.
8526 * coverage.h (get_coverage_counts): Likewise.
8527 * gcov-dump.c (dump_working_sets): Likewise.
8528 (tag_summary): Dump just single summary.
8529 * gcov-io.c (gcov_write_summary): Write just histogram
8531 (gcov_read_summary): Read just single summary.
8532 (compute_working_sets): Use gcov_summary struct.
8533 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
8534 of GCOV_COUNTERS_SUMMABLE.
8535 (GCOV_COUNTERS_SUMMABLE): Remove.
8536 (GCOV_FIRST_VALUE_COUNTER): Replace with
8537 GCOV_COUNTER_V_INTERVAL.
8538 (struct gcov_ctr_summary): Remove.
8539 (struct gcov_summary): Directly use fields of former
8541 (compute_working_sets): Use gcov_summary struct.
8542 * gcov.c (read_count_file): Do not use ctrs fields.
8543 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
8545 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
8547 * profile.c: Likewise.
8548 * profile.h: Likewise.
8550 2018-06-05 Martin Liska <mliska@suse.cz>
8552 PR gcov-profile/84846
8553 * gcov.c (output_lines): Print working directory only
8554 in intermediate format.
8556 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
8558 * config/s390/s390-builtin-types.def: Add void function type.
8559 * config/s390/s390-builtins.def: Use the function type for the
8562 2018-06-04 Jim Wilson <jimw@sifive.com>
8564 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
8566 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
8567 and maybe_eh_return. Change regno to unsigned int. Use new args to
8568 handle EH_RETURN_DATA_REGNO registers properly.
8569 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
8570 (riscv_expand_epilogue): Update comment. Change argument name and
8571 type. Update code to use new name and type. Pass new args to
8572 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
8574 * config/riscv/riscv.md (NORMAL_RETURN): New.
8575 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
8576 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
8577 (eh_return): Call gen_eh_return_internal and emit barrier.
8578 (eh_return_internal): Call riscv_expand_epilogue.
8580 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
8582 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
8583 bit_insertion field and declare can_be_merged_into method.
8584 (merged_store_group::can_be_merged_into): New method.
8585 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
8586 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
8587 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
8589 2018-06-04 Richard Biener <rguenther@suse.de>
8591 PR tree-optimization/85955
8592 * builtins.c (fold_builtin_sincos): Convert pointers to
8593 destination to appropriate type before dereferencing.
8595 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
8597 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
8599 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
8601 * expr.c (expand_expr_real_1): Force the operand into memory if
8602 its TYPE_MODE is BLKmode and if there is no integer mode for
8603 the number of bits being extracted.
8605 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8609 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
8610 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
8612 2018-06-04 Richard Biener <rguenther@suse.de>
8614 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
8615 (cleanup_tree_cfg_noloop): ... single caller. Do
8616 start_recording_case_labels later.
8618 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
8620 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
8621 to _IMMINTRIN_H_INCLUDED.
8622 * config/i386/pconfigintrin.h: Ditto.
8623 * config/i386/waitpkgintrin.h: Ditto.
8624 * config/i386/immintrin.h: Add includes for sgxintrin.h,
8625 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
8626 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
8627 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
8628 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
8629 waitpkgintrin.h and cldemoteintrin.h.
8631 2018-06-04 Richard Biener <rguenther@suse.de>
8633 PR tree-optimization/86038
8634 * tracer.c (find_best_successor): Check probability for
8635 being initialized, bail out if not.
8637 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
8640 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
8641 of bits to ignore when comparing architectures.
8643 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8645 PR tree-optimization/69615
8646 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
8647 maximum or minimum of the type, try to merge it also as if
8648 range1 is + [-, x - 1] or + [x + 1, -].
8651 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
8653 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
8655 PR tree-optimization/86034
8656 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
8657 the unsigned bitfield type in a bit insertion sequence if it does not
8658 have a larger precision than the bitfield size.
8659 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
8661 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
8663 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
8665 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8667 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
8668 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
8669 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
8670 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
8672 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8674 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
8675 Disable -fdelete-null-pointer-checks for ELF toolchain.
8677 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8678 Kito Cheng <kito.cheng@gmail.com>
8680 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
8681 (nds32le-*-*, nds32be-*-*): Integrate checking process.
8682 (nds32*-*-*): Add glibc and uclibc conditions.
8683 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
8684 (TARGET_EXCEPT_UNWIND_INFO): Define.
8685 * config/nds32/elf.h: New file.
8686 * config/nds32/linux.h: New file.
8687 * config/nds32/nds32-elf.opt: New file.
8688 * config/nds32/nds32-linux.opt: New file.
8689 * config/nds32/nds32-fp-as-gp.c
8690 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
8691 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
8693 (nds32_asm_file_end): Ditto.
8694 (nds32_print_operand): Ditto.
8695 (nds32_insert_attributes): Ditto.
8696 (nds32_init_libfuncs): New function.
8697 (TARGET_HAVE_TLS): Define.
8698 (TARGET_INIT_LIBFUNCS): Define.
8699 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
8701 (TARGET_ELF): Apply different mcmodel setting.
8702 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
8703 been migrated into elf.h and linux.h files.
8704 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
8705 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
8706 (mcmodel): The content has been migrated into nds32-elf.opt and
8707 nds32-linux.opt files.
8708 * config/nds32/t-elf: New file.
8709 * config/nds32/t-linux: New file.
8711 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8712 Shiva Chen <shiva0217@gmail.com>
8714 * config/nds32/constants.md (unspec_volatile_element): Add
8715 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
8716 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
8718 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
8719 (make_pass_nds32_fp_as_gp): Declare.
8720 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
8722 (nds32_asm_function_end_prologue): Remove unused asm output.
8723 (nds32_asm_function_begin_epilogue): Remove unused asm output.
8724 (nds32_asm_file_start): Output necessary fp_as_gp information.
8725 (nds32_option_override): Adjust register usage.
8726 (nds32_expand_prologue): Consider fp_as_gp situation.
8727 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
8728 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
8731 (simple_return): Ditto.
8732 (omit_fp_begin): Output special directive for fp_as_gp.
8733 (omit_fp_end): Output special directive for fp_as_gp.
8734 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
8735 mforbid-fp-as-gp): New options.
8737 2018-06-01 Mark Wielaard <mark@klomp.org>
8739 * dwarf2out.c (dwarf2out_finish): Remove generation of
8740 DW_AT_loclists_base.
8742 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
8744 * gimple-ssa-store-merging.c: Include gimple-fold.h.
8745 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
8746 (struct merged_store_group): Add bit_insertion field.
8747 (dump_char_array): Use standard hexadecimal format.
8748 (merged_store_group::merged_store_group): Set bit_insertion to false.
8749 (merged_store_group::apply_stores): Use optimal buffer size. Deal
8750 with BIT_INSERT_EXPR stores. Move up code updating the mask and
8751 also print the mask in the dump file.
8752 (pass_store_merging::gate): Minor tweak.
8753 (imm_store_chain_info::coalesce_immediate): Fix wrong association
8754 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
8755 stores with INTEGER_CST stores.
8756 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
8757 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
8758 and use it throughout. Generate bit insertion sequences if need be.
8759 (pass_store_merging::process_store): Remove redundant condition.
8760 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
8762 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
8764 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
8765 the 128-bit floating point types. Fix function comment.
8767 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8769 * config/aarch64/aarch64-simd.md
8770 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
8772 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
8775 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8777 PR tree-optimization/85989
8778 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
8780 (backprop::intersect_uses): Check it when deciding whether this
8781 is a backedge reference.
8782 (backprop::process_block): Add each phi to m_visited_phis
8783 after visiting it, then clear it at the end.
8785 2018-06-01 Richard Biener <rguenther@suse.de>
8787 * tree-vectorizer.h (vect_dr_stmt): New function.
8788 (vect_get_load_cost): Adjust.
8789 (vect_get_store_cost): Likewise.
8790 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
8791 Use vect_dr_stmt instead of DR_SMTT.
8792 (vect_record_base_alignments): Likewise.
8793 (vect_calculate_target_alignment): Likewise.
8794 (vect_compute_data_ref_alignment): Likewise and make static.
8795 (vect_update_misalignment_for_peel): Likewise.
8796 (vect_verify_datarefs_alignment): Likewise.
8797 (vector_alignment_reachable_p): Likewise.
8798 (vect_get_data_access_cost): Likewise. Pass down
8799 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
8800 (vect_get_peeling_costs_all_drs): Likewise.
8801 (vect_peeling_hash_get_lowest_cost): Likewise.
8802 (vect_enhance_data_refs_alignment): Likewise.
8803 (vect_find_same_alignment_drs): Likewise.
8804 (vect_analyze_data_refs_alignment): Likewise.
8805 (vect_analyze_group_access_1): Likewise.
8806 (vect_analyze_group_access): Likewise.
8807 (vect_analyze_data_ref_access): Likewise.
8808 (vect_analyze_data_ref_accesses): Likewise.
8809 (vect_vfa_segment_size): Likewise.
8810 (vect_small_gap_p): Likewise.
8811 (vectorizable_with_step_bound_p): Likewise.
8812 (vect_prune_runtime_alias_test_list): Likewise.
8813 (vect_analyze_data_refs): Likewise.
8814 (vect_supportable_dr_alignment): Likewise.
8815 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8816 (vect_gen_prolog_loop_niters): Likewise.
8817 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8818 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
8820 (vect_recog_mask_conversion_pattern): Likewise.
8821 (vect_try_gather_scatter_pattern): Likewise.
8822 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
8823 to vect_get_store_cost.
8824 (vect_get_store_cost): Get stmt_info instead of DR.
8825 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
8826 (vect_get_load_cost): Get stmt_info instead of DR.
8828 2018-06-01 Richard Biener <rguenther@suse.de>
8831 * gimple-fold.c (var_decl_component_p): Also allow offsetted
8832 vars wrapped in MEM_REFs.
8834 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8836 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8837 Fix subreg tests so that we only return a choice between
8838 GENERAL_REGS and FP_REGS if the original classes included both.
8840 2018-06-01 Richard Biener <rguenther@suse.de>
8843 * tree-ssa-structalias.c (get_function_part_constraint):
8844 Handle NULL fi->decl.
8845 (find_func_aliases_for_call): Properly handle indirect
8846 fi from direct call.
8847 (find_func_clobbers): Likewise.
8848 (ipa_pta_execute): Likewise.
8849 (create_variable_info_for): For functions that are ifunc_resolver
8850 resolve to a varinfo that contains the result of the resolver call.
8851 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
8854 2018-05-31 Michael Collison <michael.collison@arm.com>
8856 * config/aarch64/aarch64.md:
8857 (*fix_to_zero_extenddfdi2): New pattern.
8858 * gcc.target/aarch64/fix_extend1.c: New testcase.
8860 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
8864 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
8865 and BUILT_IN_STRNCMP_EQ.
8866 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
8867 BUILT_IN_STRNCMP_EQ.
8868 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
8869 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8870 (gimple_fold_builtin): Likewise.
8871 * tree-ssa-strlen.c (compute_string_length): New function.
8872 (determine_min_obsize): New function.
8873 (handle_builtin_string_cmp): New function to handle calls to
8874 string compare functions.
8875 (strlen_optimize_stmt): Add handling to builtin string compare
8877 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
8878 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8879 * tree.c (build_common_builtin_nodes): Add new defines of
8880 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
8882 2018-05-31 Jakub Jelinek <jakub@redhat.com>
8885 * bb-reorder.c (pass_partition_blocks::gate): Return false for
8886 functions with naked attribute.
8888 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
8890 * config/i386/sse.md (avx_vec_concat<mode>):
8891 Substitute concat_tg_mode mode attribute with xtg_mode.
8892 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
8893 (concat_tg_mode): Remove mode attribute.
8895 2018-05-31 Martin Sebor <msebor@redhat.com>
8898 * calls.c (alloc_max_size): Correct a logic error/typo.
8899 Treat excessive arguments as infinite. Warn for invalid arguments.
8900 * doc/invoke.texi (-Walloc-size-larger-than): Update.
8902 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
8905 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
8906 and movx for Haswell.
8908 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
8909 Cesar Philippidis <cesar@codesourcery.com>
8912 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
8913 when emitting error on private/firstprivate reductions.
8914 * omp-low.c (lower_omp_target): Avoid reference-type processing
8915 on pointers for firstprivate clause.
8917 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8919 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
8922 * config/aarch64/aarch64-simd.md
8923 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
8924 (aarch64_ld1_x3_<mode>): Likewise
8925 (aarch64_st1x2<VALLDIF:mode>): Likewise
8926 (aarch64_st1_x2_<mode>): Likewise
8927 (aarch64_st1x3<VALLDIF:mode>): Likewise
8928 (aarch64_st1_x3_<mode>): Likewise
8929 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
8930 (vld1_s8_x3): Likewise.
8931 (vld1_u16_x3): Likewise.
8932 (vld1_s16_x3): Likewise.
8933 (vld1_u32_x3): Likewise.
8934 (vld1_s32_x3): Likewise.
8935 (vld1_u64_x3): Likewise.
8936 (vld1_s64_x3): Likewise.
8937 (vld1_f16_x3): Likewise.
8938 (vld1_f32_x3): Likewise.
8939 (vld1_f64_x3): Likewise.
8940 (vld1_p8_x3): Likewise.
8941 (vld1_p16_x3): Likewise.
8942 (vld1_p64_x3): Likewise.
8943 (vld1q_u8_x3): Likewise.
8944 (vld1q_s8_x3): Likewise.
8945 (vld1q_u16_x3): Likewise.
8946 (vld1q_s16_x3): Likewise.
8947 (vld1q_u32_x3): Likewise.
8948 (vld1q_s32_x3): Likewise.
8949 (vld1q_u64_x3): Likewise.
8950 (vld1q_s64_x3): Likewise.
8951 (vld1q_f16_x3): Likewise.
8952 (vld1q_f32_x3): Likewise.
8953 (vld1q_f64_x3): Likewise.
8954 (vld1q_p8_x3): Likewise.
8955 (vld1q_p16_x3): Likewise.
8956 (vld1q_p64_x3): Likewise.
8957 (vst1_s64_x2): Likewise.
8958 (vst1_u64_x2): Likewise.
8959 (vst1_f64_x2): Likewise.
8960 (vst1_s8_x2): Likewise.
8961 (vst1_p8_x2): Likewise.
8962 (vst1_s16_x2): Likewise.
8963 (vst1_p16_x2): Likewise.
8964 (vst1_s32_x2): Likewise.
8965 (vst1_u8_x2): Likewise.
8966 (vst1_u16_x2): Likewise.
8967 (vst1_u32_x2): Likewise.
8968 (vst1_f16_x2): Likewise.
8969 (vst1_f32_x2): Likewise.
8970 (vst1_p64_x2): Likewise.
8971 (vst1q_s8_x2): Likewise.
8972 (vst1q_p8_x2): Likewise.
8973 (vst1q_s16_x2): Likewise.
8974 (vst1q_p16_x2): Likewise.
8975 (vst1q_s32_x2): Likewise.
8976 (vst1q_s64_x2): Likewise.
8977 (vst1q_u8_x2): Likewise.
8978 (vst1q_u16_x2): Likewise.
8979 (vst1q_u32_x2): Likewise.
8980 (vst1q_u64_x2): Likewise.
8981 (vst1q_f16_x2): Likewise.
8982 (vst1q_f32_x2): Likewise.
8983 (vst1q_f64_x2): Likewise.
8984 (vst1q_p64_x2): Likewise.
8985 (vst1_s64_x3): Likewise.
8986 (vst1_u64_x3): Likewise.
8987 (vst1_f64_x3): Likewise.
8988 (vst1_s8_x3): Likewise.
8989 (vst1_p8_x3): Likewise.
8990 (vst1_s16_x3): Likewise.
8991 (vst1_p16_x3): Likewise.
8992 (vst1_s32_x3): Likewise.
8993 (vst1_u8_x3): Likewise.
8994 (vst1_u16_x3): Likewise.
8995 (vst1_u32_x3): Likewise.
8996 (vst1_f16_x3): Likewise.
8997 (vst1_f32_x3): Likewise.
8998 (vst1_p64_x3): Likewise.
8999 (vst1q_s8_x3): Likewise.
9000 (vst1q_p8_x3): Likewise.
9001 (vst1q_s16_x3): Likewise.
9002 (vst1q_p16_x3): Likewise.
9003 (vst1q_s32_x3): Likewise.
9004 (vst1q_s64_x3): Likewise.
9005 (vst1q_u8_x3): Likewise.
9006 (vst1q_u16_x3): Likewise.
9007 (vst1q_u32_x3): Likewise.
9008 (vst1q_u64_x3): Likewise.
9009 (vst1q_f16_x3): Likewise.
9010 (vst1q_f32_x3): Likewise.
9011 (vst1q_f64_x3): Likewise.
9012 (vst1q_p64_x3): Likewise.
9014 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9016 * config/msp430/msp430.c (msp430_output_labelref): Prepend
9017 user_label_prefix to name.
9019 * tree-core.h: Update comment about the format of NAME string
9020 passed to handler in attribute_spec.
9022 * config/msp430/msp430.md: Remove erroneous subreg expression from
9023 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
9024 zero_extend{q,h}isi2.
9026 2018-05-30 Borislav Petkov <bp@suse.de>
9028 * doc/extend.texi: Document some architecture specific
9029 constraints and sort entries.
9031 2018-05-30 Martin Sebor <msebor@redhat.com>
9034 * builtins.c (expand_builtin_stpcpy_1): New function.
9035 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
9036 only if the former succeeds.
9038 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
9040 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
9043 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
9045 * doc/invoke.texi (-flinker-output): Document
9047 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
9049 * passes.c (ipa_write_summaries): Only modify statements if body
9051 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
9052 incrementally linking.
9053 (ipa_passes): Likewise.
9054 * lto-cgraph.c (lto_output_node): When incrementally linking do not
9055 pass down resolution info.
9056 * common.opt (flag_incremental_link): Update info.
9057 * gcc.c (plugin specs): Turn flinker-output=* to
9058 -plugin-opt=-linker-output-known
9059 * toplev.c (compile_file): Also cut compilation when doing incremental
9061 * flag-types. (enum lto_partition_model): Add
9062 LTO_LINKER_OUTPUT_NOLTOREL.
9063 (invoke.texi): Add -flinker-output docs.
9064 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
9065 link same way as WPA; do not stream in dead initializers.
9067 * dwarf2out.c (dwarf2out_die_ref_for_decl,
9068 darf2out_register_external_decl): Support incremental link.
9070 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
9072 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
9074 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
9076 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
9077 it down to simple_object_copy_lto_debug_sections.
9078 (run_gcc): Determine incremental LTO link time and configure
9079 lto1 into non-wpa mode, disable renaming of debug sections.
9081 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
9083 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
9084 descriptions of various incorrectly documented functions.
9086 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
9089 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9090 address check not strict.
9092 2018-05-30 Richard Biener <rguenther@suse.de>
9094 PR tree-optimization/85964
9095 * tracer.c (better_p): Drop initialized count check, we only
9096 call the function with initialized counts now.
9097 (find_best_successor): Do find a best edge if one
9098 has uninitialized count.
9099 (find_best_predecessor): Likewise. Do BB frequency check only
9100 if count is initialized.
9102 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
9104 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
9105 (aarch64_ldrstr_offset_compare): New.
9106 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
9107 load/store orderings.
9108 (aarch64_gen_adjusted_ldpstp): Likewise.
9110 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
9112 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
9113 Check for subset of GENERAL_REGS and FP_REGS.
9114 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
9117 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
9119 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
9120 and wi::to_poly_offset. Add the current offset and then check
9121 whether the sum fits, rather than using an unchecked addition of
9122 a checked term. Check for a shwi rather than a uhwi.
9123 * expr.c (get_bit_range): Use tree_to_poly_uint64.
9124 (store_constructor): Use poly_int_tree_p.
9125 (expand_expr_real_1): Likewise.
9126 * function.c (assign_temp): Likewise.
9127 * fold-const.c (const_binop): Use poly_int_tree_p and
9129 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
9131 * ipa-icf-gimple.c (func_checker::compare_operand): Use
9132 to_poly_offset for MEM offsets.
9133 * ipa-icf.c (sem_variable::equals): Likewise.
9134 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
9135 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
9136 wi::to_poly_offset for BIT_FIELD_REF offsets.
9137 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
9139 * var-tracking.c (emit_note_insn_var_location): Use
9140 tree_to_poly_uint64.
9142 2018-05-29 Jim Wilson <jimw@sifive.com>
9144 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
9146 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
9149 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
9150 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
9152 (sse4_1_round<mode>2): Use nonimmediate_operand
9153 for operand 1 predicate.
9155 2018-05-29 Martin Sebor <msebor@redhat.com>
9156 Richard Biener <rguenther@suse.de>
9159 * calls.c (get_size_range): Call determine_value_range instead
9160 of get_value_range..
9161 * tree-vrp.h (determine_value_range): Declared new function.
9162 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
9164 2018-05-29 Richard Biener <rguenther@suse.de>
9166 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
9167 sure to use non-pattern stmts for get_earlier_stmt arguments.
9168 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
9169 called on pattern stmts.
9170 (get_later_stmt): Likewise.
9172 2018-05-29 Martin Liska <mliska@suse.cz>
9174 PR gcov-profile/85759
9175 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
9178 2018-05-29 Jakub Jelinek <jakub@redhat.com>
9180 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
9182 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
9186 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
9187 VEC_PACK_FLOAT_EXPR): New tree codes.
9188 * tree-pretty-print.c (op_code_prio): Handle
9189 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
9190 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
9191 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
9192 * tree-inline.c (estimate_operator_cost): Likewise.
9193 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
9194 * fold-const.c (const_binop): Likewise.
9195 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
9196 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
9197 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9198 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
9199 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
9200 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
9201 * expr.c (expand_expr_real_2): Likewise.
9202 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
9203 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
9204 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
9206 * optabs.c (expand_widen_pattern_expr): For
9207 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
9208 sign from result type rather than operand's type.
9209 (expand_binop_directly): For vec_packu_float_optab and
9210 vec_packs_float_optab allow result type to be different from operand's
9212 * optabs-tree.c (optab_for_tree_code): Handle
9213 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
9214 VEC_PACK_FLOAT_EXPR. Formatting fixes.
9215 * tree-vect-generic.c (expand_vector_operations_1): Handle
9216 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
9217 VEC_PACK_FLOAT_EXPR.
9218 * tree-vect-stmts.c (supportable_widening_operation): Handle
9220 (supportable_narrowing_operation): Handle FLOAT_EXPR.
9221 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
9222 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
9223 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
9224 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
9226 (vec_pack<floatprefix>_float_<mode>): New expander.
9227 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
9229 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
9230 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
9231 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
9232 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
9233 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
9235 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
9236 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
9237 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
9238 VEC_PACK_FLOAT_EXPR): Document.
9240 2018-05-29 Richard Biener <rguenther@suse.de>
9242 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
9244 (stmt_vec_info_vec): Make pointer.
9245 (init_stmt_vec_info_vec): Remove.
9246 (free_stmt_vec_info_vec): Likewise.
9247 (set_stmt_vec_info_vec): New function.
9248 (free_stmt_vec_infos): Likewise.
9249 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
9250 (set_vinfo_for_stmt): Likewise.
9251 (get_earlier_stmt): Likewise.
9252 (get_later_stmt): Likewise.
9253 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
9254 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
9255 (vec_info::~vec_info): Free stmt_vec_infos.
9256 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
9257 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
9258 (pass_slp_vectorize::execute): Likewise.
9259 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
9260 (free_stmt_vec_info_vec): Likewise.
9261 (set_stmt_vec_info_vec): New function.
9262 (free_stmt_vec_infos): Likewise.
9263 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
9264 the global stmt_vec_info_vec.
9265 * tree-parloops.c (gather_scalar_reductions): Use
9266 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
9269 2018-05-29 Richard Biener <rguenther@suse.de>
9271 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
9273 2018-05-29 Martin Liska <mliska@suse.cz>
9274 David Malcolm <dmalcolm@redhat.com>
9276 * vec.c (test_reverse): New.
9277 (vec_c_tests): Add new test.
9278 * vec.h (vl_ptr>::reverse): New function.
9280 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
9282 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
9284 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
9287 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
9289 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
9291 2018-05-28 Richard Biener <rguenther@suse.de>
9293 PR tree-optimization/85933
9294 * tree-vect-data-refs.c (vect_record_base_alignments): Only
9295 look at stmts marked as vectorizable.
9297 2018-05-28 Richard Biener <rguenther@suse.de>
9299 PR tree-optimization/85934
9300 * tree-vect-generic.c (expand_vector_operations_1): Hoist
9301 vector boolean check before scalar optimization.
9303 2018-05-28 Jakub Jelinek <jakub@redhat.com>
9305 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
9308 2018-05-28 Mark Wielaard <mark@klomp.org>
9310 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
9311 if it is an expression containing a minus sign.
9313 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
9315 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
9317 2018-05-27 Paul Koning <ni1d@arrl.net>
9319 * config/pdp11/pdp11.md (truncsihi2): Remove.
9321 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
9322 Chung-Ju Wu <jasonwucj@gmail.com>
9324 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
9326 (unaligned_store_dw): Ditto.
9327 * config/nds32/nds32-memory-manipulation.c
9328 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
9329 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
9330 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
9331 (emit_setmem_word_loop): Rename to ...
9332 (emit_setmem_doubleword_loop): ... this.
9333 (nds32_gen_dup_4_byte_to_word_value): New function.
9334 (nds32_gen_dup_8_byte_to_double_word_value): New function.
9335 (nds32_expand_setmem_loop): Refine implementation.
9336 (nds32_expand_setmem_loop_v3m): Ditto.
9337 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
9340 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
9342 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
9344 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
9346 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
9347 (nds32_init_machine_status): Initialize machine->attr_naked_p and
9348 machine->attr_no_prologue_p.
9349 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
9350 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
9351 (nds32_expand_epilogue): Consider attr_naked_p.
9352 (nds32_expand_epilogue_v3pop): Likewise.
9353 (nds32_can_use_return_insn): Likewise.
9354 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
9355 attr_no_prologue_p fields.
9356 * config/nds32/nds32.opt (mret-in-naked-func): New option.
9358 2018-05-27 Jakub Jelinek <jakub@redhat.com>
9361 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
9363 * config/i386/sse.md
9364 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
9366 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
9368 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
9370 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
9372 (*<floatsuffix>floatv2div2sf2): Rename to ...
9373 (*float<floatunssuffix>v2div2sf2): ... this.
9374 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
9375 (float<floatunssuffix>v2div2sf2_mask): ... this.
9376 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
9377 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
9378 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
9380 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
9382 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
9384 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
9386 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
9388 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
9390 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
9391 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
9392 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
9393 gen_ufix_truncv8dfv8si2.
9394 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
9395 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
9396 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
9397 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
9398 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
9399 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
9400 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
9401 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
9403 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
9407 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
9409 2018-05-25 Jim Wilson <jimw@sifive.com>
9411 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
9412 * config/riscv/riscv.c (struct machine_function): Add
9413 interrupt_handler_p and attribute_checked_p fields.
9414 (riscv_attribute_table): Add interrupt.
9415 (riscv_interrupt_type_p): New.
9416 (riscv_save_reg_p): Save extra regs for interrupt handler.
9417 (riscv_use_save_libcall): Return false for interrupt handler.
9418 (riscv_first_stack_step): Add forward declaration.
9419 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
9420 for interrupt handler with large frame. Use it for saved reg list.
9421 (riscv_expand_prologue): Move flag_stack_usage_info support to
9422 eliminate duplication.
9423 (riscv_expand_epilogue): Generate mret for interrupt handler.
9424 (riscv_epilogue_uses): New.
9425 (riscv_can_use_return_insn): Return false for interrupt handler.
9426 (riscv_function_ok_for_sibcall): Likewise.
9427 (riscv_set_current_function): Add interrupt handler support.
9428 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
9429 * config/riscv/riscv.md (UNSPECV_MRET): New.
9431 (riscv_frflags, riscv_fsflags): Use tab after opcode.
9433 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
9435 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
9437 PR tree-optimization/85712
9438 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
9439 this candidate has already been replaced in-situ by a copy.
9441 2018-05-25 Jason Merrill <jason@redhat.com>
9443 PR c++/80485 - inline function non-zero address.
9444 * symtab.c (nonzero_address): Check DECL_COMDAT.
9446 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
9449 * config/alpha/alpha.md (ashlsi3): New insn pattern.
9450 (*ashlsi_se): Rename from *ashldi_se. Define as sign
9451 extension of SImode operation. Use const123_operand predicate.
9452 (*saddsi_1): Remove.
9453 (*saddl_se_1): Ditto.
9455 (*ssubl_se_1): Ditto.
9456 * config/alpha/predicates.md (const123_operand): New predicate.
9457 * config/alpha/constraints.md (P): Use IN_RANGE.
9459 2018-05-25 Richard Biener <rguenther@suse.de>
9461 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
9463 (ref_maybe_used_by_stmt_p): Likewise.
9464 (stmt_may_clobber_ref_p): Likewise.
9465 (stmt_may_clobber_ref_p_1): Likewise.
9466 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
9468 (ref_maybe_used_by_stmt_p): Likewise.
9469 (stmt_may_clobber_ref_p): Likewise.
9470 (stmt_may_clobber_ref_p_1): Likewise.
9471 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
9472 the alias oracle to disambiguate DRs with stmts DR analysis
9474 (vect_analyze_data_refs): Do not give up on not analyzable
9475 DRs for BB vectorization. Remove code truncating the dataref
9478 2018-05-25 Jakub Jelinek <jakub@redhat.com>
9481 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
9482 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
9483 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
9485 2018-05-25 Richard Biener <rguenther@suse.de>
9487 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
9488 function, combining stmt data ref gathering and fatal analysis
9490 (vect_analyze_data_refs): Remove now redudnant code and simplify.
9491 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
9492 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
9493 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
9494 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
9496 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9498 PR tree-optimization/85720
9499 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
9500 SCC if all partitions are builtins.
9501 (version_loop_by_alias_check): New parameter. Generate cancelable
9502 runtime alias check if all partitions are builtins.
9503 (distribute_loop): Update call to above function.
9505 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9507 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
9508 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
9509 (parm_default_def_partition_arg): Ditto.
9510 (set_parm_default_def_partition): Ditto.
9511 (get_parm_default_def_partitions): Ditto and make it static.
9512 (get_undefined_value_partitions): Ditto and make it static.
9513 (remove_ssa_form): Refactor call to init_var_map here.
9514 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
9515 computation for loop region.
9516 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
9517 (register_default_def): Delete.
9518 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
9519 (parm_default_def_partition_arg): Ditto.
9520 (set_parm_default_def_partition): Ditto.
9521 (get_parm_default_def_partitions): Ditto and make it static.
9522 (get_undefined_value_partitions): Ditto and make it static.
9523 (coalesce_with_default, coalesce_with_default): Update comment.
9524 (create_coalesce_list_for_region): New func factored out from
9525 create_outofssa_var_map.
9526 (populate_coalesce_list_for_outofssa): New func factored out from
9527 create_outofssa_var_map and coalesce_ssa_name.
9528 (create_outofssa_var_map): Delete.
9529 (coalesce_ssa_name): Refactor to support live range computation.
9530 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
9531 (get_parm_default_def_partitions): Delete.
9532 (get_undefined_value_partitions): Ditto.
9533 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
9534 computation for loop region.
9535 (new_tree_live_info, loe_visit_block): Ditto.
9536 (live_worklist, set_var_live_on_entry): Ditto.
9537 (calculate_live_on_exit, verify_live_on_entry): Ditto.
9538 * tree-ssa-live.h (struct _var_map): New fields.
9539 (init_var_map): Change decl.
9540 (region_contains_p): New.
9542 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9544 * tree-ssa-live.h (live_merge_and_clear): Delete.
9546 2018-05-25 Richard Biener <rguenther@suse.de>
9549 * tree-dump.c (dequeue_and_dump): Remove access to removed
9550 operand 2 of a SWITCH_EXPR.
9552 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9554 * doc/sourcebuild.texi (vect_double_cond_arith): Include
9555 multiplication and division.
9556 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
9557 (cond_udiv@var{m}, cond_umod@var{m}): Document.
9558 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
9559 (cond_udiv_optab, cond_umod_optab): New optabs.
9560 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
9561 (IFN_COND_RDIV): New internal functions.
9562 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
9563 TRUNC_MOD_EXPR and RDIV_EXPR.
9564 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
9565 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
9567 (SVE_INT_BINARY): Include mult.
9568 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
9569 (optab, sve_int_op): Handle mult.
9570 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
9572 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
9573 for SVE_INT_BINARY_SD.
9575 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9577 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
9578 (optab, sve_int_op): Handle div and udiv.
9579 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
9580 for SVE_INT_BINARY_SD.
9581 (*<optab><mode>3): New insn for the same.
9583 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9585 * tree-vect-patterns.c: Include predict.h.
9586 (vect_recog_divmod_pattern): Restrict check for division support
9587 to when optimizing for size.
9589 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9591 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
9592 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
9593 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
9594 (gimple_match_op::set_op): Likewise.
9595 (gimple_resimplify4): Declare.
9596 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
9597 (expr::gen_transform): Likewise.
9598 (decision_tree::gen): Generate a simplification routine for 4 operands.
9599 * gimple-match-head.c (gimple_simplify): Add an overload for
9600 4 operands. In the top-level function, handle up to 4 call
9601 arguments and call gimple_resimplify4.
9602 (gimple_resimplify4): New function.
9603 (build_call_internal): Pass a fourth operand.
9604 (maybe_push_to_seq): Likewise.
9605 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
9606 Fold VEC_COND_EXPRs of an operation and a default value into
9607 an IFN_COND_* function if possible.
9608 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
9610 (SVE_COND_FP_BINARY): Include them.
9611 (optab, sve_fp_op): Handle them.
9612 (SVE_INT_BINARY_REV): New code iterator.
9613 (SVE_COND_FP_BINARY_REV): New int iterator.
9614 (commutative): New int attribute.
9615 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
9617 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
9619 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
9620 (*cond_<optab><mode>): New patterns for reversed operands.
9622 2018-05-25 Richard Biener <rguenther@suse.de>
9624 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
9625 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
9626 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
9627 (STMT_VINFO_GROUPED_ACCESS): Adjust.
9628 * tree-vect-data-refs.c (everywhere): Adjust users.
9629 * tree-vect-loop.c (everywhere): Likewise.
9630 * tree-vect-slp.c (everywhere): Likewise.
9631 * tree-vect-stmts.c (everywhere): Likewise.
9632 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
9634 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9636 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
9638 (gcc_cv_as_section_exclude): ... this.
9639 Try Solaris as #exclude syntax.
9640 * configure: Regenerate.
9641 * config.in: Regenerate.
9642 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
9644 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
9645 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
9647 * varasm.c (default_elf_asm_named_section): Don't check if
9648 HAVE_GAS_SECTION_EXCLUDE is defined.
9650 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9652 * doc/md.texi: Update the documentation of the cond_* optabs
9653 to mention the new final operand. Fix GET_MODE_NUNITS call.
9654 Describe the scalar case too.
9655 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
9656 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
9658 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
9659 (get_conditional_internal_fn): Update comment.
9660 * tree-vect-loop.c (vectorizable_reduction): Pass the original
9661 accumulator value as a final argument to conditional functions.
9662 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
9663 a define_expand and add an "else" operand. Assert for now that
9664 the else operand is equal to operand 2. Use SVE_INT_BINARY and
9665 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
9666 (*cond_<optab><mode>): New patterns.
9667 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
9668 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9669 (UNSPEC_COND_EOR): Delete.
9670 (optab): Remove associated mappings.
9671 (SVE_INT_BINARY): New code iterator.
9672 (sve_int_op): Remove int attribute and add "minus" to the code
9674 (SVE_COND_INT_OP): Delete.
9675 (SVE_COND_FP_OP): Rename to...
9676 (SVE_COND_FP_BINARY): ...this.
9678 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9680 * optabs.c (can_reuse_operands_p): New function.
9681 (maybe_legitimize_operands): Try to reuse the results for
9684 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9686 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
9687 Add {q} suffix to insn mnemonic.
9689 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9691 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
9692 (msp430_warn_func_return): New.
9694 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
9696 * fold-const.c (tree_nonzero_bits): New function.
9697 * fold-const.h (tree_nonzero_bits): Likewise.
9698 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
9699 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
9701 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
9705 * varasm.c (assemble_alias): Check ifunc_resolver only on
9708 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9711 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
9712 when memory input operand is handled.
9714 2018-05-24 Luis Machado <luis.machado@linaro.org>
9716 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
9718 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
9720 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9722 * match.pd: Delay FMA folds until after vectorization.
9724 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
9727 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9728 address check not strict.
9730 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9732 * gimple-match.h (gimple_match_op): New class.
9733 (mprts_hook): Replace parameters with a gimple_match_op *.
9734 (maybe_build_generic_op): Likewise.
9735 (gimple_simplified_result_is_gimple_val): Replace parameters with
9736 a const gimple_match_op *.
9737 (gimple_simplify): Replace code_helper * and tree * parameters with
9738 a gimple_match_op * parameter.
9739 (gimple_resimplify1): Replace code_helper *, tree and tree *
9740 parameters with a gimple_match_op * parameter.
9741 (gimple_resimplify2): Likewise.
9742 (gimple_resimplify3): Likewise.
9743 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
9744 parameters with a gimple_match_op * parameter.
9745 * gimple-match-head.c (gimple_simplify): Change prototypes of
9746 auto-generated functions to take a gimple_match_op * instead of
9747 separate code_helper * and tree * parameters. Make the same
9748 change in the top-level overload and update calls to the
9749 gimple_resimplify routines. Update calls to the auto-generated
9750 functions and to maybe_push_res_to_seq in the publicly-facing
9751 operation-specific gimple_simplify overloads.
9752 (gimple_match_op::MAX_NUM_OPS): Define.
9753 (gimple_resimplify1): Replace rcode and ops with a single res_op
9754 parameter. Update call to gimple_simplify.
9755 (gimple_resimplify2): Likewise.
9756 (gimple_resimplify3): Likewise.
9757 (mprts_hook): Replace parameters with a gimple_match_op *.
9758 (maybe_build_generic_op): Likewise.
9759 (build_call_internal): Replace type, nargs and ops with
9760 a gimple_match_op *.
9761 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
9762 with a single gimple_match_op *. Update calls to mprts_hook,
9763 build_call_internal and gimple_simplified_result_is_gimple_val.
9764 Factor out code that is common to the tree_code and combined_fn cases.
9765 * genmatch.c (expr::gen_transform): Replace tem_code and
9766 tem_ops with a gimple_match_op called tem_op. Update calls
9767 to the gimple_resimplify functions and maybe_push_res_to_seq.
9768 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
9769 res_ops. Update call to the gimple_resimplify functions.
9770 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
9771 (decision_tree::gen): Make the functions take a gimple_match_op *
9772 called res_op instead of separate res_code and res_ops parameters.
9773 Update call accordingly.
9774 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
9775 and ops with a single res_op parameter. Update calls to
9776 maybe_build_generic_op and maybe_push_res_to_seq.
9777 (fold_stmt_1): Update calls to gimple_simplify and
9778 replace_stmt_with_simplification.
9779 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
9780 and gimple_simplified_result_is_gimple_val.
9781 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
9783 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
9784 with a gimple_match_op *.
9785 (vn_nary_build_or_lookup): Likewise. Update call to
9786 vn_nary_build_or_lookup_1.
9787 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
9788 gimple_match_op *. Update calls to the gimple_resimplify routines
9789 and to gimple_simplified_result_is_gimple_val.
9790 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
9791 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
9792 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
9793 (visit_nary_op): Likewise.
9794 (visit_reference_op_load): Likewise.
9796 2018-05-23 Luis Machado <luis.machado@linaro.org>
9798 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
9799 modifier for printing the step amount.
9801 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
9804 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
9807 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
9809 * doc/sourcebuild.texi (Endianness): New subsubsection.
9811 2018-05-23 Luis Machado <luis.machado@linaro.org>
9813 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9814 <prefetch_dynamic_strides>: New const bool field.
9815 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9816 prefetch_dynamic_strides.
9817 (exynosm1_prefetch_tune): Likewise.
9818 (thunderxt88_prefetch_tune): Likewise.
9819 (thunderx_prefetch_tune): Likewise.
9820 (thunderx2t99_prefetch_tune): Likewise.
9821 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
9823 (aarch64_override_options_internal): Update to set
9824 PARAM_PREFETCH_DYNAMIC_STRIDES.
9825 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9826 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9827 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9828 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9829 prefetch-dynamic-strides setting.
9831 2018-05-23 Luis Machado <luis.machado@linaro.org>
9833 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9834 <minimum_stride>: New const int field.
9835 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9836 minimum_stride field defaulting to -1.
9837 (exynosm1_prefetch_tune): Likewise.
9838 (thunderxt88_prefetch_tune): Likewise.
9839 (thunderx_prefetch_tune): Likewise.
9840 (thunderx2t99_prefetch_tune): Likewise.
9841 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
9842 <default_opt_level>: Set to 3.
9843 (aarch64_override_options_internal): Update to set
9844 PARAM_PREFETCH_MINIMUM_STRIDE.
9845 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9846 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9847 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9848 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9849 stride is constant and is below the minimum stride threshold.
9851 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9853 * config/arm/arm-cpus.in (mode26): Delete.
9854 (armv4): Delete mode26 reference.
9855 * config/arm/arm.c (arm_configure_build_target): Delete use of
9858 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9860 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
9862 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
9863 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
9865 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
9866 (floatunsdidf2): Ditto.
9868 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9870 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
9871 (fixuns_trunc<mode>si2_avx512f): Ditto.
9872 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
9873 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
9874 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
9876 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9878 PR rtl-optimization/79985
9879 * df-scan.c (df_insn_refs_collect): Remove special case for
9880 global registers and asm statements.
9882 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9884 * extend.texi (Global Register Variables): Rewrite the bullet list.
9885 Note that the register is available for allocation. Note that access
9886 via inline asm must use constraints. Add note about async-signal
9887 handlers. Remove paragraph about automagic register selection.
9889 2018-05-23 Richard Biener <rguenther@suse.de>
9891 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
9892 of fixed offset from memset VN.
9894 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
9896 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
9898 (alloc_cand_and_find_basis): Initialize first_interp field.
9899 (slsr_process_mul): Modify first_interp field.
9900 (slsr_process_add): Likewise.
9901 (slsr_process_cast): Modify first_interp field for each new
9903 (slsr_process_copy): Likewise.
9904 (dump_candidate): Dump first_interp field.
9905 (replace_mult_candidate): Process all interpretations, not just
9907 (replace_rhs_if_not_dup): Likewise.
9908 (replace_one_candidate): Likewise.
9910 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
9912 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
9914 (aarch64_needs_frame_chain): New function.
9915 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
9917 2018-05-23 Sudakshina Das <sudi.das@arm.com>
9920 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
9921 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
9922 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
9923 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
9924 as true for strict-align.
9925 (aarch64_can_inline_p): Perform checks even when callee has no
9926 attributes to check for strict alignment.
9927 * doc/extend.texi (AArch64 Function Attributes): Document
9929 * doc/invoke.texi: (AArch64 Options): Likewise.
9931 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
9933 PR tree-optimization/85853
9934 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
9935 the handling of the root of the node to...
9936 (vect_slp_analyze_node_operations_1): ...this new function,
9937 and run the whole thing with the child nodes' def types
9938 set according to their SLP node's def type.
9940 2018-05-23 Richard Biener <rguenther@suse.de>
9943 * tree-data-ref.c (create_runtime_alias_checks): Defer
9944 and ignore overflow warnings.
9946 2018-05-23 Yury Gribov <tetra2005@gmail.com>
9948 PR tree-optimization/85822
9949 * tree-vrp.c (is_masked_range_test): Fix handling of negative
9952 2018-05-23 Richard Biener <rguenther@suse.de>
9954 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9955 memset constants via native_interpret_expr.
9957 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
9960 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
9962 (cgraph_node::create_alias): Likewise.
9963 (cgraph_node::get_availability): Check ifunc_resolver instead
9964 of looking up ifunc attribute.
9965 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9966 * varasm.c (do_assemble_alias): Likewise.
9967 (assemble_alias): Likewise.
9968 (default_binds_local_p_3): Likewise.
9969 * cgraph.h (cgraph_node): Add ifunc_resolver.
9970 (cgraph_node::only_called_directly_or_aliased_p): Return false
9972 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
9974 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
9975 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
9976 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
9977 instead of looking up ifunc attribute.
9979 2018-05-22 Luis Machado <luis.machado@linaro.org>
9981 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
9983 2018-05-22 Martin Sebor <msebor@redhat.com>
9986 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
9987 only when expasion succeeds.
9988 (expand_builtin_strcmp): Same.
9989 (expand_builtin_strncmp): Same.
9991 2018-05-22 Martin Sebor <msebor@redhat.com>
9993 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
9995 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9996 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9998 * config/aarch64/aarch64-ldpstp.md: Replace uses of
9999 aarch64_mem_pair_operand with memory_operand and delete operand swapping
10001 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
10002 Add check for legitimate_address.
10003 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
10004 (aarch64_swap_ldrstr_operands): New.
10005 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
10008 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
10009 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10011 * config/aarch64/aarch64.md: New patterns to generate stp
10013 (store_pair_sw, store_pair_dw): New patterns to generate stp for
10014 single words and double words.
10015 (load_pair_sw, load_pair_dw): Likewise.
10016 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
10018 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
10020 * config/aarch64/aarch64-ldpstp.md: Modify peephole
10021 for different mode ldpstp and add peephole for merged zero stores.
10022 Likewise for loads.
10023 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
10025 (aarch64_gen_store_pair): Rename calls to match new patterns.
10026 (aarch64_gen_load_pair): Rename calls to match new patterns.
10027 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
10028 (load_pair<DREG:mode><DREG2:mode>): ... This.
10029 (store_pair<mode>): Rename to...
10030 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
10031 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
10032 New mode iterators.
10033 (V_INT_EQUIV): Handle SImode.
10034 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
10037 2018-05-22 Martin Sebor <msebor@redhat.com>
10040 * calls.c (maybe_warn_nonstring_arg): Use string length to set
10041 or ajust the presumed bound on an operation to avoid unnecessary
10044 2018-05-22 Martin Sebor <msebor@redhat.com>
10046 PR tree-optimization/85826
10047 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
10048 assuming that a DECL necesarily has a constant size.
10050 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
10052 PR middle-end/85862
10053 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
10055 2018-05-22 Richard Biener <rguenther@suse.de>
10057 PR tree-optimization/85834
10058 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
10059 non-constant and non-zero memset arguments.
10061 2018-05-22 Martin Liska <mliska@suse.cz>
10064 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
10066 2018-05-22 Richard Biener <rguenther@suse.de>
10068 PR tree-optimization/85863
10069 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
10070 comparisons when vectype is specified.
10071 (vectorizable_condition): Do not specify vectype for
10072 vect_is_simple_cond when SLP vectorizing.
10074 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
10077 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
10078 define __ibm128 as long double.
10079 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
10080 as a distinct type when IEEE 128-bit support is enabled.
10081 (init_float128_ieee): Fix up conversions between IFmode and IEEE
10082 128-bit types to use the correct functions.
10083 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
10084 convert between 128-bit floating point types that have different
10085 modes but the same representation, instead of using gen_lowpart to
10087 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
10089 (IFKF_reg): New attributes to give the register constraints for
10091 (extend<mode>tf2_internal): New insns to mark an explicit
10092 conversion between 128-bit floating point types that have a
10093 different mode but share the same representation.
10095 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
10097 PR tree-optimization/85814
10098 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
10099 a null return from get_strinfo when unsharing the next
10100 strinfo in the chain.
10102 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10105 * varasm.c (weak_finish): Clean up weak_decls.
10107 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10109 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
10110 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
10111 UNSPEC_UADALP values.
10112 * config/aarch64/iterators.md (ABAL): New int iterator.
10115 (sur): Add mappings for the above.
10116 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
10118 (aarch64_<sur>abal<mode>_4): Likewise.
10119 (aarch64_<sur>adalp<mode>_3): Likewise.
10120 (<sur>sadv16qi): New define_expand.
10122 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
10124 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
10125 (*movdf_internal): Ditto.
10126 (*rcpsf2_sse): Ditto.
10127 (*rsqrtsf2_sse): Ditto.
10128 (*sqrt<mode>2_sse): Ditto.
10130 2018-05-21 Tamar Christina <tamar.christina@arm.com>
10132 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
10134 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
10135 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
10136 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
10137 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
10139 * config/aarch64/arm_neon.h: Likewise.
10140 * config/aarch64/iterators.md (VQ_I): New.
10142 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
10144 * config.gcc: Add arc/t-multilib-linux to tmake_file for
10146 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
10149 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
10151 * config/nds32/constraints.md (S): New constraint.
10152 * config/nds32/nds32.md (call_internal): Use constraint S.
10153 (call_value_internal): Likewise.
10154 (sibcall_internal): Likewise.
10155 (sibcall_value_internal): Likewise.
10157 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
10158 Chung-Ju Wu <jasonwucj@gmail.com>
10160 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
10161 into consideration.
10163 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
10164 Chung-Ju Wu <jasonwucj@gmail.com>
10166 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
10167 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
10168 (nds32_rtx_costs_impl): Simplify.
10169 (nds32_address_cost_impl): Simplify.
10170 (nds32_init_rtx_costs): New function.
10171 (nds32_rtx_costs_speed_prefer): Likewise.
10172 (nds32_rtx_costs_size_prefer): Likewise.
10173 (nds32_address_cost_speed_prefer): Likewise.
10174 (nds32_address_cost_speed_fwprop): Likewise.
10175 (nds32_address_cost_size_prefer): Likewise.
10176 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
10177 * config/nds32/nds32.c (nds32_option_override): Use
10178 nds32_init_rtx_costs function.
10180 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
10182 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
10183 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
10184 (TARGET_PIPELINE_N8): Likewise.
10185 (TARGET_PIPELINE_N10): Likewise.
10186 (TARGET_PIPELINE_N13): Likewise.
10187 (TARGET_PIPELINE_GRAYWOLF): Likewise.
10189 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
10191 * config/nds32/nds32-fpu.md: Update copyright year.
10193 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10195 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
10197 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10199 * config/nds32/nds32.c
10200 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
10201 * config/nds32/nds32.opt (minline-asm-r15): New option.
10203 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10205 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
10207 * config/nds32/nds32.md (abssi2): New pattern.
10209 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
10211 * config/i386/i386.md (rex64namesuffix): New mode attribute.
10212 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
10213 Merge insn pattern from sse_cvtsi2ss<round_name> and
10214 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
10215 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
10216 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
10217 using SWI48 mode iterator.
10218 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
10219 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
10220 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
10221 pattern from sse_cvttss2si<round_saeonly_name>
10222 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
10223 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
10224 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
10225 using SWI48 mode iterator.
10226 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
10227 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
10228 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
10229 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
10230 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
10231 using SWI48 mode iterator.
10232 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
10233 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
10234 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
10235 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
10236 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
10237 SWI48 mode iterator.
10238 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
10239 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
10240 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
10241 pattern from sse_cvttsd2si<round_saeonly_name>
10242 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
10244 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10246 * config/nds32/nds32-md-auxiliary.c
10247 (nds32_valid_smw_lwm_base_p): Refine.
10248 (nds32_output_smw_single_word): Refine.
10249 (nds32_output_smw_double_word): New.
10250 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
10252 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10254 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
10255 (nds32_output_stack_pop): Refine.
10256 (nds32_expand_unaligned_load): Refine.
10257 (nds32_expand_unaligned_store): Refine.
10259 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10260 Chung-Ju Wu <jasonwucj@gmail.com>
10262 * config/nds32/constants.md: Add TP_REGNUM constant.
10263 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
10264 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
10266 * config/nds32/nds32-doubleword.md: Consider flag_pic.
10267 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
10268 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
10269 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
10270 and PIC code generation.
10271 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
10273 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
10275 * config/nds32/nds32.md: Support TLS and PIC.
10276 * config/nds32/nds32.c: Support TLS and PIC.
10277 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
10278 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
10281 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10283 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
10284 mode with E_ prefix.
10286 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10287 Chung-Ju Wu <jasonwucj@gmail.com>
10289 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
10290 * config/nds32/nds32-md-auxiliary.c
10291 (symbolic_reference_mentioned_p): New.
10292 (nds32_legitimize_ict_address): New.
10293 (nds32_expand_ict_move): New.
10294 (nds32_indirect_call_referenced_p): New.
10295 (nds32_symbol_binds_local_p): Delete.
10296 (nds32_long_call_p): Modify.
10297 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
10298 * config/nds32/nds32-protos.h
10299 (symbolic_reference_mentioned_p): Declare.
10300 (nds32_legitimize_ict_address): Declare.
10301 (nds32_expand_ict_move): Declare.
10302 (nds32_indirect_call_referenced_p): Declare.
10303 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
10304 (nds32_relax_group): Use nds32_ict_const_p as condition.
10305 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
10306 (nds32_asm_file_start): Output ict_model directive in asm code.
10307 (nds32_legitimate_address_p): Consider indirect call.
10308 (nds32_print_operand): Consider indirect call.
10309 (nds32_print_operand_address): Consider indirect call.
10310 (nds32_insert_attributes): Handle "indirect_call" attribute.
10311 (TARGET_LEGITIMATE_ADDRESS_P): Define.
10312 (TARGET_LEGITIMATE_CONSTANT_P): Define.
10313 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
10314 (TARGET_DELEGITIMIZE_ADDRESS): Define.
10315 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
10316 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
10317 (TARGET_ICT_MODEL_SMALL): Define.
10318 (TARGET_ICT_MODEL_LARGE): Define.
10319 * config/nds32/nds32.md (movsi): Consider ict model.
10320 (call, call_value): Consider ict model.
10321 (sibcall, sibcall_value): Consider ict model.
10322 * config/nds32/nds32.opt (mict-model): New option.
10323 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
10326 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
10327 Monk Chiang <sh.chiang04@gmail.com>
10328 Jim Wilson <jimw@sifive.com>
10330 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
10331 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
10332 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
10333 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
10334 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
10335 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
10336 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
10337 compute save_libcall_adjustment properly.
10338 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
10339 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
10340 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
10341 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
10342 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
10343 (ABI_SPEC): Handle mabi=ilp32e.
10344 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
10345 (RVE): Add RVE mask.
10346 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
10347 <-march>: Add rv32e as an example.
10349 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
10352 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
10353 (intra_create_variable_infos): Handle C++ constructors.
10355 2018-05-18 Martin Liska <mliska@suse.cz>
10357 * passes.def: Remove a redundant pass.
10359 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
10362 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
10364 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10366 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
10368 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
10370 (armv2, armv2a, armv3, armv3m): Delete architectures.
10371 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
10372 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
10373 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
10375 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
10376 (*mulsidi3adddi): Likewise.
10377 (mulsidi3): Likewise.
10378 (*mulsidi3_nov6): Likewise.
10379 (umulsidi3): Likewise.
10380 (umulsidi3_nov6): Likewise.
10381 (umaddsidi4): Likewise.
10382 (*umulsidi3adddi): Likewise.
10383 (smulsi3_highpart): Likewise.
10384 (*smulsi3_highpart_nov6): Likewise.
10385 (umulsi3_highpart): Likewise.
10386 (*umulsi3_highpart_nov6): Likewise.
10387 * config/arm/arm.h (arm_arch3m): Delete.
10388 * config/arm/arm.c (arm_arch3m): Delete.
10389 (arm_option_override_internal): Update armv3-related comment.
10390 (arm_configure_build_target): Delete use of isa_bit_mode32.
10391 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
10392 (arm_rtx_costs_internal): Delete check of arm_arch3m.
10393 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
10394 (mulsa3): Likewise.
10395 (mulusa3): Likewise.
10396 * config/arm/arm-protos.h (arm_arch3m): Delete.
10397 * config/arm/arm-tables.opt: Regenerate.
10398 * config/arm/arm-tune.md: Likewise.
10399 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
10400 deleted architectures.
10402 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10404 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
10405 (armv5t, armv5te): New features.
10406 (ARMv5, ARMv5e): Delete fgroups.
10407 (ARMv5t, ARMv5te): Adjust for above changes.
10408 (ARMv6m): Likewise.
10409 (armv5, armv5e): Delete arches.
10410 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
10412 (*call_reg_arm): Likewise.
10413 (*call_value_reg_armv5): Likewise.
10414 (*call_value_reg_arm): Likewise.
10415 (*call_symbol): Likewise.
10416 (*call_value_symbol): Likewise.
10417 (*sibcall_insn): Likewise.
10418 (*sibcall_value_insn): Likewise.
10419 (clzsi2): Likewise.
10420 (prefetch): Likewise.
10421 (define_split and define_peephole2 dependent on arm_arch5):
10423 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
10425 (TARGET_ARM_QBIT): Likewise.
10426 (TARGET_DSP_MULTIPLY): Likewise.
10427 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
10428 (arm_arch5, arm_arch5e): Delete.
10429 (arm_arch5t, arm_arch5te): Declare.
10430 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
10431 (arm_arch5t): Declare.
10432 (arm_option_reconfigure_globals): Update for the above.
10433 (arm_options_perform_arch_sanity_checks): Update comment, replace
10434 use of arm_arch5 with arm_arch5t.
10435 (use_return_insn): Likewise.
10436 (arm_emit_call_insn): Likewise.
10437 (output_return_instruction): Likewise.
10438 (arm_final_prescan_insn): Likewise.
10439 (arm_coproc_builtin_available): Likewise.
10440 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
10441 arm_arch5e with arm_arch5t and arm_arch5te.
10442 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
10443 (arm_arch5t, arm_arch5te): Declare.
10444 * config/arm/arm-tables.opt: Regenerate.
10445 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
10446 * config/arm/t-multilib: Likewise.
10447 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
10448 instead of arm_arch5.
10449 (*call_reg_thumb1): Likewise.
10450 (*call_value_reg_thumb1_v5): Likewise.
10451 (*call_value_reg_thumb1): Likewise.
10452 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
10454 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
10456 2018-05-18 Martin Liska <mliska@suse.cz>
10458 PR gcov-profile/84846
10459 * doc/gcov.texi: Document -t option of gcov tool.
10461 2018-05-18 Martin Liska <mliska@suse.cz>
10463 PR gcov-profile/84846
10464 * gcov.c (print_usage): Add new -t option.
10465 (process_args): Handle the option.
10466 (generate_results): Use stdout as output when requested by
10469 2018-05-18 Martin Liska <mliska@suse.cz>
10471 PR gcov-profile/84846
10472 * coverage.c (coverage_init): Write PWD to .gcno file.
10473 * doc/gcov.texi: Document how working directory is printed.
10474 * gcov-dump.c (dump_gcov_file): Print PWD.
10475 * gcov.c (output_intermediate_file): Likewise.
10476 (read_graph_file): Read PWD string.
10477 (output_lines): Print PWD.
10479 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10481 PR middle-end/85817
10482 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
10483 for retval and return false if all args to phi are zero.
10485 2018-05-18 Richard Biener <rguenther@suse.de>
10487 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
10489 (evrp_dom_walker::before_dom_children): Call it.
10491 2018-05-18 Richard Biener <rguenther@suse.de>
10493 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
10494 results when processing array refs with variable index.
10496 2018-05-18 Toon Moene <toon@moene.org>
10498 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
10499 directly after that of -floop-interchange. Indicate that both
10500 options are enabled by default when specifying -O3.
10502 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10504 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
10505 iterator. Delete separate integer-mode vec_set<mode> expander.
10506 (aarch64_simd_vec_setv2di): Delete.
10507 (vec_setv2di): Delete.
10508 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
10509 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
10510 the "w, r" alternative.
10512 2018-05-18 Martin Liska <mliska@suse.cz>
10514 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
10515 * tree-pass.h (make_pass_lower_switch_O0): New function.
10516 * tree-switch-conversion.c (node_has_low_bound): Remove.
10517 (node_has_high_bound): Likewise.
10518 (node_is_bounded): Likewise.
10519 (class pass_lower_switch): Make it a template type and create
10521 (pass_lower_switch::execute): Add template argument.
10522 (make_pass_lower_switch): New function.
10523 (make_pass_lower_switch_O0): New function.
10524 (do_jump_if_equal): Remove.
10525 (emit_case_nodes): Simplify to just handle all 3 cases and leave
10526 all the hard work to tree optimization passes.
10528 2018-05-18 Martin Liska <mliska@suse.cz>
10530 * dbgcnt.c (limit_low): Renamed from limit.
10531 (limit_high): New variable.
10532 (dbg_cnt_is_enabled): Check for upper limit.
10533 (dbg_cnt): Adjust dumping.
10534 (dbg_cnt_set_limit_by_index): Add new argument for high
10536 (dbg_cnt_set_limit_by_name): Likewise.
10537 (dbg_cnt_process_single_pair): Parse new format.
10538 (dbg_cnt_process_opt): Use strtok.
10539 (dbg_cnt_list_all_counters): Remove 'value' and add
10541 * doc/invoke.texi: Document changes.
10543 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
10545 * doc/sourcebuild.texi (scalar_all_fma): Document.
10546 * tree.def (FMA_EXPR): Delete.
10547 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
10548 * internal-fn.c (ternary_direct): New macro.
10549 (expand_ternary_optab_fn): Likewise.
10550 (direct_ternary_optab_supported_p): Likewise.
10551 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
10552 * builtins.c (fold_builtin_fma): Delete.
10553 (fold_builtin_3): Don't call it.
10554 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
10555 * expr.c (expand_expr_real_2): Likewise.
10556 * fold-const.c (operand_equal_p): Likewise.
10557 (fold_ternary_loc): Likewise.
10558 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
10559 * gimple.c (DEFTREECODE): Likewise.
10560 * gimplify.c (gimplify_expr): Likewise.
10561 * optabs-tree.c (optab_for_tree_code): Likewise.
10562 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
10563 * tree-eh.c (operation_could_trap_p): Likewise.
10564 (stmt_could_throw_1_p): Likewise.
10565 * tree-inline.c (estimate_operator_cost): Likewise.
10566 * tree-pretty-print.c (dump_generic_node): Likewise.
10567 (op_code_prio): Likewise.
10568 * tree-ssa-loop-im.c (stmt_cost): Likewise.
10569 * tree-ssa-operands.c (get_expr_operands): Likewise.
10570 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
10571 * fold-const-call.h (fold_fma): Delete.
10572 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
10573 CFN_FNMA and CFN_FNMS.
10574 (fold_fma): Delete.
10575 * genmatch.c (combined_fn): New enum.
10576 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
10577 (commutative_op): New function.
10578 (commutate): Use it. Handle more than 2 operands.
10579 (dt_operand::gen_gimple_expr): Use commutative_op.
10580 (parser::parse_expr): Allow :c to be used with non-binary
10581 operators if the commutative operand is known.
10582 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
10583 CFN_FMS, CFN_FNMA and CFN_FNMS.
10584 (backprop::process_assign_use): Remove FMA_EXPR handling.
10585 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
10586 (gen_hsa_fma): New function.
10587 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
10588 IFN_FNMA and IFN_FNMS.
10589 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
10590 * gimple-fold.h (follow_all_ssa_edges): Declare.
10591 * gimple-fold.c (follow_all_ssa_edges): New function.
10592 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
10593 gimple_build interface and use follow_all_ssa_edges to fold the result.
10594 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
10595 instead of checking for optabs directly.
10596 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
10597 rather than FMA_EXPRs.
10598 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
10599 call to IFN_FMA instead of an FMA_EXPR.
10601 2018-05-17 Jim Wilson <jimw@sifive.com>
10603 * expr.c (do_tablejump): When converting index to Pmode, if we have a
10604 sign extended promoted subreg, and the range does not have the sign bit
10605 set, then do a sign extend.
10607 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
10608 test, check for sign extended subreg and/or constant operands, and
10609 do a sign extend in that case.
10611 2018-05-17 Steve Ellcey <sellcey@cavium.com>
10613 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
10614 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
10616 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
10617 Change logics_shift_reg to logics_shift_imm.
10618 (thunderx2t99_fp_loadpair_basic): Delete.
10619 (thunderx2t99_fp_storepair_basic): Delete.
10620 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
10621 (thunderx2t99_asimd_polynomial): Delete.
10622 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
10623 and neon_fp_mul_d_scalar_q.
10624 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
10625 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
10626 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
10627 (thunderx2t99_asimd_lut): Add missing tbl types.
10628 (thunderx2t99_asimd_ext): Delete.
10629 (thunderx2t99_asimd_load1_1_mult): Delete.
10630 (thunderx2t99_asimd_load1_2_mult): Delete.
10631 (thunderx2t99_asimd_load1_ldp): New.
10632 (thunderx2t99_asimd_load1): New.
10633 (thunderx2t99_asimd_load2): Add missing *load2* types.
10634 (thunderx2t99_asimd_load3): New.
10635 (thunderx2t99_asimd_load4): New.
10636 (thunderx2t99_asimd_store1_1_mult): Delete.
10637 (thunderx2t99_asimd_store1_2_mult): Delete.
10638 (thunderx2t99_asimd_store2_mult): Delete.
10639 (thunderx2t99_asimd_store2_onelane): Delete.
10640 (thunderx2t99_asimd_store_stp): New.
10641 (thunderx2t99_asimd_store1): New.
10642 (thunderx2t99_asimd_store2): New.
10643 (thunderx2t99_asimd_store3): New.
10644 (thunderx2t99_asimd_store4): New.
10646 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
10648 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
10649 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
10651 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
10652 Segher Boessenkool <segher@kernel.crashing.org>
10655 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
10658 2018-05-17 Richard Biener <rguenther@suse.de>
10660 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
10661 for pruning loop and prune defs feeding only already visited PHIs.
10663 2018-05-17 Richard Biener <rguenther@suse.de>
10665 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
10667 2018-05-17 Bin Cheng <bin.cheng@arm.com>
10668 Richard Biener <rguenther@suse.de>
10670 PR tree-optimization/85793
10671 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
10672 for VMAT_ELEMENTWISE.
10674 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10676 * internal-fn.h (lookup_internal_fn): Declare
10677 * internal-fn.c (lookup_internal_fn): New function.
10678 * gimple.c (gimple_build_call_from_tree): Handle calls to
10679 internal functions.
10680 * gimple-pretty-print.c (dump_gimple_call): Print "." before
10681 internal function names.
10682 * tree-pretty-print.c (dump_generic_node): Likewise.
10683 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
10685 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10687 * gimple-fold.h (gimple_build): Make the function forms take
10688 combined_fn rather than built_in_function.
10689 (gimple_simplify): Likewise.
10690 * gimple-match-head.c (gimple_simplify): Likewise.
10691 * gimple-fold.c (gimple_build): Likewise.
10692 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
10693 rather than gimple_build_call_internal.
10694 (get_initial_defs_for_reduction): Likewise.
10695 (vect_create_epilog_for_reduction): Likewise.
10696 (vectorizable_live_operation): Likewise.
10698 2018-05-17 Martin Liska <mliska@suse.cz>
10700 * gimple-ssa-sprintf.c (format_directive): Do not use
10701 space in between 'G_' and '('.
10703 2018-05-17 Jakub Jelinek <jakub@redhat.com>
10706 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
10707 even if the mask is not all ones.
10710 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
10712 (ix86_gimple_fold_builtin): Likewise.
10715 * config/i386/i386.c: Include tree-vector-builder.h.
10716 (ix86_vector_shift_count): New function.
10717 (ix86_fold_builtin): Fold shift builtins by scalar count.
10718 (ix86_gimple_fold_builtin): Likewise.
10720 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
10721 _mm512_setzero): New intrinsics.
10723 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
10724 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10726 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
10727 code generation for cases where splatting a value is not useful.
10728 * simplify-rtx.c (simplify_ternary_operation): Simplify
10729 vec_merge across a vec_duplicate and a paradoxical subreg forming
10730 a vector mode to a vec_concat.
10732 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
10734 * config.gcc: Support "goldmont-plus".
10735 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10737 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10738 PROCESSOR_GOLDMONT_PLUS.
10739 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
10740 (processor_target_table): Add "goldmont-plus".
10741 (PTA_GOLDMONT_PLUS): Define.
10742 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
10743 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
10744 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
10745 (fold_builtin_cpu): Add "goldmont-plus".
10746 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
10747 (ix86_option_override_internal): Add "goldmont-plus".
10748 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
10749 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
10750 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
10751 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
10753 2018-05-17 Richard Biener <rguenther@suse.de>
10755 PR tree-optimization/85757
10756 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
10757 remove defs that only feed that PHI from further processing.
10759 2018-05-16 Jim Wilson <jimw@sifive.com>
10761 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
10763 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
10765 2018-05-16 Mark Wielaard <mark@klomp.org>
10767 * dwarf2out.c (count_index_strings): New function.
10768 (output_indirect_strings): Call count_index_strings and generate
10769 header for dwarf_version >= 5.
10771 2018-05-16 Mark Wielaard <mark@klomp.org>
10773 * dwarf2out.c (dwarf_FORM): New function.
10774 (set_indirect_string): Use dwarf_FORM.
10775 (reset_indirect_string): Likewise.
10776 (size_of_die): Likewise.
10777 (value_format): Likewise.
10778 (output_die): Likewise.
10779 (add_skeleton_AT_string): Likewise.
10780 (output_macinfo_op): Likewise.
10781 (index_string): Likewise.
10782 (output_index_string_offset): Likewise.
10783 (output_index_string): Likewise.
10784 (count_index_strings): Likewise.
10786 2018-05-16 Carl Love <cel@us.ibm.com>
10788 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10789 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
10791 2018-05-16 Martin Jambor <mjambor@suse.cz>
10793 * ipa-prop.c (ipa_free_all_edge_args): Remove.
10794 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
10796 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
10798 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
10799 (fnma<mode>4): Likewise.
10800 (fms<mode>4): Likewise.
10801 (fnms<mode>4): Likewise.
10802 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
10803 (aarch64_fnma<mode>4): Likewise.
10804 (aarch64_fms<mode>4): Likewise.
10805 (aarch64_fnms<mode>4): Likewise.
10806 (aarch64_fnmadd<mode>4): Likewise.
10808 2018-05-16 Jason Merrill <jason@redhat.com>
10810 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
10812 2018-05-16 Richard Biener <rguenther@suse.de>
10814 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
10815 (dump_stmt_cost): Declare.
10816 (add_stmt_cost): Dump cost we add.
10817 (add_stmt_costs): New function.
10818 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
10819 No longer exported.
10820 (vect_analyze_stmt): Adjust prototype.
10821 (vectorizable_condition): Likewise.
10822 (vectorizable_live_operation): Likewise.
10823 (vectorizable_reduction): Likewise.
10824 (vectorizable_induction): Likewise.
10825 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
10826 cost vector to pass to vectorizable_ and record afterwards.
10827 (vect_model_reduction_cost): Take cost vector argument and adjust.
10828 (vect_model_induction_cost): Likewise.
10829 (vectorizable_reduction): Likewise.
10830 (vectorizable_induction): Likewise.
10831 (vectorizable_live_operation): Likewise.
10832 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
10833 SLP_TREE_NUMBER_OF_VEC_STMTS.
10834 (vect_analyze_slp_cost_1): Remove.
10835 (vect_analyze_slp_cost): Likewise.
10836 (vect_slp_analyze_node_operations): Take visited args and
10837 a target cost vector. Avoid processing already visited stmt sets.
10838 (vect_slp_analyze_operations): Use a local cost vector to gather
10839 costs and register those of non-discarded instances.
10840 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
10841 (vect_schedule_slp_instance): Remove copying of
10842 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
10844 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
10845 adding cost. Record cost entry location.
10846 (vect_prologue_cost_for_slp_op): Function to compute cost of
10847 a constant or invariant generated for SLP vect in the prologue,
10848 split out from vect_analyze_slp_cost_1.
10849 (vect_model_simple_cost): Make static. Adjust for SLP costing.
10850 (vect_model_promotion_demotion_cost): Likewise.
10851 (vect_model_store_cost): Likewise, make static.
10852 (vect_model_load_cost): Likewise.
10853 (vectorizable_bswap): Add cost vector arg and adjust.
10854 (vectorizable_call): Likewise.
10855 (vectorizable_simd_clone_call): Likewise.
10856 (vectorizable_conversion): Likewise.
10857 (vectorizable_assignment): Likewise.
10858 (vectorizable_shift): Likewise.
10859 (vectorizable_operation): Likewise.
10860 (vectorizable_store): Likewise.
10861 (vectorizable_load): Likewise.
10862 (vectorizable_condition): Likewise.
10863 (vectorizable_comparison): Likewise.
10864 (can_vectorize_live_stmts): Likewise.
10865 (vect_analyze_stmt): Likewise.
10866 (vect_transform_stmt): Adjust calls to vectorizable_*.
10867 * tree-vectorizer.c: Include gimple-pretty-print.h.
10868 (dump_stmt_cost): New function.
10870 2018-05-16 Richard Biener <rguenther@suse.de>
10872 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
10873 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
10874 * tree-ssa-dse.c: Include tree-ssa-loop.h.
10875 (check_name): New callback.
10876 (dse_classify_store): Track cycles via a visited bitmap of PHI
10877 defs and simplify handling of in-loop and across loop dead stores
10878 and properly fail for loop-variant refs. Handle byte-tracking with
10879 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
10882 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
10884 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
10885 (vect_get_mask_type_for_stmt): Likewise.
10886 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
10888 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
10889 to determine the statement's vector type and the vector type that
10890 should be used for calculating nunits. Deal with cases in which
10891 the type has to be deferred.
10892 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
10893 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
10894 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
10895 (vect_determine_vf_for_stmt): New functions, split out from...
10896 (vect_determine_vectorization_factor): ...here.
10897 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
10898 (vect_get_mask_type_for_stmt): New functions, split out from
10899 vect_determine_vectorization_factor.
10901 2018-05-16 Richard Biener <rguenther@suse.de>
10903 * tree-cfg.c (verify_gimple_assign_ternary): Properly
10904 verify the [VEC_]COND_EXPR embedded comparison.
10906 2018-05-15 Martin Sebor <msebor@redhat.com>
10908 PR tree-optimization/85753
10909 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
10910 RECORD_TYPE in addition to ARRAY_TYPE.
10912 2018-05-15 Martin Sebor <msebor@redhat.com>
10914 PR middle-end/85643
10915 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
10917 2018-05-15 Richard Biener <rguenther@suse.de>
10919 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
10920 add by_clobber_p one. Change algorithm to collect all defs
10921 representing uses we need to walk and try reducing them to
10922 a single one before failing.
10923 (dse_dom_walker::dse_optimize_stmt): Adjust.
10925 2018-05-13 Mark Wielaard <mark@klomp.org>
10927 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
10928 (size_of_loc_descr): Likewise.
10929 (output_loc_operands): Likewise.
10930 (output_loc_operands_raw): Likewise.
10931 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
10932 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
10933 (hash_loc_operands): Likewise.
10934 (compare_loc_operands): Likewise.
10936 2018-05-14 Mark Wielaard <mark@klomp.org>
10938 * dwarf2out.c (count_index_addrs): New function.
10939 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
10941 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10943 PR tree-optimization/83648
10944 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
10945 return value as malloc candidate.
10947 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10950 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
10951 param as true in call to suggest_attribute.
10953 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
10955 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
10956 -mreadonly-in-sdata.
10958 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10960 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
10962 (aarch64_crypto_aesd_fused): Likewise.
10964 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
10966 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
10967 (movsi_aarch64): Likewise.
10968 (load_pairsi): Likewise.
10969 (load_pairdi): Likewise.
10970 (store_pairsi): Likewise.
10971 (store_pairdi): Likewise.
10972 (load_pairsf): Likewise.
10973 (load_pairdf): Likewise.
10974 (store_pairsf): Likewise.
10975 (store_pairdf): Likewise.
10976 (zero_extend): Likewise.
10977 (trunc): Swap alternatives.
10978 (fcvt_target): Add '?' to prefer w over r.
10980 2018-05-14 Jakub Jelinek <jakub@redhat.com>
10983 * config/i386/i386.md: Disallow non-commutative arithmetics in
10984 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
10985 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
10986 in the peephole2 before it.
10988 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
10990 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
10991 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
10992 (ix86_handle_option): Handle -mcldemote.
10993 * config.gcc: New header.
10994 * config/i386/cldemoteintrin.h: New file.
10995 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
10996 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10998 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10999 OPTION_MASK_ISA_CLDEMOTE.
11000 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
11001 (ix86_valid_target_attribute_inner_p): Ditto.
11002 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
11003 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
11004 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
11005 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
11006 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
11008 * config/i386/i386.opt: Add -mcldemote.
11009 * config/i386/x86intrin.h: New header.
11010 * doc/invoke.texi: Add -mcldemote.
11012 2018-05-14 Richard Biener <rguenther@suse.de>
11014 * doc/match-and-simplify.texi: Adjust :s documentation.
11016 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
11018 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
11019 intended memcpy size.
11020 (REORDER_45): Likewise.
11022 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
11024 * sort.cc: New file.
11025 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
11026 * vec.c (qsort_chk): Use gcc_qsort.
11027 * Makefile.in (OBJS-libcommon): Add sort.o.
11028 (build/sort.o): New target. Use it...
11029 (BUILD_RTL): ... here, and...
11030 (build/gencfn-macros): ... here, and...
11031 (build/genmatch): ... here.
11033 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
11034 Chung-Ju Wu <jasonwucj@gmail.com>
11036 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
11037 * config/nds32/nds32-graywolf.md: New file.
11038 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
11039 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
11041 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
11042 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
11043 * config/nds32/nds32.md (pipeline_model): Add graywolf.
11044 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
11045 * config/nds32/pipelines.md: Include n15 settings.
11047 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
11048 Chung-Ju Wu <jasonwucj@gmail.com>
11050 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
11051 * config/nds32/nds32-n13.md: New file.
11052 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
11053 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
11055 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
11056 * config/nds32/nds32.md (pipeline_model): Add n13.
11057 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
11058 * config/nds32/pipelines.md: Include n13 settings.
11060 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
11061 Chung-Ju Wu <jasonwucj@gmail.com>
11063 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
11064 * config/nds32/nds32-n10.md: New file.
11065 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
11066 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
11068 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
11069 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
11070 * config/nds32/nds32.md (pipeline_model): Add n10.
11071 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
11072 * config/nds32/pipelines.md: Include n10 settings.
11074 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
11075 Kito Cheng <kito.cheng@gmail.com>
11076 Chung-Ju Wu <jasonwucj@gmail.com>
11078 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
11079 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11080 Add enum values for DSP extension instructions.
11081 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
11083 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
11084 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
11085 New code iterators.
11086 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
11087 * config/nds32/nds32-dspext.md: New file for DSP implementation.
11088 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
11089 * config/nds32/nds32-intrinsic.md: Likewise.
11090 * config/nds32/nds32_intrinsic.h: Likewise.
11091 * config/nds32/nds32-md-auxiliary.c: Likewise.
11092 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
11093 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
11094 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
11095 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
11096 * config/nds32/nds32-protos.h: New declarations for DSP extension.
11097 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
11098 TYPE_DMAC in switch statement.
11099 * config/nds32/nds32.c: New checking and implementation for DSP
11100 extension instructions.
11101 * config/nds32/nds32.h: Likewise.
11102 * config/nds32/nds32.md: Likewise.
11103 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
11104 * config/nds32/predicates.md: Implement new predicates for DSP
11107 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11109 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
11110 Reformat alternatives and attributes so it is easier to identify
11111 which constraints/attributes go with which instruction.
11112 (mov<mode>_hardfloat32, FMOVE64): Likewise.
11113 (mov<mode>_softfloat32, FMOVE64): Likewise.
11114 (mov<mode>_hardfloat64, FMOVE64): Likewise.
11115 (mov<mode>_softfloat64, FMOVE64): Likewise.
11117 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
11119 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11121 (Basic PowerPC Built-in Functions): The new name of the
11122 subsection previously known as "PowerPC Built-in Functions".
11123 (Basic PowerPC Built-in Functions Available on all Configurations):
11125 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
11126 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
11127 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
11128 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
11130 2018-05-11 Martin Jambor <mjambor@suse.cz>
11133 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
11136 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
11139 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
11141 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
11143 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
11144 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
11145 (ix86_handle_option): Handle -mwaitpkg.
11146 * config.gcc: New header.
11147 * config/i386/cpuid.h (bit_WAITPKG): New bit.
11148 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
11149 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
11151 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11152 OPTION_MASK_ISA_WAITPKG.
11153 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
11154 (ix86_option_override_internal): Add PTA_WAITPKG.
11155 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
11156 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
11157 IX86_BUILTIN_TPAUSE.
11158 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
11159 __builtin_ia32_umwait and __builtin_ia32_tpause.
11160 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
11161 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
11162 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
11163 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
11164 UNSPECV_TPAUSE): New.
11165 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
11166 * config/i386/i386.opt: Add -mwaitpkg.
11167 * config/i386/waitpkgintrin.h: New file.
11168 * config/i386/x86intrin.h: New header.
11169 * doc/invoke.texi: Add -mwaitpkg.
11171 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
11174 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
11176 (cortex-m0): Use armv6s-m isa.
11177 (cortex-m0plus): Likewise.
11178 (cortex-m1): Likewise.
11179 (cortex-m0.small-multiply): Likewise.
11180 (cortex-m0plus.small-multiply): Likewise.
11181 (cortex-m1.small-multiply): Likewise.
11183 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
11184 Jakub Jelinek <jakub@redhat.com>
11186 PR tree-optimization/85692
11187 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
11188 source permute as well.
11190 2018-05-11 Martin Liska <mliska@suse.cz>
11193 * doc/extend.texi: Document LLVM style format for no_sanitize
11196 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11198 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
11199 mode_supports_vsx_dform_quad to mode_supports_dq_form.
11200 (mode_supports_vsx_dform_quad): Likewise.
11201 (mode_supports_vmx_dform): Move these functions to be next to the
11202 other mode_supports functions.
11203 (mode_supports_dq_form): Likewise.
11204 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
11205 mode_supports_dq_form.
11206 (reg_offset_addressing_ok_p): Likewise.
11207 (offsettable_ok_by_alignment): Likewise.
11208 (rs6000_legitimate_offset_address_p): Likewise.
11209 (legitimate_lo_sum_address_p): Likewise.
11210 (rs6000_legitimize_address): Likewise.
11211 (rs6000_legitimize_reload_address): Likewise.
11212 (rs6000_secondary_reload_inner): Likewise.
11213 (rs6000_preferred_reload_class): Likewise.
11214 (rs6000_output_move_128bit): Likewise.
11216 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
11218 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
11219 Generate SImode target register for null target.
11220 <case IX86_BUILTIN_XGETBV>: Ditto.
11221 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
11222 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
11224 2018-05-10 Carl Love <cel@us.ibm.com>
11226 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
11227 dcbtt and dcbtstt if operands[2] is 0.
11229 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
11232 * config/i386/sse.md (usadv64qi): New expander.
11234 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
11236 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
11237 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
11238 -maltivec=be support.
11239 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
11240 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
11241 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
11242 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
11243 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
11244 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
11245 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
11246 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
11247 altivec_vsumsws): Adjust.
11248 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
11249 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
11250 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
11252 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
11253 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
11254 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
11255 (altivec_lve<VI_char>x): Delete expand.
11256 (*altivec_lve<VI_char>x_internal): Rename to...
11257 (altivec_lve<VI_char>x): ... this.
11258 (altivec_lvxl_<mode>): Delete expand.
11259 (*altivec_lvxl_<mode>_internal): Rename to ...
11260 (altivec_lvxl_<mode>): ... this.
11261 (altivec_stvxl_<mode>): Delete expand.
11262 (*altivec_stvxl_<mode>_internal): Rename to ...
11263 (altivec_stvxl_<mode>): ... this.
11264 (altivec_stve<VI_char>x): Delete expand.
11265 (*altivec_stve<VI_char>x_internal): Rename to ...
11266 (altivec_stve<VI_char>x): ... this.
11267 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
11268 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
11269 reduc_plus_scal_<mode>): Adjust.
11270 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
11272 (rs6000_cpu_cpp_builtins): Adjust.
11273 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
11274 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
11275 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
11276 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
11277 -maltivec=be support.
11278 (rs6000_split_vec_extract_var): Adjust.
11279 (rs6000_split_v4si_init): Adjust.
11280 (swap_selector_for_mode): Delete.
11281 (altivec_expand_lvx_be, altivec_expand_stvx_be,
11282 altivec_expand_stvex_be): Delete.
11283 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
11284 -maltivec=be support.
11285 (rs6000_gimple_fold_builtin): Ditto.
11286 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
11288 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
11289 (TARGET_DIRECT_MOVE_64BIT): Adjust.
11290 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
11291 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
11292 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
11293 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
11294 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
11295 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
11296 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
11297 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
11298 anonymous split): Adjust.
11299 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
11300 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
11302 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
11304 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
11305 when --with-gxx-include-dir is also specified.
11306 * configure: Regenerate.
11308 2018-05-09 Jim Wilson <jimw@sifive.com>
11311 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
11312 * config/riscv/t-withmultilib: New.
11313 * config/riscv/withmultilib.h: New.
11314 * doc/install.texi: Document RISC-V --with-multilib-list support.
11316 2018-05-09 Richard Biener <rguenther@suse.de>
11318 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
11320 (vect_bb_vectorization_profitable_p): Adjust. Compute
11321 actual scalar cost using the cost vector and the add_stmt_cost
11324 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11326 PR rtl-optimization/85645
11327 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
11328 in the REG_CFA_REGISTER note for LR, don't leave it empty.
11330 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11332 PR rtl-optimization/85645
11333 * shrink-wrap.c (spread_components): Return a boolean saying if
11334 anything was changed.
11335 (try_shrink_wrapping_separate): Iterate spread_components until
11336 nothing changes anymore.
11338 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11340 PR rtl-optimization/85645
11341 * regrename.c (build_def_use): Also kill the chains that include the
11342 destination of a REG_CFA_REGISTER note.
11344 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11346 PR rtl-optimization/85645
11347 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
11348 insn that has a REG_CFA_REGISTER note.
11350 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
11352 * cfgexpand.c (expand_clobber): New function.
11353 (expand_gimple_stmt_1): Use it.
11354 * tree-vect-stmts.c (vect_clobber_variable): New function,
11356 (vectorizable_simd_clone_call): ...here.
11357 (vectorizable_store): Emit a clobber either side of an
11358 IFN_STORE_LANES sequence.
11359 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
11361 2018-05-09 Tom de Vries <tom@codesourcery.com>
11364 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
11365 (define_insn "trap_if_false"): Add exit after trap.
11367 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
11369 PR rtl-optimization/85638
11370 * bb-reorder.c: Include common/common-target.h.
11371 (create_forwarder_block): New function extracted from...
11372 (fix_up_crossing_landing_pad): ...here. Rename into...
11373 (dw2_fix_up_crossing_landing_pad): ...this.
11374 (sjlj_fix_up_crossing_landing_pad): New function.
11375 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
11376 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
11377 from both partitions and exit the loop after one iteration.
11379 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
11382 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11384 (Basic PowerPC Built-in Functions): The new name of the
11385 subsection previously known as "PowerPC Built-in Functions".
11386 (Basic PowerPC Built-in Functions Available on all Configurations):
11388 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
11390 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
11392 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
11394 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
11397 2018-05-08 Jim Wilson <jimw@sifive.com>
11399 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
11400 (LD_EMUL_SUFFIX): New.
11401 (LINK_SPEC): Use it.
11403 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
11405 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11407 (Basic PowerPC Built-in Functions): The new name of the
11408 subsection previously known as "PowerPC Built-in Functions".
11409 (Basic PowerPC Built-in Functions Available on all Configurations):
11411 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
11413 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
11415 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
11417 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
11420 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11423 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
11424 after cmpelim optimization.
11426 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
11428 * config.gcc: Support "goldmont".
11429 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
11430 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11431 PROCESSOR_GOLDMONT.
11432 * config/i386/i386.c (m_GOLDMONT): Define.
11433 (processor_target_table): Add "goldmont".
11434 (PTA_GOLDMONT): Define.
11435 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
11436 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
11437 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
11438 (fold_builtin_cpu): Add "goldmont".
11439 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
11440 (ix86_option_override_internal): Add "goldmont".
11441 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
11442 (processor_type): Add PROCESSOR_GOLDMONT.
11443 * config/i386/i386.md: Add CPU "glm".
11444 * config/i386/glm.md: New file.
11445 * config/i386/x86-tune.def: Add m_GOLDMONT.
11446 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
11448 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11451 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
11453 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
11454 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
11455 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
11458 * config/i386/i386.c (ix86_fold_builtin): Handle
11459 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
11462 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
11463 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
11465 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
11468 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
11469 (check_arch): Likewise.
11470 (check_fpu): Return the result rather than printing it.
11471 (end arch): Fix operator precedence.
11472 (end cpu): Likewise.
11473 (END): Print the result from check_fpu.
11475 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
11476 Alan Hayward <alan.hayward@arm.com>
11477 David Sherwood <david.sherwood@arm.com>
11479 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
11480 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
11481 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
11482 (*fcmuo<mode>_and): New patterns.
11484 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
11486 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
11487 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
11488 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
11489 (cmp_op, sve_imm_con): New code attributes.
11490 (SVE_COND_INT_CMP, imm_con): Delete.
11491 (cmp_op): Remove above unspecs from int attribute.
11492 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
11494 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
11495 comparison-specific unspecs.
11496 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
11497 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
11498 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
11499 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
11500 (*vec_fcm<cmp_op><mode>): Rename to...
11501 (*fcm<cmp_op><mode>): ...this and adjust likewise.
11502 (*vec_fcmuo<mode>): Rename to...
11503 (*fcmuo<mode>): ...this and adjust likewise.
11504 (*pred_fcm<cmp_op><mode>): New pattern.
11505 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
11506 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
11508 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
11510 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
11511 (aarch64_emit_sve_predicated_cond): New function.
11512 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
11513 (aarch64_emit_unspec_cond_or): Replace with...
11514 (aarch64_emit_sve_or_conds): ...this new function. Use
11515 aarch64_emit_sve_ptrue_op for the individual comparisons and
11516 aarch64_emit_binop to OR them together.
11517 (aarch64_emit_inverted_unspec_cond): Replace with...
11518 (aarch64_emit_sve_inverted_cond): ...this new function. Use
11519 aarch64_emit_sve_ptrue_op for the comparison and
11520 aarch64_emit_unop to invert the result.
11521 (aarch64_expand_sve_vec_cmp_float): Update after the above
11522 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
11524 2018-05-07 Nathan Sidwell <nathan@acm.org>
11526 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
11527 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
11528 (Backwards Compatibility): Likewise.
11530 2018-05-07 Luis Machado <luis.machado@linaro.org>
11534 2018-05-07 Luis Machado <luis.machado@linaro.org>
11536 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11537 <prefetch_dynamic_strides>: New const bool field.
11538 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11539 prefetch_dynamic_strides.
11540 (exynosm1_prefetch_tune): Likewise.
11541 (thunderxt88_prefetch_tune): Likewise.
11542 (thunderx_prefetch_tune): Likewise.
11543 (thunderx2t99_prefetch_tune): Likewise.
11544 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11546 (aarch64_override_options_internal): Update to set
11547 PARAM_PREFETCH_DYNAMIC_STRIDES.
11548 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11549 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11550 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11551 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11552 prefetch-dynamic-strides setting.
11554 2018-05-07 Luis Machado <luis.machado@linaro.org>
11556 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11557 <minimum_stride>: New const int field.
11558 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11559 minimum_stride field.
11560 (exynosm1_prefetch_tune): Likewise.
11561 (thunderxt88_prefetch_tune): Likewise.
11562 (thunderx_prefetch_tune): Likewise.
11563 (thunderx2t99_prefetch_tune): Likewise.
11564 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11565 (aarch64_override_options_internal): Update to set
11566 PARAM_PREFETCH_MINIMUM_STRIDE.
11567 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11568 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11569 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11570 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11571 stride is constant and is below the minimum stride threshold.
11573 2018-05-07 Luis Machado <luis.machado@linaro.org>
11575 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
11578 2018-05-07 Luis Machado <luis.machado@linaro.org>
11580 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11581 <prefetch_dynamic_strides>: New const bool field.
11582 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11583 prefetch_dynamic_strides.
11584 (exynosm1_prefetch_tune): Likewise.
11585 (thunderxt88_prefetch_tune): Likewise.
11586 (thunderx_prefetch_tune): Likewise.
11587 (thunderx2t99_prefetch_tune): Likewise.
11588 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11590 (aarch64_override_options_internal): Update to set
11591 PARAM_PREFETCH_DYNAMIC_STRIDES.
11592 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11593 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11594 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11595 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11596 prefetch-dynamic-strides setting.
11598 2018-05-07 Luis Machado <luis.machado@linaro.org>
11600 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11601 <minimum_stride>: New const int field.
11602 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11603 minimum_stride field.
11604 (exynosm1_prefetch_tune): Likewise.
11605 (thunderxt88_prefetch_tune): Likewise.
11606 (thunderx_prefetch_tune): Likewise.
11607 (thunderx2t99_prefetch_tune): Likewise.
11608 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11609 (aarch64_override_options_internal): Update to set
11610 PARAM_PREFETCH_MINIMUM_STRIDE.
11611 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11612 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11613 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11614 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11615 stride is constant and is below the minimum stride threshold.
11617 2018-05-06 Jakub Jelinek <jakub@redhat.com>
11620 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
11621 the type is addressable. Don't force op into register if it has
11624 2018-05-05 Roland McGrath <mcgrathr@google.com>
11627 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
11628 any section for which we don't know a specific type it should have,
11629 regardless of name. Previously this was done only for the exact
11630 names ".init_array", ".fini_array", and ".preinit_array".
11631 (default_elf_asm_named_section): Add comment about
11632 relationship with default_section_type_flags and SECTION_NOTYPE.
11633 (get_section): Don't consider it a type conflict if one side has
11634 SECTION_NOTYPE and the other doesn't, as long as neither has the
11635 SECTION_BSS et al used in the default_section_type_flags logic.
11637 2018-05-05 Tom de Vries <tom@codesourcery.com>
11640 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
11641 (workaround_barsyncs): New function.
11642 (nvptx_reorg): Use workaround_barsyncs.
11643 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11644 (define_expand "nvptx_membar_cta"): New define_expand.
11645 (define_insn "*nvptx_membar_cta"): New insn.
11647 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
11649 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
11650 To improve optimization opportunities.
11651 * builtin-types.def: The new needed builtin types for the above.
11653 2018-05-04 Richard Biener <rguenther@suse.de>
11655 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
11656 * gimple-ssa-store-merging.c
11657 (imm_store_chain_info::output_merged_store): Remove redundant create,
11658 release split_store vector contents on failure.
11659 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
11660 scalar stmt vector on cache hit.
11662 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
11664 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
11666 * config.gcc (powerpc-xilinx-eabi*): Remove.
11667 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
11669 (fusion_addis_mem_combo_load): Ditto.
11670 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
11672 (rs6000_cpu_cpp_builtins): Ditto.
11673 * config/rs6000/rs6000-linux.c
11674 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
11675 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
11676 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
11678 (rs6000_setup_reg_addr_masks): Ditto.
11679 (rs6000_init_hard_regno_mode_ok): Ditto.
11680 (rs6000_option_override_internal): Ditto.
11681 (legitimate_lo_sum_address_p): Ditto.
11682 (rs6000_legitimize_address): Ditto.
11683 (rs6000_legitimize_reload_address): Ditto.
11684 (rs6000_legitimate_address_p): Ditto.
11685 (abi_v4_pass_in_fpr): Ditto.
11686 (setup_incoming_varargs): Ditto.
11687 (rs6000_gimplify_va_arg): Ditto.
11688 (rs6000_split_multireg_move): Ditto.
11689 (rs6000_savres_strategy): Ditto.
11690 (rs6000_emit_prologue_components): Ditto.
11691 (rs6000_emit_epilogue_components): Ditto.
11692 (rs6000_emit_prologue): Ditto.
11693 (rs6000_emit_epilogue): Ditto.
11694 (rs6000_elf_file_end): Ditto.
11695 (rs6000_function_value): Ditto.
11696 (rs6000_libcall_value): Ditto.
11697 * config/rs6000/rs6000.h: Ditto.
11698 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
11699 (TARGET_MINMAX): ... this. New.
11700 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
11701 * config/rs6000/rs6000.md: Remove Xilinx FP support.
11702 (*movsi_internal1_single): Delete.
11703 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
11704 mfpu=, mxilinx-fpu): Delete.
11705 * config/rs6000/singlefp.h: Delete.
11706 * config/rs6000/sysv4.h: Remove Xilinx FP support.
11707 * config/rs6000/t-rs6000: Ditto.
11708 * config/rs6000/t-xilinx: Delete.
11709 * config/rs6000/titan.md: Adjust for fp_type removal.
11710 * config/rs6000/vsx.md: Remove Xilinx FP support.
11711 (VStype_simple): Delete.
11712 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
11713 * config/rs6000/xfpu.h: Delete.
11714 * config/rs6000/xfpu.md: Delete.
11715 * config/rs6000/xilinx.h: Delete.
11716 * config/rs6000/xilinx.opt: Delete.
11717 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
11718 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
11720 2018-05-04 Tom de Vries <tom@codesourcery.com>
11723 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
11726 2018-05-04 Richard Biener <rguenther@suse.de>
11728 PR middle-end/85627
11729 * tree-complex.c (update_complex_assignment): We are always in SSA form.
11730 (expand_complex_div_wide): Likewise.
11731 (expand_complex_operations_1): Likewise.
11732 (expand_complex_libcall): Preserve EH info of the original stmt.
11733 (tree_lower_complex): Handle removed blocks.
11734 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
11735 on complex multiplication and division libcall builtins.
11737 2018-05-04 Richard Biener <rguenther@suse.de>
11739 PR middle-end/85574
11740 * fold-const.c (negate_expr_p): Restrict negation of operand
11741 zero of a division to when we know that can happen without
11743 (fold_negate_expr_1): Likewise.
11745 2018-05-04 Jakub Jelinek <jakub@redhat.com>
11748 * real.h (real_nextafter): Declare.
11749 * real.c (real_nextafter): New function.
11750 * fold-const-call.c (fold_const_nextafter): New function.
11751 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
11752 CASE_CFN_NEXTTOWARD.
11753 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
11754 even when arg1_mode is different from arg0_mode.
11756 2018-05-03 Nathan Sidwell <nathan@acm.org>
11758 * doc/extend.texi (Deprecated Features): Remove
11759 -ffriend-injection.
11760 (Backwards Compatibility): Likewise.
11761 * doc/invoke.texi (C++ Language Options): Likewise.
11762 (C++ Dialect Options): Likewise.
11764 2018-05-03 Jakub Jelinek <jakub@redhat.com>
11767 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
11768 _mm512_mask_mullox_epi64): New intrinsics.
11770 2018-05-03 Tom de Vries <tom@codesourcery.com>
11773 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11774 dump files): Add offload-tree.
11776 2018-05-03 Richard Biener <rguenther@suse.de>
11778 PR tree-optimization/85615
11779 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
11780 to loops not nested in BBs loop father to avoid creating multi-entry
11783 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11785 PR tree-optimization/70291
11786 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
11787 arguments. Change return type to tree. Emit libcall as a new
11788 statement rather than replacing existing one when inplace_p is true.
11789 (expand_complex_multiplication_components): New function.
11790 (expand_complex_multiplication): Expand floating-point complex
11791 multiplication using the above.
11792 (expand_complex_division): Rename inner_type parameter to type.
11793 Update expand_complex_libcall call-site.
11794 (expand_complex_operations_1): Update expand_complex_multiplication
11795 and expand_complex_division call-sites.
11797 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11800 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11801 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11802 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
11803 the highest significant bit of the shift count mask is clear. In
11804 check whether and[sq]i3 is needed verify that all significant bits
11805 of the shift count other than the highest are set.
11807 2018-05-02 Tom de Vries <tom@codesourcery.com>
11810 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
11811 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
11812 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
11813 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
11814 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
11815 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
11816 __builtin_goacc_parlevel_size.
11818 2018-05-02 Richard Biener <rguenther@suse.de>
11820 PR tree-optimization/85597
11821 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
11822 do not use split vect_get_vec_defs call but call vect_get_slp_defs
11825 2018-05-02 Tom de Vries <tom@codesourcery.com>
11828 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11829 dump files): Add ltrans-tree.
11831 2018-05-02 Tom de Vries <tom@codesourcery.com>
11834 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11835 dump files): Add wpa-ipa.
11837 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
11839 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
11840 powerpc*-*-linux*paired* target.
11841 * config/rs6000/750cl.h: Delete.
11842 * config/rs6000/paired.h: Delete.
11843 * config/rs6000/paired.md: Delete.
11844 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
11846 * config/rs6000/rs6000-builtin.def: Remove paired float support.
11847 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
11848 comment. Remove paired float support.
11849 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
11850 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
11852 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
11853 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
11855 * config/rs6000/rs6000.c: Remove paired float support.
11856 (paired_expand_vector_init, paired_expand_vector_move,
11857 paired_emit_vector_compare, paired_emit_vector_cond_expr,
11858 (paired_expand_lv_builtin, paired_expand_stv_builtin,
11859 paired_expand_builtin, paired_expand_predicate_builtin,
11860 paired_init_builtins): Delete.
11861 * config/rs6000/rs6000.h: Remove paired float support.
11862 * config/rs6000/rs6000.md: Remove paired float support.
11863 (move_from_CR_ov_bit): Delete.
11864 * config/rs6000/rs6000.opt (mpaired): Delete.
11865 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
11866 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
11868 2018-05-02 Richard Biener <rguenther@suse.de>
11870 PR middle-end/85567
11871 * gimplify.c (gimplify_save_expr): When in SSA form allow
11872 SAVE_EXPRs to compute to SSA vars.
11874 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11877 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11878 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11879 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
11880 clobber operands[2], instead use a new pseudo. Formatting fixes.
11882 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
11884 PR tree-optimization/85586
11885 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
11886 exit early for statements in the same group if the accesses are
11889 2018-05-02 Tom de Vries <tom@codesourcery.com>
11892 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
11895 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
11897 PR tree-optimization/85143
11898 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
11900 2018-05-01 Tom de Vries <tom@codesourcery.com>
11903 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
11904 not found" error message.
11906 2018-05-01 Tom de Vries <tom@codesourcery.com>
11909 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
11910 * vec.c (test_ordered_remove_if): New function.
11911 (vec_c_tests): Call test_ordered_remove_if.
11912 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
11913 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
11914 * tree-vect-patterns.c (vect_pattern_recog_1): Use
11915 VEC_ORDERED_REMOVE_IF.
11917 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11919 PR tree-optimization/82665
11920 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
11921 pointer subtraction where arguments come from a memchr call.
11923 2018-05-01 Jakub Jelinek <jakub@redhat.com>
11925 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
11926 --push-state --as-needed and --pop-state instead of --as-needed and
11927 --no-as-needed if ld supports it.
11928 * configure: Regenerated.
11931 * doc/install.texi2html: Replace _002d with - and _002a with * in
11932 generated html files using sed.
11934 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11937 * gcc-rich-location.c (blank_line_before_p): New function.
11938 (use_new_line): New function.
11939 (gcc_rich_location::add_fixit_insert_formatted): New function.
11940 * gcc-rich-location.h
11941 (gcc_rich_location::add_fixit_insert_formatted): New function.
11943 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11945 * selftest.c (assert_streq): Rename "expected" and "actual" to
11946 "val1" and "val2". Extend NULL-handling to cover both inputs
11947 symmetrically, while still requiring both to be non-NULL for a pass.
11948 * selftest.h (assert_streq): Rename "expected" and "actual" to
11950 (ASSERT_EQ): Likewise.
11951 (ASSERT_EQ_AT): Likewise.
11952 (ASSERT_KNOWN_EQ): Likewise.
11953 (ASSERT_KNOWN_EQ_AT): Likewise.
11954 (ASSERT_NE): Likewise.
11955 (ASSERT_MAYBE_NE): Likewise.
11956 (ASSERT_MAYBE_NE_AT): Likewise.
11957 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
11958 the assertion to pass.
11959 (ASSERT_STREQ_AT): Likewise.
11961 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11963 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
11964 interaction with -pie.
11966 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11968 * selftest.h: Fix alphabetization of per-source-file selftest
11971 2018-04-30 Jason Merrill <jason@redhat.com>
11973 PR c++/61982 - dead stores to destroyed objects.
11974 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
11977 2018-04-30 Jason Merrill <jason@redhat.com>
11979 * tree.c (build_clobber): New.
11980 * tree.h: Declare it.
11981 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
11983 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11985 * diagnostic-show-locus.c (layout::layout): Update for
11986 location_get_source_line returning a char_span.
11987 (struct char_span): Move to input.h.
11988 (struct correction): Update for fields in char_span becoming
11990 (struct source_line): Update for location_get_source_line
11991 returning a char_span.
11992 (layout::print_line): Likewise.
11993 * edit-context.c (edited_file::print_content): Likewise.
11994 (edited_file::print_diff_hunk): Likewise.
11995 (edited_file::print_run_of_changed_lines): Likewise.
11996 (edited_file::get_num_lines): Likewise.
11997 (edited_line::edited_line): Likewise.
11998 * final.c (asm_show_source): Likewise.
11999 * input.c (location_get_source_line): Convert return type
12000 from const char * to char_span, losing the final "line_len"
12002 (dump_location_info): Update for the above.
12003 (get_substring_ranges_for_loc): Likewise. Use a char_span
12004 when handling the literal within the line.
12005 (test_reading_source_line): Update for location_get_source_line
12006 returning a char_span.
12007 * input.h (class char_span): Move here from
12008 diagnostic-show-locus.c, converting from a struct to a class.
12009 Make data members private.
12010 (char_span::operator bool): New.
12011 (char_span::length): New.
12012 (char_span::get_buffer): New.
12013 (char_span::operator[]): New.
12014 (char_span::subspan): Make const.
12015 (char_span::xstrdup): New.
12016 (location_get_source_line): Convert return type from const char *
12017 to char_span, losing the final "line_size" param.
12019 2018-04-30 Jan Hubicka <jh@suse.cz>
12021 * lto-wrapper.c (ltrans_priorities): New static var.
12022 (cmp_priority): New.
12023 (run_gcc): Read priorities and if doing parallel build order
12024 the Makefile by them.
12026 2018-04-30 David Malcolm <dmalcolm@redhat.com>
12028 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
12030 2018-04-30 Richard Biener <rguenther@suse.de>
12032 * tree-cfg.c (verify_address): Remove base argument, add
12033 flag whether to check TREE_ADDRESSABLE and do that.
12034 (verify_expr): Remove.
12035 (verify_types_in_gimple_reference): Add pieces from verify_expr.
12036 (verify_gimple_assign_single): Likewise.
12037 (verify_gimple_switch): Likewise.
12038 (verify_expr_location_1): Dereference tp once. Add (disabled)
12039 piece from verify_expr.
12040 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
12042 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
12044 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
12046 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
12048 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
12049 (small_data_pattern): Likewise.
12050 (arc_rewrite_small_data): Likewise.
12051 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
12052 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
12053 (get_symbol_alignment): New function.
12054 (legitimate_small_data_address_p): Likewise.
12055 (legitimate_scaled_address): Update, call
12056 legitimate_small_data_address_p.
12057 (output_sdata): New static variable.
12058 (arc_print_operand): Update how we handle small data operands.
12059 (arc_print_operand_address): Likewise.
12060 (arc_legitimate_address_p): Update, use
12061 legitimate_small_data_address_p.
12062 (arc_rewrite_small_data_p): Remove.
12063 (arc_rewrite_small_data_1): Likewise.
12064 (arc_rewrite_small_data): Likewise.
12065 (small_data_pattern): Likewise.
12066 (compact_sda_memory_operand): Update to use
12067 legitimate_small_data_address_p and get_symbol_alignment.
12068 (prepare_move_operands): Don't rewite sdata pattern.
12069 (prepare_extend_operands): Remove.
12070 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
12072 (zero_extendqisi2): Likewise.
12073 (zero_extendhisi2): Likewise.
12074 (extendqihi2): Likewise.
12075 (extendqisi2): Likewise.
12076 (extendhisi2): Likewise.
12077 (addsi3): Likewise.
12078 (subsi3): Likewise.
12079 (andsi3): Likewise.
12080 * config/arc/constraints.md (Usd): Change it to memory constraint.
12082 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
12084 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
12085 as source of std instructions.
12086 * config/arc/arc.md (movsi_insn): Update pattern predicate to
12087 allow 6-bit constants as source for store instructions.
12088 (movdi_insn): Update instruction pattern to allow 6-bit constants
12089 as source for store instructions.
12091 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
12093 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
12095 2018-04-30 Nathan Sidwell <nathan@acm.org>
12096 Sandra Loosemore <sandra@codesourcery.com>
12098 * dumpfile.c (dump_open): Allow '-' for stdout.
12099 * doc/invoke.texi (Developer Options): Document dump filename
12100 determination early. Document stdin/stdout selection.
12102 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
12104 Microblaze Target: PIC data text relative
12106 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
12107 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
12109 * config/microblaze/microblaze.h (microblaze_constant_address_p):
12110 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
12111 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
12112 New addressing mode for data-text relative position indepenedent code.
12113 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
12114 'ADDRESS_SYMBOLIC_TXT_REL'.
12115 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
12116 (microblaze_legitimate_pic_operand): Exclude function calls from
12117 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
12118 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
12120 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
12121 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
12122 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
12123 for 'address + offset'.
12124 (microblaze_expand_prologue): Add new function prologue call for
12126 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
12127 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
12128 table in case of TARGET_PIC_DATA_TEXT_REL.
12129 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
12130 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
12131 Add new macros 'UNSPEC_TEXT',
12132 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
12133 + exclude function calls from 'UNSPEC_PLT' in case of data text
12135 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
12136 new target hook for generating address diff vector tables in case of
12138 * doc/tm.texi : Regenerate.
12139 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
12140 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
12141 of addr diff vector generation.
12142 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
12143 target hook definition.
12144 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
12145 Add default function for generate_pic_addr_diff_vec -> flag_pic.
12146 * doc/invoke.texi (Add new pic option): Add new microblaze pic
12147 option for data text relative.
12149 2018-04-30 Richard Biener <rguenther@suse.de>
12151 * tree-chrec.h (evolution_function_is_constant_p): Remove
12153 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
12155 2018-04-30 Richard Biener <rguenther@suse.de>
12158 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
12160 2018-04-30 Richard Biener <rguenther@suse.de>
12162 PR tree-optimization/28364
12163 PR tree-optimization/85275
12164 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
12165 copying first exit test.
12167 2018-04-28 Mark Wielaard <mark@klomp.org>
12169 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
12170 dwarf_version >= 5.
12171 (dwarf_AT): Handle DW_AT_addr_base.
12172 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
12174 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
12177 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
12178 (*ashl<dwi>3_doubleword_mask_1): Ditto.
12179 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
12180 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
12182 2018-04-28 Richard Biener <rguenther@suse.de>
12184 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
12185 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
12186 to reflect use. Only add interesting stmts.
12188 2018-04-27 Martin Jambor <mjambor@suse.cz>
12191 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
12192 the jump function allows for passing through aggregate values.
12194 2018-04-27 David Malcolm <dmalcolm@redhat.com>
12196 * input.h (in_system_header_at): Convert from macro to inline
12198 (from_macro_expansion_at): Likewise.
12199 (from_macro_definition_at): Likewise.
12201 2018-04-27 Jeff Law <law@redhat.com>
12203 * config.gcc: Mark tile* targets as deprecated/obsolete.
12205 2018-04-27 Richard Biener <rguenther@suse.de>
12207 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
12210 2018-04-27 Richard Biener <rguenther@suse.de>
12212 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
12214 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
12216 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
12217 with Yd constraint. Set "preferred_for_speed" attribute from
12218 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
12219 with Yd constraint.
12220 (*movdi_internal): Ditto.
12221 (movti_interunit splitters): Remove
12222 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
12223 (movdi_interunit splitters): Ditto.
12224 * config/i386/constraints.md (Ye): Remove.
12225 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
12227 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12230 * config/aarch64/constraints.md (Usg): Limit to 31.
12231 (Usj): Limit to 63.
12233 2018-04-27 Jakub Jelinek <jakub@redhat.com>
12235 PR tree-optimization/85529
12236 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
12237 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
12238 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
12239 zero extension or masking of the MSB bit.
12240 (optimize_range_tests): Add FIRST_BB argument, pass it through
12241 to optimize_range_tests_var_bound.
12242 (maybe_optimize_range_tests, reassociate_bb): Adjust
12243 optimize_range_tests callers.
12245 2018-04-26 Richard Biener <rguenther@suse.de>
12246 Jakub Jelinek <jakub@redhat.com>
12248 * cgraph.h (symbol_table): Just declare debug method here.
12249 * symtab.c (symbol_table::debug): Define.
12251 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
12253 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
12255 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
12257 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
12258 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
12259 (*movdi_internal): Substitute Yi and Yj constraint with x
12260 and Ym and Yn constraint with y constraint. Update "isa"
12261 attribute and set "preferred_for_speed" attribute from
12262 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
12263 (*movsi_internal): Ditto.
12264 (*movdf_internal): Ditto.
12265 (*movsf_internal): Ditto.
12266 (*zero_extendsidi2): Ditto.
12267 * config/i386/sse.md (vec_set<mode>_0): Ditto.
12268 (sse2_loadld): Ditto.
12269 (*vec_extract<ssevecmodelower>_0): Ditto.
12270 (*vec_extractv4si_0_zext_sse4): Ditto.
12271 (vec_concatv2di): Ditto.
12272 (*vec_dup<mode>): Ditto.
12273 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
12274 * config/i386/constraints.md (Yi): Remove.
12279 2018-04-26 Nathan Sidwell <nathan@acm.org>
12281 * dumpfile.c (dump_open): New.
12282 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
12283 (dump_finish): Detect stdio/stderr by value not name.
12285 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
12287 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
12289 2018-04-26 Tom de Vries <tom@codesourcery.com>
12292 * config/nvptx/nvptx.c (verify_neutering_jumps)
12293 (verify_neutering_labels): New function
12294 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
12296 2018-04-26 Tom de Vries <tom@codesourcery.com>
12299 * config/nvptx/nvptx.c (needs_neutering_p): New function.
12300 (nvptx_single): Use needs_neutering_p to skip over insns that do not
12303 2018-04-26 Richard Biener <rguenther@suse.de>
12304 Tom de Vries <tom@codesourcery.com>
12307 * lto-streamer-out.c (output_function): Fixup loops if required to match
12308 discovery done in the reader.
12310 2018-04-26 Richard Biener <rguenther@suse.de>
12312 PR tree-optimization/85116
12313 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
12314 have a loop exit from the single latch predecessor. Remove
12315 case of header with just condition.
12316 (ch_base::copy_headers): Exclude infinite loops from any
12318 (pass_ch::execute): Record exits.
12320 2018-04-26 Richard Biener <rguenther@suse.de>
12322 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
12323 prologue cost vector and pass it to vect_get_load_cost.
12324 (vect_get_peeling_costs_all_drs): Likewise.
12325 (vect_peeling_hash_get_lowest_cost): Likewise.
12326 (vect_enhance_data_refs_alignment): Likewise.
12328 2018-04-26 Richard Biener <rguenther@suse.de>
12330 PR middle-end/85450
12331 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
12332 checking of integer<->pointer conversions.
12333 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
12334 sign-/zero-extending pointer types.
12335 (expand_omp_for_static_chunk): Likewise.
12337 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
12338 Jean Lee <xiaoyur347@gmail.com>
12340 * config/mips/mips.c (mips_asan_shadow_offset): New function.
12341 (TARGET_ASAN_SHADOW_OFFSET): Define.
12342 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
12343 true for -fsanitize=address.
12345 2018-04-25 Mark Wielaard <mark@klomp.org>
12347 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
12350 2018-04-25 Jakub Jelinek <jakub@redhat.com>
12352 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
12353 than "alu", remove explicit "memory" and "imm_disp" attributes.
12354 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
12356 PR middle-end/85414
12357 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
12358 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
12359 gen_lowpart_no_emit.
12361 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
12364 * config/i386/i386.c (ix86_expand_builtin): Change memory
12365 operand to XI, extend p0 to Pmode.
12366 * config/i386/i386.md: Change unspec volatile and operand
12367 1 mode to XI, change operand 0 mode to P.
12369 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12371 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
12372 GET_MODE_MASK before any checking.
12373 (nds32_can_use_bset_p): Likewise.
12374 (nds32_can_use_btgl_p): Likewise.
12376 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12378 * config/nds32/nds32-doubleword.md: New define_split pattern for
12379 illegal register number.
12381 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12383 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
12385 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12387 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
12389 2018-04-25 Richard Biener <rguenther@suse.de>
12391 * lto-streamer.h (LTO_major_version): Bump to 8.
12393 2018-04-25 Jakub Jelinek <jakub@redhat.com>
12395 * BASE-VER: Set to 9.0.0.
12397 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
12399 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
12400 in __abskf2 and __powikf2.
12402 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12405 * config/aarch64/constraints.md (Usg, Usj): New constraints.
12406 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
12407 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
12408 Use the above on operand 2. Reindent.
12409 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
12411 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
12414 * common/config/i386/i386-common.c (ix86_handle_option): Don't
12416 * config/i386/i386.opt (mcet): Removed.
12417 * doc/install.texi: Remove -mcet documentation.
12418 * doc/invoke.texi: Likewise.
12420 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
12423 * doc/install.texi: Remove -mcet from bootstrap-cet.
12425 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12428 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
12429 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
12433 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
12434 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
12435 containing a CONST_VECTOR.
12437 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
12439 * doc/install.texi: Update newlib dependency for nvptx.
12441 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12444 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
12445 instead of INTVAL when shifting x left.
12447 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
12449 PR tree-optimization/85478
12450 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
12451 vect_grouped_store_supported for single element vectors.
12453 2018-04-24 Richard Biener <rguenther@suse.de>
12456 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
12457 load cost increase to the case of non-constant step.
12459 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12462 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
12463 destination if any_malformed_asm.
12465 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
12467 PR middle-end/85496
12468 * expr.c (store_field): In the bitfield case, if the value comes from
12469 a function call and is returned in registers by means of a PARALLEL,
12470 do not change the mode of the temporary unless BLKmode and VOIDmode.
12472 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
12474 PR rtl-optimization/85423
12475 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
12476 dependencies to debug insns when the previous insn is non-debug.
12478 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
12480 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
12481 enums into a single definition.
12482 (fls): Fix predicates and printing.
12485 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
12487 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
12488 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
12489 and short u6 immediate.
12490 (check_if_valid_sleep_operand): Remove.
12491 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
12493 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12495 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
12496 flag_always_save_lp condition.
12497 * config/nds32/nds32.opt (malways-save-lp): New option.
12499 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12501 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
12502 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
12503 * config/nds32/nds32.h
12504 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
12505 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
12507 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12509 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
12510 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
12512 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12513 Chung-Ju Wu <jasonwucj@gmail.com>
12515 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
12517 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
12518 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
12520 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12522 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
12524 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12526 * config/nds32/nds32-protos.h (nds32_data_alignment,
12527 nds32_local_alignment): Declare.
12528 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
12529 nds32_local_alignment): New functions.
12530 (TARGET_CONSTANT_ALIGNMENT): Define.
12531 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
12533 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12535 * config/nds32/nds32.c
12536 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
12537 (TARGET_MODES_TIEABLE_P): Likewise.
12539 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12541 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
12542 level Ofast and Og.
12544 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
12545 Chung-Ju Wu <jasonwucj@gmail.com>
12547 * config/nds32/constants.md (unspec_volatile_element): Add enum values
12548 for unaligned access.
12549 * config/nds32/nds32-intrinsic.c: Implementation of expanding
12551 * config/nds32/nds32-intrinsic.md: Likewise.
12552 * config/nds32/nds32_intrinsic.h: Likewise.
12553 * config/nds32/nds32.h (nds32_builtins): Likewise.
12554 * config/nds32/nds32.opt (munaligned-access): New option.
12555 * config/nds32/nds32.c (nds32_asm_file_start): Display
12556 flag_unaligned_access status.
12558 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
12560 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
12561 -mno-relax is present.
12562 * config/riscv/linux.h (LINK_SPEC): Ditto.
12564 2018-04-20 Martin Sebor <msebor@redhat.com>
12567 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
12569 (gimple_fold_builtin_stxcpy_chk): Same.
12570 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
12572 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
12575 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
12576 __powikf2 when long double is IEEE 128-bit.
12578 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
12580 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
12581 step to make sure stack always aligned.
12583 2018-04-20 Carl Love <cel@us.ibm.com>
12586 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
12587 size check for arg0.
12589 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
12590 Tom de Vries <tom@codesourcery.com>
12593 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
12594 Emit insns for calls too.
12595 (nvptx_find_par): Always look for worker-level predecessor insn.
12596 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
12598 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
12599 (nvptx_process_pars): Propagate frames for calls.
12601 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
12604 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
12606 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
12607 (ix86_handle_option): Don't handle OPT_mibt.
12608 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
12610 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
12612 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
12613 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
12614 (ix86_target_macros): Define __CET__ with flag_cf_protection
12615 for -fcf-protection.
12616 * config/i386/i386.c (isa2_opts): Remove -mibt.
12617 * config/i386/i386.h (TARGET_IBT): Removed.
12618 (TARGET_IBT_P): Likewise.
12619 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
12620 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
12621 * config/i386/i386.opt (mcet): Update help message.
12622 (mshstk): Likewise.
12624 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
12625 -mcet as an alias for -mshstk.
12627 2018-04-20 Richard Biener <rguenther@suse.de>
12629 PR middle-end/85475
12630 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
12631 complexity by forcing a single use of the multiply operand.
12633 2018-04-20 Martin Jambor <mjambor@suse.cz>
12636 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
12637 recursion dependency to only apply to non-clones.
12639 2018-04-20 Martin Jambor <mjambor@suse.cz>
12642 * ipa-cp.c (create_specialized_node): Check that clones of
12643 self-recursive edges exist during IPA-CP.
12645 2018-04-19 Toon Moene <toon@moene.org>
12647 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
12650 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12652 PR tree-optimization/85467
12653 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
12654 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
12655 VECTOR_CST element to type.
12657 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12660 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
12661 * config/i386/i386.md (builtin_setjmp_setup): Removed.
12662 (builtin_longjmp): Likewise.
12663 (save_stack_nonlocal): New pattern.
12664 (restore_stack_nonlocal): Likewise.
12666 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12669 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12670 Replace ASM_OUTPUT_LABEL with fprintf.
12672 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12675 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12676 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
12677 * config/i386/i386-c.c (ix86_target_macros_internal): Also
12678 define __IBT__ and __SHSTK__ for -fcf-protection.
12679 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
12681 (ix86_trampoline_init): Likewise.
12682 (x86_output_mi_thunk): Likewise.
12683 (ix86_notrack_prefixed_insn_p): Likewise.
12684 (ix86_option_override_internal): Don't disallow -fcf-protection.
12685 * config/i386/i386.md (rdssp<mode>): Also enable for
12687 (incssp<mode>): Likewise.
12688 (nop_endbr): Likewise.
12689 * config/i386/i386.opt (mcet): Change help message to built-in
12692 (mshstk): Likewise.
12693 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
12694 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
12695 enable CET built-in functions.
12697 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
12699 * common/config/i386/i386-common.c
12700 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
12701 OPTION_MASK_ISA_MOVDIRI_UNSET,
12702 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
12703 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
12704 * config.gcc (movdirintrin.h): New header.
12705 * config/i386/cpuid.h (bit_MOVDIRI,
12706 bit_MOVDIR64B): New bits.
12707 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
12709 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
12710 (VOID, PVOID, PCVOID)): New function types.
12711 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
12712 __builtin_ia32_directstoreu_u64,
12713 __builtin_ia32_movdir64b): New builtins.
12714 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
12715 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
12717 (ix86_valid_target_attribute_inner_p): Ditto.
12718 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
12719 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
12720 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
12721 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
12722 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
12723 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
12724 (movdiri<mode>, movdir64b_<mode>): New.
12725 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
12726 * config/i386/immintrin.h: Include movdirintrin.h.
12727 * config/i386/movdirintrin.h: New file.
12728 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
12730 2018-04-19 Richard Biener <rguenther@suse.de>
12732 PR middle-end/85455
12733 * cfg.c (clear_bb_flags): When loop state says we have
12734 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
12736 2018-04-19 Richard Biener <rguenther@suse.de>
12738 PR tree-optimization/84737
12739 * tree-vect-data-refs.c (vect_copy_ref_info): New function
12740 copying restrict info.
12741 (vect_setup_realignment): Use it.
12742 * tree-vectorizer.h (vect_copy_ref_info): Declare.
12743 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
12744 the first DR to all generated stores.
12745 (vectorizable_load): Likewise for loads.
12747 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12749 PR tree-optimization/85446
12750 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
12751 the integral and pointer types to have the same precision.
12753 * doc/install.texi: Document --disable-cet being the default and
12756 2018-04-18 Martin Liska <mliska@suse.cz>
12758 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
12761 2018-04-18 Martin Liska <mliska@suse.cz>
12764 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12767 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12768 arguments if they are comparable.
12770 2018-04-18 Martin Liska <mliska@suse.cz>
12773 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12776 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12779 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
12782 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
12783 ENDBR after calling __morestack.
12785 2018-04-18 David Malcolm <dmalcolm@redhat.com>
12788 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
12789 by using gcc_base_ver to generate a gcc_driver_version, and use
12790 it when generating GCC_DRIVER_NAME.
12791 * configure: Regenerate.
12793 2018-04-18 Jakub Jelinek <jakub@redhat.com>
12796 * config.gcc: Obsolete powerpc*-*-*spe*.
12798 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12801 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
12802 (stabstr_D): Change type of unum from unsigned int to
12803 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
12806 2018-04-17 Jim Wilson <jimw@sifive.com>
12809 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
12810 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
12811 Set arg_pointer_offset after using pretend_args_size.
12813 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12815 PR rtl-optimization/85431
12816 * dse.c (record_store): Ignore zero width stores.
12819 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
12820 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
12821 __builtin_stack_restore rather than after it.
12822 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
12823 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
12824 argument instead of virtual_dynamic_stack_rtx.
12826 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
12828 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
12830 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
12831 Add note to error message to explain internal mapping of overloaded
12832 built-in function name to non-overloaded built-in function name.
12833 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
12836 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12839 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
12840 where the inputs overlap with the output.
12842 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12844 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
12845 (=v, v) alternative and explicit "memory" attribute.
12846 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
12847 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12849 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
12850 "sselog1" type instead of "sselog".
12851 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
12852 "sselog". Remove explicit "memory" attribute.
12853 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
12854 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12856 (vec_extract_hi_v32hi): Merge all alternatives into one, use
12857 "sselog1" type instead of "sselog". Remove explicit "memory"
12859 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
12860 use "sselog1" type instead of "sselog". Remove explicit "memory"
12862 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
12863 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12865 (vec_extract_hi_v64qi): Merge all alternatives into one, use
12866 "sselog1" type instead of "sselog". Remove explicit "memory"
12868 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
12869 use "sselog1" type instead of "sselog". Remove explicit "memory"
12873 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
12875 PR middle-end/85414
12876 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
12879 2018-04-17 Martin Jambor <mjambor@suse.cz>
12882 * ipa-cp.c (create_specialized_node): Call
12883 expand_all_artificial_thunks if necessary.
12885 2018-04-17 Martin Liska <mliska@suse.cz>
12888 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
12889 in message, remote space in between '_G' and '('.
12891 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12894 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
12895 avx512f_vmcmp<mode>3<round_saeonly_name>,
12896 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
12897 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
12898 avx512f_rndscale<mode><round_saeonly_name>,
12899 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
12900 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
12901 Use %<iptr>2 instead of %2 for -masm=intel.
12902 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
12903 avx512f_vcvttss2usi<round_saeonly_name>,
12904 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
12906 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
12907 avx512f_vcvttsd2usi<round_saeonly_name>,
12908 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
12909 Use %q1 instead of %1 for -masm=intel.
12910 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
12911 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
12912 of %3 for -masm=intel.
12913 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
12915 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
12917 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
12919 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
12920 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
12922 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
12924 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
12925 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
12926 %g1 and one with %0 and %1.
12927 (avx512er_vmrcp28<mode><round_saeonly_name>,
12928 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
12929 %1 for -masm=intel.
12930 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
12931 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
12932 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
12933 of %0 and %{%4%} for -masm=intel.
12934 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
12935 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
12936 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
12937 order of %0 and %{%5%}%{z%} for -masm=intel.
12939 2018-04-17 Jan Hubicka <jh@suse.cz>
12942 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
12944 2018-04-17 Martin Liska <mliska@suse.cz>
12947 * multiple_target.c (create_dispatcher_calls): Set apostrophes
12948 for target_clone error message. Make default implementation
12949 clone to be a local declaration.
12950 (separate_attrs): Add new argument and check for an empty
12952 (expand_target_clones): Handle it.
12953 (ipa_target_clone): Make redirection just for target_clones
12956 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
12957 Tom de Vries <tom@codesourcery.com>
12959 PR middle-end/84955
12960 * omp-expand.c (expand_oacc_for): Add dummy false branch for
12961 tiled basic blocks without omp continue statements.
12963 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
12966 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
12967 vec_extract expression as having side effects to make sure it gets
12970 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
12973 * config/i386/i386.c (get_builtin_code_for_version): Check
12976 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
12979 * config.gcc: Support "skylake".
12980 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12982 * config/i386/i386.c (m_SKYLAKE): Define.
12983 (processor_target_table): Add "skylake".
12984 (ix86_option_override_internal): Add "skylake".
12985 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
12986 PROCESSOR_CANNONLAKE.
12987 (get_builtin_code_for_version): Fix priority for
12988 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
12989 PROCESSOR_SKYLAKE-AVX512.
12990 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
12991 (processor_type): Add PROCESSOR_SKYLAKE.
12993 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
12994 Jason Merrill <jason@redhat.com>
12997 * convert.c (convert_to_integer_1): Use direct recursion for
12998 enumeral types and types with a precision less than the number
12999 of bits in their mode.
13001 2018-04-16 Julia Koval <julia.koval@intel.com>
13004 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
13005 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
13007 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
13010 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
13011 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
13012 and -mno-direct-move.
13014 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
13017 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
13018 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
13019 Ensure negative shifts result in {0}.
13021 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
13023 PR rtl-optimization/79916
13024 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
13025 regs (if any) to define how to gnerate SD moves when LRA is in
13028 2018-04-13 Jakub Jelinek <jakub@redhat.com>
13030 PR rtl-optimization/85393
13031 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
13032 * except.c (expand_dw2_landing_pad_for_region): Make static.
13033 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
13034 a label and unconditional jump to old_bb, rather than
13035 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
13038 PR rtl-optimization/85376
13039 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
13040 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
13041 instead of a specific value.
13043 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
13044 Bin Cheng <bin.cheng@arm.com>
13046 PR tree-optimization/82965
13047 PR tree-optimization/83991
13048 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
13049 by_profile_only parameter.
13050 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
13051 information if the loop was predicted to iterate too many times.
13052 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
13054 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
13057 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
13060 2018-04-13 Martin Liska <mliska@suse.cz>
13061 Jakub Jelinek <jakub@redhat.com>
13063 PR middle-end/81657
13064 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
13065 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
13066 * builtins.c (expand_builtin_memory_copy_args): Use
13067 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
13068 handle dest_addr == pc_rtx.
13070 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
13073 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
13074 asked to not generate direct moves.
13075 (fix_trunc<mode>si2_stfiwx): Similar.
13076 (fix_trunc<mode>si2_internal): Similar.
13078 2018-04-12 Jakub Jelinek <jakub@redhat.com>
13081 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
13082 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
13083 lookup if dest in some wider mode is known to be const0_rtx and
13084 if so, record permanent equivalence for it to be ZERO_EXTEND of
13085 the narrower mode destination.
13087 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
13089 * lto-streamer-out.c (output_function): Revert 259346.
13090 * omp-expand.c (expand_oacc_for): Likewise.
13092 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
13094 PR rtl-optimization/85354
13095 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
13096 * sel-sched.c (sel_global_init): ... here.
13098 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
13101 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
13102 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
13103 mode for PE-COFF targets.
13104 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
13105 (i386_pe_asm_lto_end): Likewise.
13106 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
13107 (TARGET_ASM_LTO_END): Likewise.
13108 * config/i386/winnt.c (saved_debug_info_level): New static variable.
13109 (i386_pe_asm_lto_start): New function.
13110 (i386_pe_asm_lto_end): Likewise.
13112 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
13113 Richard Biener <rguenther@suse.de>
13115 PR middle-end/84955
13116 * lto-streamer-out.c (output_function): Fix CFG loop state before
13118 * omp-expand.c (expand_oacc_for): Handle calls to internal
13119 functions like regular functions.
13121 2018-04-12 Richard Biener <rguenther@suse.de>
13124 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
13125 for the early LTO debug to properly generate references to it
13126 during DIE emission. Do not re-use that for the skeleton for
13128 (dwarf2out_early_finish): Likewise.
13130 2018-04-12 Jakub Jelinek <jakub@redhat.com>
13133 * config/i386/sse.md
13134 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
13135 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
13136 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
13137 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
13138 and output is a reg, avoid creating invalid lowpart subreg, but
13139 instead split into a 512-bit move. Don't split if not AVX512VL,
13140 input is xmm16+ reg and output is a mem.
13141 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
13142 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
13143 xmm16+ reg and output is a mem.
13145 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13147 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
13148 also for flag_dwarf2_cfi_asm.
13150 2018-04-12 Jakub Jelinek <jakub@redhat.com>
13152 PR rtl-optimization/85342
13153 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
13154 a bool scalar var inside of the loop instead. Don't try to update
13155 recog_data.operand after failed apply_change_group.
13157 2018-04-12 Tom de Vries <tom@codesourcery.com>
13160 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
13161 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
13162 array with flexible array member as array without given dimension.
13163 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
13164 argument for undefined param to true.
13166 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
13169 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
13170 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
13171 from PowerPC section.
13172 * config/rs6000/sysv4.opt (mcall-): Improve help text.
13173 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
13174 help text that is too long.
13175 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
13176 help text that is too long.
13177 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
13178 help text that is too long.
13180 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
13182 * config/alpha/alpha.md (stack_probe_internal): Rename
13183 from "probe_stack". Update all callers.
13185 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13187 PR rtl-optimization/84566
13188 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
13189 sched_macro_fuse_insns.
13191 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13194 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
13195 (compute_block_dependences): ... from here.
13197 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13199 PR tree-optimization/85331
13200 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
13201 from int to HOST_WIDE_INT.
13203 2018-04-11 Martin Jambor <mjambor@suse.cz>
13206 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
13207 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
13208 not the same as the source val.
13209 (cgraph_edge_brings_value_p): New parameter.
13210 (gather_edges_for_value): Pass destination value to
13211 cgraph_edge_brings_value_p.
13212 (perhaps_add_new_callers): Likewise.
13213 (get_info_about_necessary_edges): Likewise and exclude values brought
13214 only by self-recursive edges.
13215 (create_specialized_node): Redirect only clones of self-calling edges.
13216 (+self_recursive_pass_through_p): New function.
13217 (find_more_scalar_values_for_callers_subset): Use it.
13218 (find_aggregate_values_for_callers_subset): Likewise.
13219 (known_aggs_to_agg_replacement_list): Removed.
13220 (decide_whether_version_node): Re-calculate known constants for all
13221 remaining context clones.
13223 2018-04-11 Richard Biener <rguenther@suse.de>
13226 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
13227 from early DWARF output.
13228 (dwarf2out_early_finish): Output line info unconditionally into
13229 early DWARF and add reference to it.
13231 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13234 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
13235 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
13236 other than V2DFmode using iptr mode attribute.
13237 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
13239 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13241 PR rtl-optimization/84659
13242 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
13244 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13247 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
13249 (output_loc_list): Pass address of a dummy size variable even in the
13250 locview handling loop.
13251 (index_location_lists): Add comment on why skip_loc_list_entry can't
13254 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
13257 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
13260 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
13263 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
13264 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
13265 and -mstring-compare-inline-limit.
13267 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
13270 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
13271 for stack clash protection in a register whenever we need it to be in
13274 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
13276 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
13277 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
13279 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
13282 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
13284 (mlong-double-): Ditto.
13285 * config/rs6000/sysv4.opt (msdata=): Ditto.
13286 (mtls-size=): Ditto.
13288 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13290 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
13291 erroneous entries for
13292 "vector int vec_ldl (int, long int *)", and
13293 "vector unsigned int vec_ldl (int, unsigned long int *)".
13294 Add comments and entries for
13295 "vector bool char vec_ldl (int, bool char *)",
13296 "vector bool short vec_ldl (int, bool short *)",
13297 "vector bool int vec_ldl (int, bool int *)",
13298 "vector bool long long vec_ldl (int, bool long long *)",
13299 "vector pixel vec_ldl (int, pixel *)",
13300 "vector long long vec_ldl (int, long long *)",
13301 "vector unsigned long long vec_ldl (int, unsigned long long *)".
13302 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
13303 type tree bool_long_long_type_node and correct definition of
13304 bool_V2DI_type_node to make reference to this new type tree.
13305 (rs6000_mangle_type): Replace erroneous reference to
13306 bool_long_type_node with bool_long_long_type_node.
13307 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
13308 comments to emphasize sign distinctions for char and int types and
13309 replace RS6000_BTI_bool_long constant with
13310 RS6000_BTI_bool_long_long constant. Also add comment to restrict
13311 use of RS6000_BTI_pixel.
13312 (bool_long_type_node): Remove this macro definition.
13313 (bool_long_long_type_node): New macro definition
13315 2018-04-10 Jakub Jelinek <jakub@redhat.com>
13317 PR rtl-optimization/85300
13318 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
13319 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
13320 simplify_unary_operation fails.
13322 2018-04-10 Martin Liska <mliska@suse.cz>
13324 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
13325 cgraph_edge and ipa_ref.
13327 2018-04-10 Jakub Jelinek <jakub@redhat.com>
13331 * config/i386/sse.md
13332 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
13333 computation of the VEC_MERGE selector from mask.
13334 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
13335 Fix decoding of the VEC_MERGE selector into mask.
13337 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
13339 PR tree-optimization/85286
13340 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
13342 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
13344 * final.c (final_1): Set insn_last_address as well as
13345 insn_current_address.
13347 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13350 * explow.c (emit_stack_probe): Call validize_mem on memory location
13351 before passing it to gen_probe_stack. Create address operand and
13352 legitimize it for the probe_stack_address case.
13354 2018-04-09 Jan Hubicka <jh@suse.cz>
13357 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
13358 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
13359 * tree.c (free_lang_data_in_type): Fix handling of binfos;
13361 (free_lang_data): Rebuild type inheritance graph.
13363 2018-04-09 Martin Sebor <msebor@redhat.com>
13365 * invoke.texi (-finline-small-functions): Mention other optimization
13367 (-findirect-inlining, -fpartial-inlining): Same.
13368 (-finline-functions-called-once): Same.
13369 (-freorder-blocks-and-partition): Same.
13371 2018-04-09 Jan Hubicka <jh@suse.cz>
13374 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
13375 jumps; choose last target that matches the criteria (i.e.
13376 no partition changes for non-crossing jumps).
13377 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
13378 support for redirecting crossing jumps to non-crossing.
13380 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
13382 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
13383 also for naked functions.
13385 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
13387 * config/arc/arc.md (add_shift): New pattern.
13388 (add_shift2): Likewise.
13389 (sub_shift): Likewise.
13390 (sub_shift_cmp0_noout): Likewise.
13391 (compare_si_ashiftsi): Likewise.
13392 (xbfu_cmp0_noout): New combine pattern.
13393 (xbfu_cmp0"): Likewise.
13394 (movsi_set_cc_insn): Place the predicable variant first.
13395 (commutative_binary_cmp0_noout): Remove clobber.
13396 (commutative_binary_cmp0): New pattern.
13397 (noncommutative_binary_cmp0): Likewise.
13398 (noncommutative_binary_cmp0_noout): Likewise.
13399 (noncommutative_binary_comparison_result_used): Removed.
13400 (rsub_cmp0): New pattern.
13401 (rsub_cmp0_noout): Likewise.
13402 (extzvsi): Changed, keep only meaningful variants.
13403 (SQH, SEZ): New iterators.
13404 (SQH_postfix): New mode attribute.
13405 (SEZ_prefix): New code attribute.
13406 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
13407 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
13408 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
13409 of numerical value.
13410 (noncommutative_operator): Check the availability of barrel
13413 2018-04-09 Richard Biener <rguenther@suse.de>
13415 PR tree-optimization/85284
13416 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
13417 Only use the niter constraining form of simple_iv when the exit
13418 is always executed.
13420 2018-04-09 Tom de Vries <tom@codesourcery.com>
13423 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13424 (define_expand "*memory_barrier"): New define_expand.
13425 (define_insn "memory_barrier"): New insn.
13427 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13429 PR rtl-optimization/80463
13430 PR rtl-optimization/83972
13431 PR rtl-optimization/83480
13433 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
13434 correct producer for the insn.
13435 (tidy_control_flow): Fixup seqnos in case of debug insns.
13437 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13439 PR rtl-optimization/83913
13441 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
13442 different sched-times when merging exprs.
13444 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13446 PR rtl-optimization/83962
13448 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
13449 tidy_fallthru_edge and tidy_control_flow.
13451 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13453 PR rtl-optimization/83530
13455 * sel-sched.c (force_next_insn): New global variable.
13456 (remove_insn_for_debug): When force_next_insn is true, also leave only
13457 next insn in the ready list.
13458 (sel_sched_region): When the region wasn't scheduled, make another pass
13459 over it with force_next_insn set to 1.
13461 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
13463 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
13465 * config/nds32/constants.md (unspec_volatile_element): Add enum values
13466 for interrupt control.
13467 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13468 functions for interrupt control.
13469 * config/nds32/nds32-intrinsic.md: Likewise.
13470 * config/nds32/nds32_intrinsic.h: Likewise.
13471 * config/nds32/nds32.h (nds32_builtins): Likewise.
13473 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
13475 * config/nds32/nds32.c (nds32_init_machine_status,
13476 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
13477 strict_aligned_p field.
13478 (nds32_expand_to_rtl_hook): New function.
13479 (TARGET_EXPAND_TO_RTL_HOOK): Define.
13480 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
13482 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13483 Chung-Ju Wu <jasonwucj@gmail.com>
13485 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
13486 * config/nds32/nds32-n7.md: New file.
13487 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
13488 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
13490 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
13491 * config/nds32/nds32.md (pipeline_model): Add n7.
13492 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
13493 * config/nds32/pipelines.md: Include n7 settings.
13495 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13496 Chung-Ju Wu <jasonwucj@gmail.com>
13498 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
13499 * config/nds32/nds32-e8.md: New file.
13500 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
13501 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
13503 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
13504 * config/nds32/nds32.md (pipeline_model): Add e8.
13505 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
13506 * config/nds32/pipelines.md: Include e8 settings.
13508 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13509 Chung-Ju Wu <jasonwucj@gmail.com>
13511 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
13512 * config/nds32/nds32-n8.md: New file.
13513 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
13514 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
13516 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
13517 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
13518 * config/nds32/nds32.md (pipeline_model): Add n8.
13519 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
13520 * config/nds32/pipelines.md: Include n8 settings.
13522 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13523 Chung-Ju Wu <jasonwucj@gmail.com>
13525 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
13526 * config/nds32/nds32-n9-2r1w.md: New file.
13527 * config/nds32/nds32-n9-3r2w.md: New file.
13528 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
13529 nds32_register_ports): New or modify for cpu n9.
13530 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
13532 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
13533 * config/nds32/nds32-utils.c: New file.
13534 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
13535 TARGET_MUL_SLOW): Define.
13536 * config/nds32/nds32.md (pipeline_model): New attribute.
13537 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
13538 New options that support cpu n9.
13539 * config/nds32/pipelines.md: Include n9 settings.
13540 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
13542 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
13544 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
13545 information if necessary.
13546 (output_cond_branch_compare_zero): Likewise.
13547 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
13548 (nds32_target_alignment): Refine for alignment.
13549 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
13550 (FUNCTION_BOUNDARY): Modify.
13551 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
13553 * config/nds32/nds32.opt (malways-align, malign-functions): New.
13555 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
13557 * config/nds32/constants.md (unspec_volatile_element): Add values for
13558 TLB operation and data prefetch.
13559 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13560 functions for TLB operation and data prefetch.
13561 * config/nds32/nds32-intrinsic.md: Likewise.
13562 * config/nds32/nds32_intrinsic.h: Likewise.
13563 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
13564 (nds32_print_operand): Likewise.
13565 * config/nds32/nds32.h (nds32_builtins): Likewise.
13567 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
13568 Andrew Pinski <pinsika@gcc.gnu.org>
13570 PR middle-end/82976
13571 * match.pd: Use constant_boolean_node of correct type instead of
13572 boolean_true_node or boolean_false_node for simplifying
13573 pointer comparisons to zero.
13575 2018-04-07 Jakub Jelinek <jakub@redhat.com>
13577 PR tree-optimization/80021
13578 * tree.c (verify_type_variant): Make error call in verify_variant_match
13579 translatable and remove final full stop.
13581 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13583 * config/nds32/constants.md (unspec_volatile_element): Add
13584 UNSPEC_VOLATILE_EH_RETURN.
13585 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
13586 nds32_output_stack_pop): Support dwarf exception handling process.
13587 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
13588 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
13589 exception handling process.
13590 (nds32_compute_stack_frame): Likewise.
13591 (nds32_return_addr_rtx): Likewise.
13592 (nds32_initial_elimination_offset): Likewise.
13593 (nds32_expand_prologue): Likewise.
13594 (nds32_expand_epilogue): Likewise.
13595 (nds32_dynamic_chain_address): New function.
13596 * config/nds32/nds32.h (machine_function): Add fields for dwarf
13597 exception handling.
13598 (DYNAMIC_CHAIN_ADDRESS): Define.
13599 (EH_RETURN_DATA_REGNO): Define.
13600 (EH_RETURN_STACKADJ_RTX): Define.
13601 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
13602 patterns for dwarf exception handling.
13604 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13606 * config/nds32/nds32.h: Clean up obsolete macros.
13608 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13610 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13611 Add enum values for particular instructions.
13612 * config/nds32/nds32-intrinsic.c: Implementation of expanding
13613 particular intrinsic functions.
13614 * config/nds32/nds32-intrinsic.md: Likewise.
13615 * config/nds32/nds32_intrinsic.h: Likewise.
13616 * config/nds32/nds32.h (nds32_builtins): Likewise.
13617 * config/nds32/nds32.md (type): Add pbsad and pbsada.
13618 (btst, ave): New patterns for particular instructions.
13620 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13622 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13623 Add enum values for atomic load/store and memory sync.
13624 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
13626 * config/nds32/nds32-intrinsic.md: Likewise.
13627 * config/nds32/nds32_intrinsic.h: Likewise.
13628 * config/nds32/nds32.h (nds32_builtins): Likewise.
13630 2018-04-07 Jakub Jelinek <jakub@redhat.com>
13632 PR tree-optimization/85257
13633 * fold-const.c (native_encode_vector): If not all elts could fit
13634 and off is -1, return 0 rather than offset.
13635 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
13636 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
13637 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
13638 adjust buffer in native_interpret_expr call.
13640 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13642 * config/nds32/constants.md (unspec_volatile_element): Add cache
13643 control enum values.
13644 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
13645 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
13646 * config/nds32/nds32.c (nds32_cctl_names): New.
13647 (nds32_print_operand): Handle cache control register names.
13648 * config/nds32/nds32.h (nds32_builtins): New enum values.
13649 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
13651 * config/nds32/nds32.md (type): Add mmu.
13652 * config/nds32/pipelines.md (simple_insn): Add mmu.
13654 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13656 * config/nds32/nds32.md (type): Remove call.
13657 * config/nds32/pipelines.md (simple_insn): Likewise.
13659 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13661 * config/nds32/constants.md (unspec_volatile_element): Add
13662 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
13663 UNSPEC_VOLATILE_FMFCFG.
13664 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
13665 description for fmfcfg and fmfcsr.
13666 (bdesc_1arg): Add fmtcsr.
13667 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
13668 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
13669 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
13670 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
13671 unspec_fmfcfg): New patterns.
13672 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
13673 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
13674 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
13675 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
13676 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
13677 __nds32__fmfcfg): Define.
13679 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13681 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
13682 intrinsic register names.
13683 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
13684 intrinsic register enum values and macros.
13686 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13688 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
13689 for load/store addressing form.
13690 (nds32_print_operand_address): Likewise.
13692 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
13695 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
13696 based on LABEL_REF. Remove useless assertion.
13697 (pic_address_needs_scratch): Fix formatting.
13698 (sparc_legitimize_pic_address): Minor tweaks.
13699 (sparc_delegitimize_address): Adjust assertion accordingly.
13700 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
13701 into symbolic_operand.
13702 (movsi_high_pic_label_ref): Likewise.
13703 (movsi_lo_sum_pic_label_ref): Likewise.
13704 (movdi_pic_label_ref): Likewise.
13705 (movdi_high_pic_label_ref): Likewise.
13706 (movdi_lo_sum_pic_label_ref): Likewise.
13708 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
13710 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
13711 custom LIB_SPEC setup.
13713 2018-04-06 Ruslan Bukin <br@bsdpad.com>
13714 Kito Cheng <kito.cheng@gmail.com>
13716 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
13717 * config/riscv/freebsd.h: New.
13719 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13721 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
13722 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
13725 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13726 Kito Cheng <kito.cheng@gmail.com>
13728 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
13729 nds32_output_call, nds32_symbol_binds_local_p): New functions.
13730 * config/nds32/nds32-protos.h (nds32_output_call,
13731 nds32_output_return): Declare.
13732 * config/nds32/nds32.md: Refine all the call and return patterns.
13734 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13737 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
13738 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
13740 PR rtl-optimization/84872
13741 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
13742 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
13743 EDGE_CROSSING edge.
13745 2018-04-06 Tamar Christina <tamar.christina@arm.com>
13747 * expr.c (copy_blkmode_to_reg): Revert 254862.
13748 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
13750 2018-04-06 Richard Biener <rguenther@suse.de>
13752 PR middle-end/85244
13753 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
13754 after seeing a component reference with an adjacent field. Treat
13755 refs to arrays at struct end of external decls similar to
13756 refs to unconstrained commons.
13758 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13761 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
13762 look through SAVE_EXPRs with non-side-effects argument. Adjust
13764 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
13767 2018-04-06 Richard Biener <rguenther@suse.de>
13769 PR middle-end/85180
13770 * alias.c (find_base_term): New wrapper around find_base_term
13771 unwinding CSELIB_VAL_PTR changes.
13772 (find_base_term): Do not restore CSELIB_VAL_PTR during the
13775 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13777 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
13779 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
13780 constant definitions.
13781 ("nop"): lr 0,0 -> nopr r0
13782 ("nop_lr0", "nop_lr1"): New insn definitions.
13784 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13786 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
13787 NDS32_V3PUSH_AVAILABLE_P macro.
13789 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
13790 Chung-Ju Wu <jasonwucj@gmail.com>
13792 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
13793 (nds32*-*-*): Add float and fpu_config into supported_defaults.
13794 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
13795 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
13796 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
13797 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
13798 * config/nds32/constraints.md: New constraints and checking for hard
13799 float configuration.
13800 * config/nds32/iterators.md: New mode iterator and attribute for hard
13801 float configuration.
13802 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
13804 * config/nds32/nds32-fpu.md: New file.
13805 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
13806 deal with hard float code generation.
13807 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
13809 (abi_type, float_reg_number): New enum type.
13810 * config/nds32/nds32-predicates.c: New predicates for hard float.
13811 * config/nds32/nds32-protos.h: Declare functions for hard float.
13812 * config/nds32/nds32.c: Implementation for hard float configuration.
13813 * config/nds32/nds32.h: Definitions for hard float configuration.
13814 * config/nds32/nds32.md: Include hard float machine description and
13815 modify patterns for hard float configuration.
13816 * config/nds32/nds32.opt: New options for hard float configuration.
13817 * config/nds32/predicates.md: New predicates for hard float
13820 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13822 * common/config/nds32/nds32-common.c
13823 (nds32_option_optimization_table): Enable -mreleax-hint by default.
13825 2018-04-05 Jakub Jelinek <jakub@redhat.com>
13827 PR middle-end/85195
13828 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
13829 CONSTRUCTOR_ELT (ctor, ...)->value.
13831 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
13834 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
13836 2018-04-05 Tom de Vries <tom@codesourcery.com>
13839 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
13842 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13843 Kito Cheng <kito.cheng@gmail.com>
13845 * config/nds32/constraints.md (U33): Fine-tune checking condition.
13846 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
13847 * config/nds32/nds32.h (nds32_16bit_address_type): Add
13848 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
13850 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13851 Kito Cheng <kito.cheng@gmail.com>
13853 * config/nds32/constraints.md (Ufe): New memory constraint.
13854 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
13855 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
13856 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
13858 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
13859 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
13861 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13863 * config/nds32/nds32.md: Use optimize_size in the condition for
13864 alu-shift instructions.
13866 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13868 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
13870 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13872 * config/nds32/nds32.md (negsi2): Refine pattern.
13874 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13875 Chung-Ju Wu <jasonwucj@gmail.com>
13877 * config/nds32/iterators.md (shift_rotate): New code iterator.
13878 (shift): New code attribute.
13879 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
13880 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
13881 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
13882 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
13883 bit-wise operations.
13884 (andsi3, *andsi3): Ditto.
13885 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
13886 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
13887 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
13888 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
13889 nds32_ior_operand, nds32_xor_operand): New predicates.
13891 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13893 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
13894 (addsi3, subsi3): ... this.
13896 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13898 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
13900 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13902 * config/nds32/nds32.md: Adjust indention.
13904 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13906 * config/nds32/nds32.md (feature): New attribute.
13908 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13910 * config/nds32/nds32.md (subtype): New attribute.
13912 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
13915 * config/arm/arm-builtins.c (arm_expand_builtin): Change
13916 expansion to perform a bitwise AND of the argument followed by a
13917 boolean negation of the result.
13919 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
13921 PR rtl-optimization/84878
13922 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
13923 the basic block. Assert the use reference is not artificial and that
13924 it has an associated insn.
13926 2018-04-04 Michael Matz <matz@suse.de>
13928 * builtins.c (compute_objsize): Pass correct operand
13929 to array_at_struct_end_p.
13931 2018-04-04 Richard Biener <rguenther@suse.de>
13934 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
13935 from contexts for DINFO_LEVEL_TERSE and below.
13937 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13939 * config/nds32/nds32-doubleword.md (move_<mode>): Require
13940 resiter_operand condition.
13941 * config/nds32/nds32.md (*move<mode>): Ditto.
13943 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13944 Monk Chiang <sh.chiang04@gmail.com>
13946 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
13948 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13950 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
13952 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13953 Kito Cheng <kito.cheng@gmail.com>
13955 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
13956 nds32_cond_code_str, output_cond_branch,
13957 output_cond_branch_compare_zero, nds32_expand_cbranch,
13958 nds32_expand_cstore, nds32_expand_movcc,
13959 nds32_output_cbranchsi4_equality_zero,
13960 nds32_output_cbranchsi4_equality_reg,
13961 nds32_output_cbranchsi4_equality_reg_or_const_int,
13962 nds32_output_cbranchsi4_greater_less_zero: New functions.
13963 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
13964 nds32_expand_cstore, nds32_expand_movcc,
13965 nds32_output_cbranchsi4_equality_zero,
13966 nds32_output_cbranchsi4_equality_reg,
13967 nds32_output_cbranchsi4_equality_reg_or_const_int,
13968 nds32_output_cbranchsi4_greater_less_zero): Declare.
13969 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
13970 nds32_rimm11s_operand): New predicates.
13971 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
13972 * config/nds32/nds32.md: Rewrite all the branch and conditional move
13975 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13977 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
13978 * config/nds32/nds32.md: Ditto.
13979 * config/nds32/pipelines.md: Ditto.
13981 2018-04-04 Richard Biener <rguenther@suse.de>
13983 PR tree-optimization/85168
13984 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
13985 propagating abnormals.
13987 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13989 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
13991 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13992 Kito Cheng <kito.cheng@gmail.com>
13994 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
13995 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
13996 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
13997 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
13998 * config/nds32/nds32.md (sibcall_internal): New.
13999 (sibcall_register): Remove.
14000 (sibcall_immediate): Remove.
14001 (sibcall_value_internal): New.
14002 (sibcall_value_register): Remove.
14003 (sibcall_value_immediate): Remove.
14004 * config/nds32/predicates.md (nds32_general_register_operand): New.
14005 (nds32_call_address_operand): New.
14007 2018-04-03 Jakub Jelinek <jakub@redhat.com>
14009 PR rtl-optimization/85167
14010 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
14011 bb_defs if *split_p, instead preinitialize it to NULL.
14013 PR tree-optimization/85156
14014 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
14015 evaluating the argument multiple times.
14017 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
14019 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
14021 (_mm_cvtpd_ps): Likewise.
14022 (_mm_cvttpd_epi32): Likewise.
14023 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
14024 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
14025 vector, pixel, and bool following altivec.h include.
14027 2018-04-03 Martin Sebor <msebor@redhat.com>
14029 * doc/extend.texi (Common Function Attributes): Clarify.
14030 (const attribute): Likewise.
14031 (pure attribute): Likewise.
14033 2018-04-03 Jakub Jelinek <jakub@redhat.com>
14036 * config/i386/i386.c (ix86_expand_vector_set): Use
14037 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
14039 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
14041 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
14042 instructions when changing rounding bits to preserve precision bits
14043 in the x87 control word.
14045 2018-04-03 Martin Liska <mliska@suse.cz>
14047 PR tree-optimization/82491
14048 * rtl.h (strip_offset_and_add): Replace += suboffset with
14049 poly_uint64 () + suboffset.
14051 2018-03-29 Martin Liska <mliska@suse.cz>
14052 Martin Jambor <mjambor@suse.cz>
14055 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
14056 param_type is not an integral or pointer type.
14058 2018-04-03 Richard Biener <rguenther@suse.de>
14060 * sese.h (recompute_all_dominators): Remove.
14062 2018-04-02 Martin Sebor <msebor@redhat.com>
14064 * doc/invoke.texi (-Wrestrict): Fix typos.
14066 2018-04-02 Jim Wilson <jimw@sifive.com>
14068 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
14069 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
14070 (<optab>di3, <optab>si3_extend): Likewise.
14071 (<optab>si3_mask, <optab>si3_mask_1): New.
14072 (<optab>di3_mask, <optab>di3_mask_1): New.
14073 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
14074 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
14075 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
14077 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
14079 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
14082 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
14084 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
14085 (nds32_canonicalize_comparison): New function.
14087 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
14088 Kito Cheng <kito.cheng@gmail.com>
14089 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14091 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
14092 * config/nds32/constants.md (unspec_volatile_element): Add
14093 UNSPEC_VOLATILE_RELAX_GROUP.
14094 * config/nds32/nds32-relax-opt.c: New file.
14095 * config/nds32/nds32-predicates.c
14096 (nds32_symbol_load_store_p): New function.
14097 * config/nds32/nds32-protos.h
14098 (nds32_symbol_load_store_p): Declare function.
14099 (make_pass_nds32_relax_opt): Declare new rtl pass function.
14100 * config/nds32/nds32.c
14101 (nds32_register_pass): New function to register pass.
14102 (nds32_register_passes): New function to register passes.
14103 * config/nds32/nds32.md (relax_group): New pattern.
14104 * config/nds32/nds32.opt (mrelax-hint): New option.
14105 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
14107 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
14109 * config/nds32/t-nds32: Modify files dependency.
14111 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
14113 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
14114 (PROFILE_HOOK): Define its implementation.
14116 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
14118 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
14119 type and 32-bit size.
14121 2018-04-01 Jakub Jelinek <jakub@redhat.com>
14123 PR middle-end/85090
14124 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
14125 (V_128_256): New mode iterator.
14126 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
14127 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
14128 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
14130 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
14131 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
14133 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
14136 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
14137 NaN inputs correctly.
14139 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
14142 * config/rs6000/vsx.md (??r): New mode attribute.
14143 (*vsx_mov<mode>_64bit): Use it.
14144 (*vsx_mov<mode>_32bit): Likewise.
14146 2018-03-30 Martin Sebor <msebor@redhat.com>
14148 PR tree-optimization/84818
14149 * builtins.c (check_access): Use warning_n.
14151 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14154 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
14156 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
14159 2018-03-30 Julia Koval <julia.koval@intel.com>
14162 * x86-tune.def (movx, partial_reg_dependency): Enable for
14165 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
14167 PR inline-asm/84985
14168 * lra-constraints.c (process_alt_operands): Move setting
14169 this_alternative_matches below.
14171 2018-03-29 Martin Liska <mliska@suse.cz>
14174 * doc/invoke.texi: Document how LTO works with debug info.
14175 Describe auto-load support of binutils. Mention 'x86-64'
14176 as valid option value of -march option.
14178 2018-03-29 Jakub Jelinek <jakub@redhat.com>
14180 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
14183 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
14184 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
14185 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
14188 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
14191 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
14192 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
14193 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
14194 for RS6000_BTM_POWERPC64.
14195 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
14196 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
14197 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
14200 (DIVDEU): Likewise.
14202 2018-03-28 Carl Love <cel@us.ibm.com>
14205 2017-09-27 Carl Love <cel@us.ibm.com>
14207 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14208 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14209 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14212 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14214 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
14215 instead of __vector bool.
14216 (_mm_max_pu8): Likewise.
14217 (_mm_min_pi16): Likewise.
14219 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
14222 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
14223 (DIVWEUO): Likewise.
14224 (DIVDEO): Likewise.
14225 (DIVDEUO): Likewise.
14226 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
14227 DIVWEUO and DIVDEUO.
14228 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
14229 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
14230 (div_extend): Likewise.
14231 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
14233 (__builtin_divweuo): Likewise.
14234 (__builtin_divdeo): Likewise.
14235 (__builtin_divdeuo): Likewise.
14237 2018-03-28 Jakub Jelinek <jakub@redhat.com>
14240 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
14241 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
14243 PR tree-optimization/82004
14244 * gimple-match-head.c (optimize_pow_to_exp): New function.
14245 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
14246 Don't fold to exp if optimize_pow_to_exp is false.
14248 2018-03-28 Martin Liska <mliska@suse.cz>
14251 * calls.c (initialize_argument_information): Fix trailing space.
14252 * common.opt: Fix typo and provide better explanation for
14253 -fsanitize-coverage option.
14254 * config/i386/i386.opt: Fix typo.
14256 2018-03-28 Jakub Jelinek <jakub@redhat.com>
14257 Martin Liska <mliska@suse.cz>
14260 * gimplify.c (asan_poison_variable): Don't do the check for
14261 gimplify_omp_ctxp here.
14262 (gimplify_decl_expr): Do it here.
14263 (gimplify_target_expr): Likewise.
14265 2018-03-28 Martin Liska <mliska@suse.cz>
14268 * config/i386/i386.c (ix86_function_arg_advance): Do not call
14269 chkp_type_bounds_count if MPX is not enabled.
14271 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
14273 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
14275 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
14278 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
14279 function to create the function decl for complex long double
14280 multiply and divide for -mabi=ieeelongdouble.
14281 (init_float128_ieee): Call it.
14283 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
14286 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
14287 -fcf-protection=branch -mibt.
14288 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
14290 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14293 * config/arm/arm.c (arm_valid_symbolic_address): Handle
14294 arm_word_relocations.
14296 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
14299 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
14300 extern array declarations.
14302 2018-03-27 Richard Biener <rguenther@suse.de>
14304 PR middle-end/84067
14305 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
14306 explicit single_use checks.
14308 2018-03-27 Richard Biener <rguenther@suse.de>
14310 PR tree-optimization/85082
14311 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
14314 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14316 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
14317 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
14318 Turn on fasynchronous-unwind-tables and funwind-tables.
14320 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
14323 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
14324 (*bmi_blsr_<mode>_ccz): Ditto.
14326 2018-03-26 Tom de Vries <tom@codesourcery.com>
14328 PR tree-optimization/85063
14329 * omp-general.c (offloading_function_p): New function. Factor out
14331 * omp-offload.c (pass_omp_target_link::gate): ... here.
14332 * omp-general.h (offloading_function_p): Declare.
14333 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
14334 with attribute omp declare target for offloading functions.
14336 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
14338 PR tree-optimization/84005
14339 * tree-data-ref.h (get_base_for_alignment): Declare.
14340 * tree-data-ref.c (get_base_for_alignment_1): New function.
14341 (get_base_for_alignment): Likewise.
14342 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
14343 get_base_for_alignment to find a suitable base object, instead
14344 of always using drb->base_address.
14346 2018-03-23 Jakub Jelinek <jakub@redhat.com>
14348 PR inline-asm/85022
14349 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
14350 known size by default.
14352 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
14354 PR inline-asm/85030
14355 * lra-constraints.c (process_alt_operands): Don't match BLKmode
14356 and non BLKmode operands.
14358 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14361 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
14362 Clean up attributes.
14364 2018-03-23 Richard Biener <rguenther@suse.de>
14367 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
14368 we are going to emit early debug for LTO.
14370 2018-03-23 Jakub Jelinek <jakub@redhat.com>
14372 PR inline-asm/85034
14373 * function.c (match_asm_constraints_1): Don't optimize if input
14374 doesn't satisfy general_operand predicate for output's mode.
14376 PR inline-asm/85022
14377 * alias.c (write_dependence_p): Don't require for x_canonicalized
14378 non-VOIDmode if x has VOIDmode.
14381 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
14382 just don't try to optimize it rather than assert it never happens.
14384 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
14386 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
14387 macro expansions for definition of ST_INTERNAL_<mode> and
14388 LD_INTERNAL_<mode> builtins.
14389 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
14391 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
14393 (altivec_expand_st_builtin): Likewise.
14394 (altivec_expand_builtin): Remove calls to deleted functions.
14395 (rs6000_address_for_altivec): Delete this function.
14396 * config/rs6000/vector.md: Remove expands for
14397 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
14399 2018-03-22 Sudakshina Das <sudi.das@arm.com>
14402 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
14403 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
14404 re-computing once computed.
14405 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
14406 (arm_init_machine_status): Initialize
14407 machine->static_chain_stack_bytes.
14409 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
14412 * doc/extend.texi: Add four new prototypes for vec_ld.
14413 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
14414 definitions for more logical presentation.
14415 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
14416 entries for V1TI variants of __builtin_altivec_ld builtin.
14417 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
14418 handling of V1TI variant of LVX icode pattern.
14419 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
14420 (rs6000_gimple_fold_builtin): Likewise.
14421 (altivec_init_builtins): Add code to define
14422 __builtin_altivec_lvx_v1ti function.
14424 2018-03-22 Jakub Jelinek <jakub@redhat.com>
14426 PR inline-asm/84941
14427 * function.c (match_asm_constraints_1): Don't do the optimization
14428 if input isn't a REG, SUBREG, MEM or constant.
14430 2018-03-22 Tom de Vries <tom@codesourcery.com>
14432 PR tree-optimization/84956
14433 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
14434 bb_has_abnormal_pred.
14436 2018-03-22 Jakub Jelinek <jakub@redhat.com>
14439 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
14440 DECL_INITIAL (decl) to decl at the end.
14441 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
14442 adjust the comment.
14444 2018-03-21 Joseph Myers <joseph@codesourcery.com>
14446 * doc/extend.texi (__builtin_tgmath): Document when complex
14447 integer types are treated as _Complex _Float64.
14449 2018-03-21 Tom de Vries <tom@codesourcery.com>
14451 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
14453 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14455 PR tree-optimization/84960
14456 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
14457 if it is ENTRY block, move them into single succ of ENTRY in that case.
14459 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
14461 PR tree-optimization/84811
14462 * poly-int.h (poly_span_traits): Remove the T3 parameter and
14463 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
14464 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
14465 (known_subrange_p): Update accordingly. Cast each value involved
14466 in the size comparison, rather than casting the result of the
14469 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14471 PR tree-optimization/84982
14472 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
14473 by flipping the least significant bit rather than all bits from
14474 bitpos to bitpos + bitsize - 1.
14476 2018-03-21 Nathan Sidwell <nathan@acm.org>
14478 * doc/extend.texi (Deprecated Features): Remove mention of
14479 long-deleted deprecations.
14481 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14484 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
14485 * configure: Regenerate.
14487 2018-03-21 Tom de Vries <tom@codesourcery.com>
14489 PR tree-optimization/83126
14490 * tree-parloops.c (num_phis): New function.
14491 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
14493 2018-03-21 Nathan Sidwell <nathan@acm.org>
14495 * doc/extend.texi (Deprecated Features): Update deprecated flags,
14496 mention anon-struct/union members and trailing attributes.
14498 2018-03-21 Bin Cheng <bin.cheng@arm.com>
14500 PR tree-optimization/84969
14501 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
14502 builtin memset partitions if they set different rhs values.
14504 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14506 PR rtl-optimization/84989
14507 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
14508 VEC_DUPLICATE with scalar result mode.
14510 2018-03-21 Martin Liska <mliska@suse.cz>
14513 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
14514 not intended return statement.
14516 2018-03-21 Martin Liska <mliska@suse.cz>
14519 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
14520 (chkp_find_bound_slots_1): Limit number of iterations.
14522 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
14525 * Minor grammar fixes for x86 options.
14527 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14530 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
14531 holding REG_CFA_RESTORE notes, instead turn them into a USE.
14533 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
14536 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
14537 (altivec_lvx_<mode>_1op): Likewise.
14538 (altivec_stvx_<mode>_2op): Likewise.
14539 (altivec_stvx_<mode>_1op): Likewise.
14540 (altivec_lvx_<VM2:mode>): New define_expand.
14541 (altivec_stvx_<VM2:mode>): Likewise.
14542 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
14543 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14544 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
14545 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14546 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
14547 (rs6000_gen_lvx): Likewise.
14548 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
14549 (altivec_expand_stv_builtin): Likewise.
14550 (altivec_expand_builtin): Likewise.
14551 * config/rs6000/vector.md: Likewise.
14553 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14556 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
14559 2018-03-20 Richard Biener <rguenther@suse.de>
14562 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
14563 sign-conversions as zero, fall back to standard scalar_stmt
14566 2018-03-20 Martin Liska <mliska@suse.cz>
14569 * predict.c (rebuild_frequencies): Handle case when we have
14570 PROFILE_ABSENT, but flag_guess_branch_prob is false.
14572 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14575 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
14576 flag_section_anchors.
14577 * varasm.c (use_blocks_for_decl_p): Remove hack for
14578 dw2_force_const_mem.
14581 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
14583 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
14584 be created, use lowpart_subreg of operands[0] rather than operands[0]
14586 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
14587 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
14588 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
14589 and n constraint instead of aarch64_shift_imm_di and Usd.
14590 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
14591 (*aarch64_<optab>_reg_minus<mode>3): ... this.
14593 2018-03-20 Sudakshina Das <sudi.das@arm.com>
14596 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
14597 to favor GPR over NEON registers.
14598 (<shift>di3_neon): Likewise.
14600 2018-03-20 Tom de Vries <tom@codesourcery.com>
14603 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
14604 (nvptx_process_pars): Emit bar.sync asap and alap.
14606 2018-03-20 Tom de Vries <tom@codesourcery.com>
14609 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
14610 seen_label if seen_label is already set.
14612 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14615 * config/i386/i386.c (fold_builtin_cpu): For features above 31
14616 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
14617 Use 1U instead of 1. Formatting fixes.
14620 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
14621 instead of TREE_TYPE (s1) for the return value.
14623 2018-03-19 Jakub Jelinek <jakub@redhat.com>
14625 PR tree-optimization/84946
14626 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
14627 bitsize + bitsize in poly_uint64 rather than poly_int64.
14630 * dwarf2asm.c: Include fold-const.c.
14631 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
14632 of decl rather than decl itself.
14634 PR rtl-optimization/84643
14635 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
14637 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
14640 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
14641 calling assemble_variable.
14643 2018-03-19 Sudakshina Das <sudi.das@arm.com>
14646 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
14647 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
14649 2018-03-19 Jim Wilson <jimw@sifive.com>
14652 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
14653 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
14654 (riscv_first_stack_step): Likewise.
14655 (riscv_option_override): Use STACK_BOUNDARY instead of
14656 MIN_STACK_BOUNDARY.
14657 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
14658 MIN_STACK_BOUNDARY.
14659 (BIGGEST_ALIGNMENT): Set to 128.
14660 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
14661 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
14664 2018-03-19 Richard Biener <rguenther@suse.de>
14666 PR tree-optimization/84933
14667 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
14668 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
14670 2018-03-19 Richard Biener <rguenther@suse.de>
14672 PR tree-optimization/84859
14673 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
14674 (cond_if_else_store_replacement): Perform sinking operation on
14675 single-store BBs regardless of MAX_STORES_TO_SINK setting.
14676 Generalize what a BB with a single eligible store is.
14678 2018-03-19 Richard Biener <rguenther@suse.de>
14680 PR tree-optimization/84929
14681 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
14682 chrec_is_positive against non-chrec arg.
14684 2018-03-19 Tamar Christina <tamar.christina@arm.com>
14687 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
14689 2018-03-18 Martin Liska <mliska@suse.cz>
14691 PR rtl-optimization/84635
14692 * regrename.c (build_def_use): Use matches_mode only when
14695 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
14697 PR tree-optimization/84913
14698 * tree-vect-loop.c (vectorizable_reduction): Don't try to
14699 vectorize chains of COND_EXPRs.
14701 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14703 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
14705 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14707 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
14709 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14711 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
14713 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14714 Kito Cheng <kito.cheng@gmail.com>
14716 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
14717 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
14718 (nds32_adjust_reg_alloc_order): New function.
14719 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
14721 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
14723 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
14724 nds32_print_operand, nds32_print_operand_address): Use
14725 HOST_WIDE_INT_PRINT_DEC instead.
14727 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14729 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
14731 2018-03-17 Jakub Jelinek <jakub@redhat.com>
14734 * config/i386/i386.c (initial_ix86_tune_features,
14735 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
14736 unsigned long long.
14737 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
14738 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
14739 rather than 1u << ix86_tune. Formatting fix.
14740 (ix86_option_override_internal): Change ix86_arch_mask from
14741 unsigned int to unsigned HOST_WIDE_INT, initialize to
14742 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
14743 (ix86_function_specific_restore): Likewise.
14745 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14748 * postreload.c (reload_combine_recognize_pattern): Perform
14749 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
14750 truncate_int_for_mode the result for the destination's mode.
14753 * hsa-gen.c (mem_type_for_type): Fix comment typo.
14754 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
14756 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14759 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
14762 * lra-assigns.c (lra_split_hard_reg_for): Don't use
14763 regno_allocno_class_array and sorted_pseudos.
14764 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
14765 insns where regno is used.
14767 2018-03-16 Martin Liska <mliska@suse.cz>
14770 * multiple_target.c (create_dispatcher_calls): Redirect
14771 reference in the symbol table.
14773 2018-03-16 Martin Liska <mliska@suse.cz>
14776 * multiple_target.c (create_dispatcher_calls): Redirect also
14779 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14783 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
14784 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
14785 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
14787 2018-03-16 Julia Koval <julia.koval@intel.com>
14789 * doc/invoke.texi (Skylake Server): Add CLWB.
14790 Cannonlake): Remove CLWB.
14792 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14794 PR tree-optimization/84841
14795 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
14797 (FLOAT_ONE_CONST_TYPE): Define.
14798 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
14799 (sort_by_operand_rank): Put entries with higher constant_type last
14800 rather than first to match comments.
14802 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
14804 * config/nios2/nios2.md (movsi_internal): Fix thinko in
14807 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14810 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
14811 check and use address of target always.
14813 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
14816 * config/i386/i386.c (indirect_thunk_needed): Update comments.
14817 (indirect_thunk_bnd_needed): Likewise.
14818 (indirect_thunks_used): Likewise.
14819 (indirect_thunks_bnd_used): Likewise.
14820 (indirect_return_needed): New.
14821 (indirect_return_bnd_needed): Likewise.
14822 (output_indirect_thunk_function): Add a bool argument for
14824 (output_indirect_thunk_function): Don't generate alias for
14825 function return thunk.
14826 (ix86_code_end): Call output_indirect_thunk_function to generate
14827 function return thunks.
14828 (ix86_output_function_return): Set indirect_return_bnd_needed
14829 and indirect_return_needed instead of indirect_thunk_bnd_needed
14830 and indirect_thunk_needed.
14832 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
14834 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
14835 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
14836 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
14838 2018-03-15 David Malcolm <dmalcolm@redhat.com>
14839 Paul Hua <paul.hua.gm@gmail.com>
14842 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
14844 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
14846 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
14847 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
14848 resp. SFmode cases.
14850 2018-03-15 Tamar Christina <tamar.christina@arm.com>
14853 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14854 instead of GET_MODE_SIZE when comparing Units.
14856 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14859 * varasm.c (hash_section): Return an unchangeble hash value
14860 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
14861 Return !aarch64_can_use_per_function_literal_pools_p ().
14863 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14866 * optabs.c (emit_conditional_move): Pass address of cmode's copy
14867 rather than address of cmode as last argument to prepare_cmp_insn.
14869 2018-03-15 Julia Koval <julia.koval@intel.com>
14871 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
14872 F_AVX512VNNI, F_AVX512BITALG): New.
14874 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
14877 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
14878 insn for floating-point loads and stores.
14880 2018-03-14 Carl Love <cel@us.ibm.com>
14882 * config/rs6000/rs6000-c.c: Add macro definitions for
14883 ALTIVEC_BUILTIN_VEC_PERMXOR.
14884 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
14885 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
14886 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
14887 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
14889 * config/doc/extend.texi: Add prototypes for vec_permxor.
14891 2018-03-14 David Malcolm <dmalcolm@redhat.com>
14894 * diagnostic-show-locus.c (class layout_point): Convert m_line
14895 from int to linenum_type.
14896 (line_span::comparator): Use linenum "compare" function when
14897 comparing line numbers.
14898 (test_line_span): New function.
14899 (layout_range::contains_point): Convert param "row" from int to
14901 (layout_range::intersects_line_p): Likewise.
14902 (layout::will_show_line_p): Likewise.
14903 (layout::print_source_line): Likewise.
14904 (layout::should_print_annotation_line_p): Likewise.
14905 (layout::print_annotation_line): Likewise.
14906 (layout::print_leading_fixits): Likewise.
14907 (layout::annotation_line_showed_range_p): Likewise.
14908 (struct line_corrections): Likewise for field m_row.
14909 (line_corrections::line_corrections): Likewise for param "row".
14910 (layout::print_trailing_fixits): Likewise.
14911 (layout::get_state_at_point): Likewise.
14912 (layout::get_x_bound_for_row): Likewise.
14913 (layout::print_line): Likewise.
14914 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
14915 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
14916 * input.c (selftest::test_linenum_comparisons): New function.
14917 (selftest::input_c_tests): Call it.
14918 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
14919 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
14920 * selftest.h (ASSERT_GT): New macro.
14921 (ASSERT_GT_AT): New macro.
14922 (ASSERT_LT): New macro.
14923 (ASSERT_LT_AT): New macro.
14925 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
14927 PR rtl-optimization/84780
14928 * combine.c (distribute_links): Don't make a link based on pc_rtx.
14930 2018-03-14 Martin Liska <mliska@suse.cz>
14932 * tree.c (record_node_allocation_statistics): Use
14933 get_stats_node_kind.
14934 (get_stats_node_kind): New function extracted from
14935 record_node_allocation_statistics.
14936 (free_node): Use get_stats_node_kind.
14938 2018-03-14 Richard Biener <rguenther@suse.de>
14940 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
14941 that the value-set of ANTIC_IN doesn't grow.
14944 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14946 (BB_VISITED_WITH_VISITED_SUCCS): New define.
14947 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14949 2018-03-14 Julia Koval <julia.koval@intel.com>
14951 * config.gcc (icelake-client, icelake-server): New.
14953 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
14954 (initial_ix86_arch_features): Ditto.
14955 (PTA_SKYLAKE): Add SGX.
14956 (PTA_ICELAKE): Remove.
14957 (PTA_ICELAKE_CLIENT): New.
14958 (PTA_ICELAKE_SERVER): New.
14959 (ix86_option_override_internal): Split up icelake on icelake client and
14961 (get_builtin_code_for_version): Ditto.
14962 (fold_builtin_cpu): Ditto.
14963 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
14964 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
14965 * config/i386/i386.h (processor_type): Ditto.
14966 * doc/invoke.texi: Ditto.
14968 2018-03-14 Jakub Jelinek <jakub@redhat.com>
14971 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
14972 INTEGER_CST offset, add it together with bitpos / 8 and
14973 sign extend based on POINTER_SIZE.
14977 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
14980 * config/i386/constraints.md (Yc): New register constraint.
14981 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
14982 Use Yc constraint for alternative 2 of operand 0. Remove
14983 preferred_for_speed attribute.
14985 2018-03-14 Richard Biener <rguenther@suse.de>
14987 PR tree-optimization/84830
14988 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
14989 with the old one to avoid oscillations.
14991 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
14994 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
14996 (assign_by_spills): Return a flag of reload assignment failure.
14997 Do not process the reload assignment failures. Do not spill other
14998 reload pseudos if they has the same reg class. Update n if
15000 (lra_assign): Add a return arg. Set up from the result of
15001 assign_by_spills call.
15002 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15003 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15004 usage_insns if it is not NULL.
15005 (spill_hard_reg_in_range): New function.
15006 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15007 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15008 function prototypes.
15009 (lra_assign): Change prototype.
15010 * lra.c (lra): Add code to deal with fails by splitting hard reg
15013 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
15015 * config/riscv/riscv.opt (mrelax): New option.
15016 * config/riscv/riscv.c (riscv_file_start): Emit ".option
15017 "norelax" when riscv_mrelax is disabled.
15018 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
15020 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15023 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
15024 reassociation for int modes.
15026 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
15028 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
15029 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
15031 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
15032 * config/aarch64/aarch64-sve.md
15033 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
15034 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
15035 (*extend<mode><Vwide>2): Rename to...
15036 (aarch64_sve_extend<mode><Vwide>2): ...this.
15037 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
15038 renaming the old pattern to...
15039 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
15041 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
15042 define_expand, renaming the old pattern to...
15043 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
15044 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
15045 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
15046 account when deciding which SVE instruction the optab should use.
15047 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
15049 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
15051 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
15052 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
15053 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
15054 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
15055 (tlsdesc_small_<mode>): Turn a define_expand and use
15056 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
15057 (tlsdesc_small_advsimd_<mode>): ...this.
15058 (tlsdesc_small_sve_<mode>): New pattern.
15060 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
15062 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
15063 (UNSPEC_UMUL_HIGHPART): New constants.
15064 (MUL_HIGHPART): New int iteraor.
15065 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
15066 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
15068 (*<su>mul<mode>3_highpart): New define_insn.
15070 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
15073 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
15076 2018-03-13 Martin Liska <mliska@suse.cz>
15079 * (sem_item_optimizer::sem_item_optimizer): Initialize new
15081 (sem_item_optimizer::~sem_item_optimizer): Release it.
15082 (sem_item_optimizer::merge_classes): Register variable aliases.
15083 (sem_item_optimizer::fixup_pt_set): New function.
15084 (sem_item_optimizer::fixup_points_to_sets): Likewise.
15085 * ipa-icf.h: Declare new variables and functions.
15087 2018-03-13 Jakub Jelinek <jakub@redhat.com>
15089 PR middle-end/84834
15090 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
15091 integer_pow2p@2 and test integer_pow2p in condition.
15092 (A < 0 ? C : 0): Similarly for @1.
15094 PR middle-end/84831
15095 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
15096 characters starting at p contain '\0' character, don't look beyond
15100 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
15101 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
15104 * reg-stack.c (change_stack): Change update_end var from int to
15105 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
15106 also call set_block_for_insn on the newly added insns and rescan.
15109 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
15110 on the last operand.
15113 * tree.c (stabilize_reference_1): Return save_expr (e) for
15114 STATEMENT_LIST even if it doesn't have side-effects.
15116 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
15118 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
15120 2018-03-12 Renlin Li <renlin.li@arm.com>
15122 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
15123 aarch64_output_scalar_simd_mov_immediate.
15125 2018-03-12 Martin Sebor <msebor@redhat.com>
15127 PR tree-optimization/83456
15128 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
15129 for perfectly overlapping calls to memcpy.
15130 (gimple_fold_builtin_memory_chk): Same.
15131 (gimple_fold_builtin_strcpy): Handle no-warning.
15132 (gimple_fold_builtin_stxcpy_chk): Same.
15133 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
15135 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
15137 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
15138 parameter. Use it for SFmode.
15139 (rs6000_function_arg_advance_1): Adjust.
15140 (rs6000_function_arg): Adjust.
15141 (rs6000_gimplify_va_arg): Pass false for that new parameter.
15143 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
15145 PR rtl-optimization/84169
15146 PR rtl-optimization/84780
15147 * combine.c (can_combine_p): Check for a 2-insn combination whether
15148 the destination register is used between the two insns, too.
15150 2018-03-12 Richard Biener <rguenther@suse.de>
15152 PR tree-optimization/84803
15153 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
15154 for refs DR analysis didn't process.
15156 2018-03-12 Richard Biener <rguenther@suse.de>
15158 PR tree-optimization/84777
15159 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
15160 force-vectorize loops ignore whether we are optimizing for size.
15162 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
15164 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
15165 (TARGET_MD_ASM_ADJUST): Define.
15167 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
15168 Kito Cheng <kito.cheng@gmail.com>
15169 Chung-Ju Wu <jasonwucj@gmail.com>
15171 * config/nds32/nds32.c (nds32_compute_stack_frame,
15172 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
15173 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
15174 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
15175 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
15176 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
15177 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
15178 * config/nds32/nds32.md (prologue, epilogue): Use macro
15179 NDS32_V3PUSH_AVAILABLE_P to do checking.
15181 2018-03-11 Jakub Jelinek <jakub@redhat.com>
15184 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
15185 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
15186 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
15187 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
15188 addition of most attributes on !orig_type_die or the attribute not
15189 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
15191 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15192 Chung-Ju Wu <jasonwucj@gmail.com>
15194 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
15195 __NDS32_VH__ macro.
15196 * config/nds32/nds32.opt (mvh): New option.
15198 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15199 Chung-Ju Wu <jasonwucj@gmail.com>
15201 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
15203 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
15204 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
15207 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15208 Chung-Ju Wu <jasonwucj@gmail.com>
15210 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
15212 * config/nds32/nds32-multiple.md (strlensi): New pattern.
15213 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
15215 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
15216 Kito Cheng <kito.cheng@gmail.com>
15217 Chung-Ju Wu <jasonwucj@gmail.com>
15219 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
15220 UNSPEC_FFMISM and UNSPEC_FLMISM.
15221 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
15222 for ffb, ffmism and flmism.
15223 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
15224 (unspec_ffmism): Ditto.
15225 (unspec_flmism): Ditto.
15226 (nds32_expand_builtin_impl): Check if string extension is available.
15227 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
15228 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
15230 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
15233 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
15236 * lra-assigns.c (assign_by_spills): Return a flag of reload
15237 assignment failure. Do not process the reload assignment
15238 failures. Do not spill other reload pseudos if they has the same
15240 (lra_assign): Add a return arg. Set up from the result of
15241 assign_by_spills call.
15242 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15243 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15244 usage_insns if it is not NULL.
15245 (spill_hard_reg_in_range): New function.
15246 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15247 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15248 function prototypes.
15249 (lra_assign): Change prototype.
15250 * lra.c (lra): Add code to deal with fails by splitting hard reg
15253 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
15256 * config/i386/i386.opt: Replace Enforcment with Enforcement.
15258 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
15261 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
15262 (dw_val_node): Add val_symbolic_view.
15263 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
15264 (symview_upper_bound): New.
15265 (new_line_info_table): Initialize symviews_since_reset.
15266 (dwarf2out_source_line): Count symviews_since_reset and set
15267 symview_upper_bound.
15268 (dw_val_equal_p): Handle symview.
15269 (add_AT_symview): New.
15270 (print_dw_val): Handle symview.
15271 (attr_checksum, attr_checksum_ordered): Likewise.
15272 (same_dw_val_p, size_of_die): Likewise.
15273 (value_format, output_die): Likewise.
15274 (add_high_low_attributes): Use add_AT_symview for entry_view.
15275 (dwarf2out_finish): Reset symview_upper_bound, clear
15278 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
15281 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
15282 Add strict argument and use it.
15283 (rs6000_split_multireg_move): Update for new strict argument.
15284 (mem_operand_gpr): Disallow all non-offsettable addresses.
15285 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
15287 2018-03-09 Jakub Jelinek <jakub@redhat.com>
15290 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
15291 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
15292 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
15295 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
15296 decl, use remap_type if we want to use the type.
15298 2018-03-09 Martin Sebor <msebor@redhat.com>
15300 PR tree-optimization/84526
15301 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
15303 (builtin_access::generic_overlap): Be prepared to handle non-array
15306 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
15308 PR rtl-optimization/84682
15309 * lra-constraints.c (process_address_1): Check is_address flag
15310 for address constraints.
15311 (process_alt_operands): Likewise.
15312 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
15313 preprocess_constraints.
15314 * recog.h (preprocess_constraints): Add oploc parameter.
15316 * recog.c (preprocess_constraints): Test address_operand for
15317 CT_ADDRESS constraints.
15319 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
15322 * lra-assigns.c (assign_by_spills): Return a flag of reload
15323 assignment failure. Do not process the reload assignment
15324 failures. Do not spill other reload pseudos if they has the same
15326 (lra_assign): Add a return arg. Set up from the result of
15327 assign_by_spills call.
15328 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15329 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15330 usage_insns if it is not NULL.
15331 (spill_hard_reg_in_range): New function.
15332 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15333 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15334 function prototypes.
15335 (lra_assign): Change prototype.
15336 * lra.c (lra): Add code to deal with fails by splitting hard reg
15339 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15342 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
15343 Accept complain bool parameter. Only emit errors if it is true.
15344 (arm_parse_cpu_option_name): Likewise.
15345 (arm_target_thumb_only): Adjust callers of the above.
15346 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
15347 prototype to take a default true bool parameter.
15348 (arm_parse_arch_option_name): Likewise.
15350 2018-03-09 David Malcolm <dmalcolm@redhat.com>
15351 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
15355 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
15356 * configure: Regenerate.
15357 * configure.ac ("linker --version-script option"): New.
15358 ("linker soname option"): New.
15360 2018-03-09 Richard Biener <rguenther@suse.de>
15362 PR tree-optimization/84775
15363 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
15364 immediate uses of predicate stmts and mark them modified.
15367 PR tree-optimization/84178
15368 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
15370 (version_loop_for_if_conversion): Delay update_ssa call.
15371 (tree_if_conversion): Delay update_ssa until after predicate
15374 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
15377 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
15378 when the function accesses prior frames.
15380 2018-03-08 Jakub Jelinek <jakub@redhat.com>
15383 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
15384 gen_llsym, otherwise call maybe_gen_llsym.
15386 PR inline-asm/84742
15387 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
15388 has ',' character inside of it.
15390 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15393 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
15394 as clobbering CC_REGNUM.
15396 2018-03-08 Richard Biener <rguenther@suse.de>
15398 PR middle-end/84552
15399 * tree-scalar-evolution.c: Include tree-into-ssa.h.
15400 (follow_copies_to_constant): Do not follow SSA names registered
15403 2018-03-08 Richard Biener <rguenther@suse.de>
15405 PR tree-optimization/84178
15406 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
15408 (version_loop_for_if_conversion): Delay update_ssa call.
15409 (tree_if_conversion): Delay update_ssa until after predicate
15412 2018-03-08 David Malcolm <dmalcolm@redhat.com>
15414 PR tree-optimization/84178
15415 * tree-if-conv.c (release_bb_predicate): Remove the
15416 the assertion that the stmts have NULL use_ops.
15417 Discard the statements, asserting that they haven't
15418 yet been added to a BB.
15420 2018-03-08 Richard Biener <rguenther@suse.de>
15422 PR tree-optimization/84746
15423 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
15424 (phi_translate): Pass in destination ANTIC_OUT set.
15425 (phi_translate_1): Likewise. For a simplified result lookup
15426 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
15427 (phi_translate_set): Adjust.
15428 (do_pre_regular_insertion): Likewise.
15429 (do_pre_partial_partial_insertion): Likewise.
15431 2018-03-08 Martin Liska <mliska@suse.cz>
15433 PR gcov-profile/84735
15434 * doc/gcov.texi: Document usage of profile files.
15435 * gcov-io.h: Document changes in the format.
15437 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
15441 * dwarf2out.c (struct dw_line_info_table): Update comments for
15443 (FORCE_RESET_NEXT_VIEW): New.
15444 (FORCE_RESETTING_VIEW_P): New.
15445 (RESETTING_VIEW_P): Check for -1 too.
15446 (ZERO_VIEW_P): Likewise.
15447 (new_line_info_table): Force-reset next view.
15448 (dwarf2out_begin_function): Likewise.
15449 (dwarf2out_source_line): Simplify zero_view_p initialization.
15450 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
15451 view directly. Omit view when omitting .loc at line 0.
15453 2018-03-08 Jakub Jelinek <jakub@redhat.com>
15455 PR tree-optimization/84740
15456 * tree-switch-conversion.c (process_switch): Call build_constructors
15457 only if info.phi_count is non-zero.
15459 PR tree-optimization/84739
15460 * tree-tailcall.c (find_tail_calls): Check call arguments against
15461 DECL_ARGUMENTS (current_function_decl) rather than
15462 DECL_ARGUMENTS (func) when checking for tail recursion.
15464 2018-03-07 Jakub Jelinek <jakub@redhat.com>
15466 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
15467 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
15468 Volker Reichelt's entry and add entries for people that perform
15469 GCC fuzzy testing and report numerous bugs.
15471 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
15474 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
15475 readonly data in sdata, if that is disabled.
15476 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
15477 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
15478 -mreadonly-in-sdata option.
15480 2018-03-07 Martin Sebor <msebor@redhat.com>
15482 PR tree-optimization/84468
15483 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
15484 basic block when looking for nul assignment.
15486 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
15489 * except.h (output_function_exception_table): Adjust prototype.
15490 * except.c (output_function_exception_table): Remove FNNAME parameter
15491 and add SECTION parameter. Ouput one part of the table at a time.
15492 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
15493 the first part of the exception table and emit unwind directives.
15494 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
15495 (i386_pe_seh_cold_init): Likewise.
15496 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
15497 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
15498 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
15499 (ix86_output_call_insn): Emit a nop in one more case for SEH.
15500 * config/i386/winnt.c: Include except.h.
15501 (struct seh_frame_state): Add reg_offset, after_prologue and
15502 in_cold_section fields.
15503 (i386_pe_seh_end_prologue): Set seh->after_prologue.
15504 (i386_pe_seh_cold_init): New function.
15505 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
15506 to seh->in_cold_section.
15507 (seh_emit_push): Record the offset of the push.
15508 (seh_emit_save): Record the offet of the save.
15509 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
15510 Test seh->after_prologue to disregard the epilogue.
15511 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
15512 (i386_pe_end_cold_function): New function.
15514 2018-03-07 Jakub Jelinek <jakub@redhat.com>
15517 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
15518 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
15521 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
15523 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
15524 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
15526 PR middle-end/84723
15527 * multiple_target.c: Include tree-inline.h and intl.h.
15528 (expand_target_clones): Diagnose and fail if node->definition and
15529 !tree_versionable_function_p (node->decl).
15531 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
15533 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
15535 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
15536 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
15537 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
15539 2018-03-06 Jakub Jelinek <jakub@redhat.com>
15542 * combine.c (try_combine): Use reg_or_subregno instead of handling
15543 just paradoxical SUBREGs and REGs.
15545 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
15547 * config/arc/arc.c (arc_finalize_pic): Remove function.
15548 (arc_must_save_register): We use single base PIC register, remove
15549 checks to save/restore the PIC register.
15550 (arc_expand_prologue): Likewise.
15551 * config/arc/arc-protos.h (arc_set_default_type_attributes):
15553 (arc_verify_short): Likewise.
15554 (arc_attr_type): Likewise.
15555 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
15556 (walk_stores): Likewise.
15557 (arc_address_cost): Make it static.
15558 (arc_verify_short): Likewise.
15559 (branch_dest): Likewise.
15560 (arc_attr_type): Likewise.
15561 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
15562 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
15563 (arc_final_prescan_insn): Remove inserting the nops due to
15564 hardware hazards. It is done in reorg step.
15565 (insn_length_variant_t): Remove.
15566 (insn_length_parameters_t): Likewise.
15567 (arc_insn_length_parameters): Likewise.
15568 (arc_get_insn_variants): Likewise.
15569 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
15571 2018-03-06 Jakub Jelinek <jakub@redhat.com>
15573 PR inline-asm/84683
15574 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
15577 PR tree-optimization/84687
15578 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
15580 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
15582 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15584 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
15585 Rename to ppc_speculation_barrier.
15586 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
15587 __builtin_ppc_speculation_barrier.
15589 2018-03-05 Jakub Jelinek <jakub@redhat.com>
15592 * combine.c (combine_simplify_rtx): Don't try to simplify if
15593 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
15596 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
15598 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
15599 to 32 bytes when compiling for POWER9.
15601 2018-03-05 Jakub Jelinek <jakub@redhat.com>
15604 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
15605 regparm >= 3 with no arg reg available also for calls with
15606 flag_force_indirect_call. Pass decl to ix86_function_regparm.
15609 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
15611 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
15613 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
15616 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
15618 2018-03-05 Richard Biener <rguenther@suse.de>
15620 PR tree-optimization/84486
15621 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
15622 When inserting a __builtin_assume_aligned call set the LHS
15623 SSA name alignment info accordingly.
15625 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
15627 PR tree-optimization/84114
15628 * config/aarch64/aarch64.c (aarch64_reassociation_width)
15629 Avoid reassociation of FLOAT_MODE addition.
15631 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
15633 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
15634 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
15635 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
15636 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
15637 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
15638 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
15639 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
15641 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
15642 __builtin_ia32_wbinvd): New builtins.
15643 (SPECIAL_ARGS2): New.
15644 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
15645 (SPECIAL_ARGS2): New.
15646 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
15647 (ix86_valid_target_attribute_inner_p): Ditto.
15648 (ix86_init_mmx_sse_builtins): Add special_args2.
15649 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
15650 TARGET_WBNOINVD_P): New.
15651 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
15652 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
15653 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
15654 * config/i386/immintrin.h (_wbinvd): New intrinsic.
15655 * config/i386/pconfigintrin.h: New file.
15656 * config/i386/wbnoinvdintrin.h: Ditto.
15657 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
15659 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
15661 2018-03-05 Richard Biener <rguenther@suse.de>
15663 PR tree-optimization/84670
15664 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
15666 (BB_VISITED_WITH_VISITED_SUCCS): New define.
15667 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
15668 (compute_antic_aux): Only assert the number of values in ANTIC_IN
15669 doesn't grow if all successors (recursively) were visited at least
15672 2018-03-05 Richard Biener <rguenther@suse.de>
15674 PR tree-optimization/84650
15675 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
15676 if executed in the loop pipeline.
15678 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
15680 * doc/configfiles.texi (Configuration Files): Move info about
15681 conditionalizing $target-protos.h to...
15682 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
15683 differs from $target-protos.h.
15685 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
15686 Chung-Ju Wu <jasonwucj@gmail.com>
15688 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
15689 * config/nds32/nds32-multiple.md (setmemsi): Define.
15690 * config/nds32/nds32-memory-manipulation.c
15691 (nds32_gen_dup_4_byte_to_word_value): New.
15692 (emit_setmem_word_loop): New.
15693 (emit_setmem_byte_loop): New.
15694 (nds32_expand_setmem_loop): New.
15695 (nds32_expand_setmem_loop_v3m): New.
15696 (nds32_expand_setmem_unroll): New.
15697 (nds32_expand_setmem): New.
15699 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15700 Chung-Ju Wu <jasonwucj@gmail.com>
15702 * config/nds32/nds32-memory-manipulation.c
15703 (nds32_emit_load_store): New.
15704 (nds32_emit_post_inc_load_store): New.
15705 (nds32_emit_mem_move): New.
15706 (nds32_emit_mem_move_block): New.
15707 (nds32_expand_movmemsi_loop_unknown_size): New.
15708 (nds32_expand_movmemsi_loop_known_size): New.
15709 (nds32_expand_movmemsi_loop): New.
15710 (nds32_expand_movmemsi_unroll): New.
15711 (nds32_expand_movmemqi): Rename ...
15712 (nds32_expand_movmemsi): ... to this.
15713 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
15714 (movmemsi): ... to this.
15715 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
15716 (nds32_expand_movmemsi): ... to this.
15718 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15719 Monk Chiang <sh.chiang04@gmail.com>
15720 Chung-Ju Wu <jasonwucj@gmail.com>
15722 * config/nds32/nds32-protos.h
15723 (nds32_expand_load_multiple): New arguments.
15724 (nds32_expand_store_multiple): Ditto.
15725 (nds32_valid_multiple_load_store): Rename ...
15726 (nds32_valid_multiple_load_store_p): ... to this.
15727 * config/nds32/nds32-memory-manipulation.c
15728 (nds32_expand_load_multiple): Refine implementation.
15729 (nds32_expand_store_multiple): Ditto.
15730 * config/nds32/nds32-multiple.md
15731 (load_multiple): Update nds32_expand_load_multiple interface.
15732 (store_multiple): Update nds32_expand_store_multiple interface.
15733 * config/nds32/nds32-predicates.c
15734 (nds32_valid_multiple_load_store): Rename ...
15735 (nds32_valid_multiple_load_store_p): ... to this and refine
15737 * config/nds32/predicates.md
15738 (nds32_load_multiple_and_update_address_operation): New predicate.
15739 (nds32_store_multiple_and_update_address_operation): New predicate.
15741 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15742 Chung-Ju Wu <jasonwucj@gmail.com>
15744 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
15745 (combo): New attribute.
15746 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
15748 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
15750 * config/nds32/nds32.opt: Change -mcmodel= default value.
15752 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
15753 Monk Chiang <sh.chiang04@gmail.com>
15754 Chung-Ju Wu <jasonwucj@gmail.com>
15756 * config/nds32/constants.md (unspec_element): New enum.
15757 * config/nds32/constraints.md (Umw): New constraint.
15758 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
15759 * config/nds32/nds32-intrinsic.md: Likewise.
15760 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
15761 (nds32_valid_smw_lwm_base_p): New.
15762 (nds32_output_smw_single_word): New.
15763 (nds32_output_lmw_single_word): New.
15764 (nds32_expand_unaligned_load): New.
15765 (nds32_expand_unaligned_store): New.
15766 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
15767 (nds32_output_smw_single_word): Declare.
15768 (nds32_output_lmw_single_word): Declare.
15769 (nds32_expand_unaligned_load): Declare.
15770 (nds32_expand_unaligned_store): Declare.
15771 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
15772 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
15773 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
15774 NDS32_BUILTIN_UASTORE_DW.
15775 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
15778 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
15779 Kito Cheng <kito.cheng@gmail.com>
15780 Chung-Ju Wu <jasonwucj@gmail.com>
15782 * config/nds32/nds32-intrinsic.c
15783 (nds32_expand_builtin_null_ftype_reg): Delete.
15784 (nds32_expand_builtin_reg_ftype_imm): Ditto.
15785 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
15786 (nds32_read_argument): New.
15787 (nds32_legitimize_target): Ditto.
15788 (nds32_legitimize_argument): Ditto.
15789 (nds32_check_constant_argument): Ditto.
15790 (nds32_expand_unop_builtin): Ditto.
15791 (nds32_expand_unopimm_builtin): Ditto.
15792 (nds32_expand_binop_builtin): Ditto.
15793 (nds32_builtin_decl_impl): Ditto.
15794 (builtin_description): Ditto.
15795 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
15796 (nds32_init_builtins_impl): Ditto.
15797 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
15798 (nds32_builtin_decl): New.
15799 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
15800 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
15802 2018-03-02 Jeff Law <law@redhat.com>
15804 * reorg.c (stop_search_p): Handle DEBUG_INSN.
15805 (redundant_insn, fill_simple_delay_slots): Likewise.
15806 (fill_slots_from_thread): Likewise.
15807 * resource.c (mark_referenced_resources): Likewise.
15808 (mark_set_resources, find_dead_or_set_registers): Likewise.
15810 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15812 * substring-locations.h (format_warning_va): Formatting fix for
15813 ATTRIBUTE_GCC_DIAG.
15814 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
15816 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
15817 * substring-locations.c: Include intl.h.
15818 (format_warning_va): Turned into small wrapper around
15819 format_warning_n_va, renamed to ...
15820 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
15821 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
15823 (format_warning_at_substring_n): New function.
15824 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
15825 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
15826 format_warning_at_substring with just a shorter name instead of
15827 const function pointer.
15828 (fmtwarn_n): New function.
15829 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
15830 appropriate, get rid of all the fmtstr temporaries, move conditionals
15831 with G_() wrapped string literals directly into fmtwarn arguments,
15832 cast dir.len to (int), formatting fixes.
15834 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
15836 * doc/invoke.texi: Remove "Cilk Plus" references.
15838 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15839 Richard Biener <rguenther@suse.de>
15842 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
15843 for error or warning attributes if CALL_FROM_THUNK_P is set.
15846 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15849 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
15850 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
15853 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
15854 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
15856 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
15857 instead of -1U in last predictors element's probability member.
15859 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15862 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15863 arguments if they are comparable.
15865 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15867 PR tree-optimization/84634
15868 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
15869 masks and masked_loop_p with a single loop_masks, making sure it's
15870 null for bb vectorization.
15872 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15874 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
15875 (vect_analyze_data_ref_access): Use loop->safe_len rather than
15876 loop->force_vectorize to check whether there is no alias.
15878 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15881 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
15883 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
15885 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
15886 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
15887 instead of a loop around prev_real_insn.
15888 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
15891 PR inline-asm/84625
15892 * config/i386/i386.c (ix86_print_operand): Use conditional
15893 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
15896 2018-03-02 Richard Biener <rguenther@suse.de>
15898 PR tree-optimization/84427
15899 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
15900 (bitmap_set_subtract_values): Rewrite to handle multiple
15903 (prune_clobbered_mems): Likewise.
15904 (phi_translate): Take edge instead of pred/phiblock.
15905 (phi_translate_1): Likewise.
15906 (phi_translate_set): Likewise. Insert all translated
15907 exprs for a value into the set, keeping possibly multiple
15908 expressions per value.
15909 (compute_antic_aux): Adjust for phi_translate changes.
15910 When intersecting union the expressions and prune those
15911 not in the final value set, keeping possibly multiple
15912 expressions per value. Do not use value-insertion
15913 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
15914 all expressions. Add verification that the value-sets
15915 only shrink during iteration.
15916 (compute_partial_antic_aux): Adjust for the phi_translate changes.
15917 (do_pre_regular_insertion): Likewise.
15918 (do_pre_partial_partial_insertion): Likewise.
15920 2018-03-02 Richard Biener <rguenther@suse.de>
15923 * config/darwin.c (saved_debug_info_level): New static global.
15924 (darwin_asm_lto_start): Disable debug info generation for LTO out.
15925 (darwin_asm_lto_end): Restore debug info generation settings.
15927 2018-03-01 Martin Liska <mliska@suse.cz>
15930 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
15931 volatile arguments.
15933 2018-03-01 Richard Biener <rguenther@suse.de>
15936 * dwarf2out.c (gen_variable_die): Properly handle late VLA
15937 type annotation with LTO when debug was disabled at compile-time.
15939 2018-03-01 Matthew Fortune <mfortune@gmail.com>
15941 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
15943 (mips_final_postscan_insn): Likewise.
15945 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
15947 PR rtl-optimization/84528
15948 * alias.c (init_alias_target): Add commentary.
15949 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
15950 a unique base value if the frame pointer is not eliminated
15951 to the stack pointer.
15953 2018-03-01 Tom de Vries <tom@codesourcery.com>
15955 PR rtl-optimization/83327
15956 * lra-int.h (hard_regs_spilled_into): Declare.
15957 * lra.c (hard_regs_spilled_into): Define.
15958 (init_reg_info): Init hard_regs_spilled_into.
15959 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
15960 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
15961 (process_bb_lives): Handle hard_regs_spilled_into.
15962 (lra_create_live_ranges_1): Before doing liveness propagation, clear
15963 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
15965 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
15967 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
15968 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
15969 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
15970 * config/rs6000/aix72.h: New file.
15972 2018-02-28 Jakub Jelinek <jakub@redhat.com>
15974 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
15975 instead of warning_at with conditional singular and plural messages
15979 * stor-layout.c (update_alignment_for_field): For
15980 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
15981 && !DECL_PACKED (field), do the alignment update, just use
15982 only desired_align instead of MAX (type_align, desired_align)
15984 (place_field): Don't do known_align < desired_align handling
15985 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
15986 is non-NULL, instead do it after rli->prev_field handling and
15987 only if not within a bitfield word. For DECL_PACKED (field)
15988 use type_align of BITS_PER_UNIT.
15990 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
15992 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
15993 superfluous parentheses and trailing spaces.
15995 2018-02-28 Richard Biener <rguenther@suse.de>
15997 PR tree-optimization/84584
15998 * graphite-scop-detection.c (scop_detection::add_scop): Discard
15999 SCoPs with fake exit edge.
16001 2018-02-28 Martin Liska <mliska@suse.cz>
16004 * timevar.c (timer::print): Fix format to properly print 100%
16007 2018-02-28 Richard Biener <rguenther@suse.de>
16009 PR middle-end/84607
16010 * genmatch.c (capture_info::walk_match): Do not mark
16011 captured expressions without operands as expr_p given
16012 they act more like predicates and should be subject to
16013 "lost tail" side-effect preserving.
16015 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
16017 PR rtl-optimization/81611
16018 * auto-inc-dec.c (attempt_change): Move dead note from
16019 mem_insn if it's the next use of regno
16020 (find_address): Take address use of reg holding
16021 non-incremented value. Add parm to limit search to the named
16023 (merge_in_block): Attempt to use a mem insn that is the next
16024 use of the original regno.
16026 2018-02-27 Martin Sebor <msebor@redhat.com>
16029 * doc/invoke.texi (-Wmissing-attributes): New option.
16030 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
16032 2018-02-27 Martin Sebor <msebor@redhat.com>
16034 PR translation/84207
16035 * diagnostic-core.h (warning_n, error_n, inform_n): Change
16036 n argument to unsigned HOST_WIDE_INT.
16037 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
16038 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
16039 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
16040 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
16042 2018-02-27 Richard Biener <rguenther@suse.de>
16044 PR tree-optimization/84512
16045 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
16046 Do not use the estimate returned from record_stmt_cost for
16047 the scalar iteration cost but sum properly using add_stmt_cost.
16049 2018-02-27 Richard Biener <rguenther@suse.de>
16051 PR tree-optimization/84466
16052 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
16053 Adjust last change to less strictly validate use operands.
16055 2018-02-27 Martin Liska <mliska@suse.cz>
16057 PR gcov-profile/84548
16058 * gcov.c (process_file): Allow partial overlap and consider it
16059 also as group functions.
16060 (output_lines): Properly calculate range of lines for a group.
16062 2018-02-27 Martin Liska <mliska@suse.cz>
16064 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
16065 'ggc' suffixes. Change first column width.
16066 (timer::print): Fix formatting of the column.
16068 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
16070 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
16071 preserve inline entry blocks for the sake of debug inline
16072 entry point markers alone.
16073 (remove_unused_locals): Suggest in comments a better place to
16074 force the preservation of inline entry blocks that are
16075 otherwise unused, but do not preserve them.
16077 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
16079 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
16081 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
16084 * config/i386/constraints.md (Bs): Replace
16085 ix86_indirect_branch_register with
16086 TARGET_INDIRECT_BRANCH_REGISTER.
16088 * config/i386/i386.md (indirect_jump): Likewise.
16089 (tablejump): Likewise.
16090 (*sibcall_memory): Likewise.
16091 (*sibcall_value_memory): Likewise.
16092 Peepholes of indirect call and jump via memory: Likewise.
16093 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
16094 (*sibcall_value_GOT_32): Likewise.
16095 * config/i386/predicates.md (indirect_branch_operand): Likewise.
16096 (GOT_memory_operand): Likewise.
16097 (call_insn_operand): Likewise.
16098 (sibcall_insn_operand): Likewise.
16099 (GOT32_symbol_operand): Likewise.
16100 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
16102 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
16104 PR rtl-optimization/83496
16105 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
16106 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
16107 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
16108 redundant insn, if any.
16109 (relax_delay_slots): Likewise.
16110 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
16112 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
16114 PR tree-optimization/83965
16115 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
16116 that grouped statements are part of a reduction chain. Return
16117 true if the statement is not marked as a reduction itself but
16118 is part of a group.
16119 (vect_recog_dot_prod_pattern): Don't check whether the statement
16120 is part of a group here.
16121 (vect_recog_sad_pattern): Likewise.
16122 (vect_recog_widen_sum_pattern): Likewise.
16124 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
16127 * final.c (rest_of_clean_state): Also look for calls inside sequences.
16129 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
16132 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
16134 (ix86_output_indirect_function_return): New prototype.
16135 (ix86_split_simple_return_pop_internal): Likewise.
16136 * config/i386/i386.c (indirect_return_via_cx): New.
16137 (indirect_return_via_cx_bnd): Likewise.
16138 (indirect_thunk_name): Handle return va CX_REG.
16139 (output_indirect_thunk_function): Create alias for
16140 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
16141 (ix86_output_indirect_jmp): Remove the bool argument.
16142 (ix86_output_indirect_function_return): New function.
16143 (ix86_split_simple_return_pop_internal): Likewise.
16144 * config/i386/i386.md (*indirect_jump): Don't pass false
16145 to ix86_output_indirect_jmp.
16146 (*tablejump_1): Likewise.
16147 (simple_return_pop_internal): Change it to define_insn_and_split.
16148 Call ix86_split_simple_return_pop_internal to split it for
16149 -mfunction-return=.
16150 (simple_return_indirect_internal): Call
16151 ix86_output_indirect_function_return instead of
16152 ix86_output_indirect_jmp.
16154 2018-02-26 Jakub Jelinek <jakub@redhat.com>
16157 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
16158 memset and value initialization afterwards.
16160 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
16162 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
16164 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16167 * common/config/aarch64/aarch64-common.c
16168 (aarch_option_optimization_table[]): Switch
16169 off fomit-frame-pointer
16171 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
16172 Chung-Ju Wu <jasonwucj@gmail.com>
16174 * config/nds32/nds32-multiple.md (load_multiple): Disallow
16176 (store_multiple): Ditto.
16178 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
16180 * config.gcc: Add --with-cpu support for nds32 target.
16181 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
16182 * config/nds32/nds32.opt: Add -mcpu= option.
16184 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
16186 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
16187 isel=yes): Warn for these deprecated options.
16189 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
16191 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
16192 ISA_2_5_MASKS_EMBEDDED.
16194 2018-02-23 Jakub Jelinek <jakub@redhat.com>
16196 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
16197 p->max as pointers rather than using iterative_hash_expr.
16199 2018-02-23 Carl Love <cel@us.ibm.com>
16201 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
16202 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
16204 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
16205 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
16206 P8V_BUILTIN_VEC_VUNSIGNED2.
16208 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
16211 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
16212 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
16214 * lra-constraints.c (curr_insn_transform): Set up
16215 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
16217 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
16218 * lra-eliminations.c (spill_pseudos): Ditto.
16219 (process_insn_for_elimination): Ditto.
16220 * lra-lives.c (reg_early_clobber_p): Use the new macros.
16221 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
16222 LRA_NON_CLOBBERED_ALT.
16224 2018-02-22 Martin Sebor <msebor@redhat.com>
16226 PR tree-optimization/84480
16227 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
16228 to maybe_diag_stxncpy_trunc. Call it.
16229 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
16230 from gimple_fold_builtin_strcpy. Print inlining stack.
16231 (handle_builtin_stxncpy): Print inlining stack.
16232 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
16234 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
16237 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
16238 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
16239 and -fcheck-pointer-bounds are used together.
16240 (indirect_thunk_prefix): New enum.
16241 (indirect_thunk_need_prefix): New function.
16242 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
16243 "_nt" instead of "_bnd" for NOTRACK prefix.
16244 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
16245 (output_indirect_thunk_function): Likewise.
16247 (ix86_code_end): Update output_indirect_thunk_function calls.
16248 (ix86_output_indirect_branch_via_reg): Replace
16249 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
16250 (ix86_output_indirect_branch_via_push): Likewise.
16251 (ix86_output_function_return): Likewise.
16252 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
16253 incompatible with -fcf-protection=branch and
16254 -fcheck-pointer-bounds.
16256 2018-02-22 Steve Ellcey <sellcey@cavium.com>
16259 * config/aarch64/aarch64.c (aarch64_print_address_internal):
16260 Change gcc_assert call to output_operand_lossage.
16262 2018-02-22 Steve Ellcey <sellcey@cavium.com>
16264 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
16266 2018-02-22 DJ Delorie <dj@redhat.com>
16267 Sebastian Perta <sebastian.perta@renesas.com>
16268 Oleg Endo <olegendo@gcc.gnu.org>
16270 * config/rx/rx.c (rx_rtx_costs): New function.
16271 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
16273 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
16275 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
16277 2018-02-22 Martin Liska <mliska@suse.cz>
16280 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
16281 Add "native" as a possible value.
16283 2018-02-22 Martin Liska <mliska@suse.cz>
16286 * config/i386/i386.c (ix86_option_override_internal):
16287 Add "native" as a possible value for -march and -mtune.
16289 2018-02-22 Jakub Jelinek <jakub@redhat.com>
16292 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
16293 to all type variants.
16295 PR tree-optimization/84503
16296 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
16297 width as info->bitpos + info->bitsize - start.
16298 (merged_store_group::merge_overlapping): Simplify width computation.
16299 (check_no_overlap): New function.
16300 (imm_store_chain_info::try_coalesce_bswap): Compute expected
16301 start + width and last_order of the group, fail if check_no_overlap
16303 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
16304 to group if check_no_overlap fails.
16306 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
16308 * config/rs6000/altivec.md: Delete contraint arguments to
16309 define_expand, define_split, and define_peephole2, and in
16310 define_insn_and_split if always unused.
16311 * config/rs6000/darwin.md: Ditto.
16312 * config/rs6000/dfp.md: Ditto.
16313 * config/rs6000/rs6000.md: Ditto.
16314 * config/rs6000/sync.md: Ditto.
16315 * config/rs6000/vector.md: Ditto.
16316 * config/rs6000/vsx.md: Ditto.
16318 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
16320 * config/rs6000/altivec.md: Write output control strings as braced
16321 blocks instead of double-quoted strings.
16322 * config/rs6000/darwin.md: Ditto.
16323 * config/rs6000/rs6000.md: Ditto.
16324 * config/rs6000/vector.md: Ditto.
16325 * config/rs6000/vsx.md: Ditto.
16327 2018-02-21 Jason Merrill <jason@redhat.com>
16329 PR c++/84314 - ICE with templates and fastcall attribute.
16330 * attribs.c (build_type_attribute_qual_variant): Remove assert.
16332 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
16334 * ipa-cp.c (determine_versionability): Fix comment typos.
16336 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
16339 * ipa-cp.c (determine_versionability): Do not version functions caling
16342 2018-02-21 Martin Liska <mliska@suse.cz>
16345 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
16346 Add "native" as a possible value.
16347 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
16348 the macro when native cpu detection is available.
16350 2018-02-21 Martin Liska <mliska@suse.cz>
16353 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
16354 Add "native" as a possible value.
16355 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
16356 when native cpu detection is available.
16358 2018-02-21 Jakub Jelinek <jakub@redhat.com>
16359 Martin Sebor <msebor@redhat.com>
16361 PR tree-optimization/84478
16362 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
16364 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
16365 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
16366 support which is conservatively correct, for 2 only stay conservative
16367 for maxlen. Formatting and comment capitalization fixes. Add STRICT
16368 argument to the 2 argument get_range_strlen, adjust 6 arg
16369 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
16371 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
16372 (gimple_fold_builtin_strlen): Pass true as last argument to
16375 2018-02-20 Martin Sebor <msebor@redhat.com>
16377 PR middle-end/84095
16378 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
16379 (builtin_memref::set_base_and_offset): Same. Handle inner references.
16380 (builtin_memref::builtin_memref): Factor out parts into
16381 set_base_and_offset and call it.
16383 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
16385 PR middle-end/84406
16386 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
16387 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
16388 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
16389 search at the associated MODE_INT.
16391 2018-02-20 Jeff Law <law@redhat.com>
16393 PR middle-end/82123
16394 PR tree-optimization/81592
16395 PR middle-end/79257
16396 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
16397 for range data rather than using global data.
16398 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
16399 range data rather than using global data.
16400 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
16401 pass it to children as needed.
16402 (struct directive::fmtresult): Similarly.
16403 (struct directive::set_width): Similarly.
16404 (struct directive::set_precision): Similarly.
16405 (format_integer, format_directive, parse_directive): Similarly.
16406 (format_none): Accept unnamed vr_values parameter.
16407 (format_percent, format_floating, format_character): Similarly.
16408 (format_string, format_plain): Similarly.
16409 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
16410 the EVRP range analyzer for range data rather than using global data.
16411 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
16412 gimple-ssa-evrp-analyze.h
16413 (class sprintf_dom_walker): Add after_dom_children member function.
16414 Add evrp_range_analyzer member.
16415 (sprintf_dom_walker::before_dom_children): Call into the EVRP
16416 range analyzer as needed.
16417 (sprintf_dom_walker::after_dom_children): New member function.
16418 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
16420 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
16421 (evrp_range_analyzer::pop_to_marker): Likewise.
16423 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
16425 PR tree-optimization/84419
16426 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
16427 with the required type if its current type is compatible but
16430 2018-02-20 Jakub Jelinek <jakub@redhat.com>
16432 PR middle-end/82004
16433 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
16434 after vectorization.
16436 2018-02-20 Martin Liska <mliska@suse.cz>
16439 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
16440 possible values if we don't have a hint.
16442 2018-02-20 Martin Liska <mliska@suse.cz>
16446 * final.c (shorten_branches): Build align_tab array with one
16448 * opts.c (finish_options): Add alignment option limit check.
16449 (MAX_CODE_ALIGN): Likewise.
16450 (MAX_CODE_ALIGN_VALUE): Likewise.
16451 * doc/invoke.texi: Document maximum allowed option value for
16452 all -falign-* options.
16454 2018-02-19 Jakub Jelinek <jakub@redhat.com>
16457 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
16458 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
16459 * var-tracking.c (emit_note_insn_var_location): Remove all references
16460 to NOTE_INSN_CALL_ARG_LOCATION.
16461 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
16462 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
16463 Use copy_rtx_if_shared.
16464 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
16465 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
16466 (dwarf2out_var_location): Remove handling of
16467 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
16469 * final.c (final_scan_insn): Remove all references to
16470 NOTE_INSN_CALL_ARG_LOCATION.
16471 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
16472 before dumping final insns.
16473 * except.c (emit_note_eh_region_end): Remove all references to
16474 NOTE_INSN_CALL_ARG_LOCATION.
16475 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
16476 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
16477 * config/arc/arc.c (hwloop_optimize): Likewise.
16478 * config/arm/arm.c (create_fix_barrier): Likewise.
16479 * config/s390/s390.c (s390_chunkify_start): Likewise.
16480 * config/sh/sh.c (find_barrier): Likewise.
16481 * config/i386/i386.c (rest_of_insert_endbranch,
16482 ix86_seh_fixup_eh_fallthru): Likewise.
16483 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
16484 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
16485 * config/frv/frv.c (frv_function_prologue): Likewise.
16486 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
16488 (note_outside_basic_block_p): Remove all references to
16489 NOTE_INSN_CALL_ARG_LOCATION.
16490 * gengtype.c (adjust_field_rtx_def): Likewise.
16491 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
16493 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
16494 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
16497 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
16500 PR tree-optimization/84452
16501 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
16502 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
16505 2018-02-19 Martin Liska <mliska@suse.cz>
16508 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
16510 2018-02-19 Martin Liska <mliska@suse.cz>
16511 Richard Sandiford <richard.sandiford@linaro.org>
16513 PR tree-optimization/82491
16514 * gimple-fold.c (get_base_constructor): Make earlier bail out
16517 2018-02-19 Carl Love <cel@us.ibm.com>
16519 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
16520 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
16522 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
16523 P8V_BUILTIN_VEC_NEG.
16525 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
16527 * config/rl78/rl78.md (movdf): New define expand.
16529 2018-02-19 Martin Liska <mliska@suse.cz>
16532 * doc/invoke.texi: Fix typo.
16533 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
16535 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
16537 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
16538 handle rs6000_single_float and rs6000_double_float specially for
16541 2018-02-16 Jeff Law <law@redhat.com>
16543 * config/rx/rx.c (add_pop_cfi_notes): New function.;
16544 (pop_regs): Use it.
16546 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16549 * ipa-inline.c (inline_small_functions): Fix a typo.
16551 2018-02-16 Nathan Sidwell <nathan@acm.org>
16553 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
16555 2018-02-16 Carl Love <cel@us.ibm.com>
16557 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
16558 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
16559 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
16560 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
16561 expansion to P8V_BUILTIN_VEC_FLOAT2.
16563 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
16565 PR rtl-optimization/70023
16566 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
16567 src_regno into account.
16569 2018-02-16 Carl Love <cel@us.ibm.com>
16571 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
16572 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
16573 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
16574 * config/rs6000/rs6000.c: Remove case statements for
16575 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
16576 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
16577 and P9V_BUILTIN_VEC_VINSERT4B.
16578 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
16579 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
16580 * config/rs6000/vsx.md:
16581 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
16584 2018-02-16 Carl Love <cel@us.ibm.com>
16586 * config/rs6000/altivec.h: Add builtin names vec_extract4b
16588 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
16590 * config/rs6000/rs6000-c.c: Add the definitions for
16591 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
16592 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
16593 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
16594 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
16595 definition for insert4b and define insn *insert3b_internal.
16596 * doc/extend.texi: Add documentation for vec_extract4b.
16598 2018-02-16 Nathan Sidwell <nathan@acm.org>
16600 * doc/extend.texi (Backwards Compatibility): Mention friend
16601 injection. Note for-scope is deprecated.
16602 * doc/invoke.texi (-ffriend-injection): Deprecate.
16604 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
16606 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
16607 that moved to I2, also allow destinations that are a paradoxical
16608 subreg (instead of a normal reg).
16610 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
16613 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
16616 2018-02-16 Richard Biener <rguenther@suse.de>
16618 PR tree-optimization/84037
16619 PR tree-optimization/84016
16621 * config/i386/i386.c (ix86_builtin_vectorization_cost):
16622 Adjust vec_construct for the fact we need additional higher latency
16623 128bit inserts for AVX256 and AVX512 vector builds.
16624 (ix86_add_stmt_cost): Scale vector construction cost for
16627 2018-02-16 Richard Biener <rguenther@suse.de>
16629 PR tree-optimization/84417
16630 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
16631 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
16632 (non_rewritable_lvalue_p): Likewise, use poly-ints.
16634 2018-02-16 Martin Liska <mliska@suse.cz>
16637 * internal-fn.def (ASAN_CHECK): Set proper flags.
16638 (ASAN_MARK): Likewise.
16640 2018-02-16 Julia Koval <julia.koval@intel.com>
16642 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
16643 from PTA_CANNONLAKE.
16645 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16648 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
16649 Use ++iter rather than iter++ for std::list iterators.
16650 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
16651 defer deleting them until all nodes in the forest are processed. Do
16652 free even leaf nodes. Change to_process into auto_vec.
16655 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
16656 * vec.h (vec_default_construct): Use memset instead of placement new
16657 if BROKEN_VALUE_INITIALIZATION is defined.
16658 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
16659 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
16662 PR rtl-optimization/83723
16663 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
16664 * lra.c (lra_substitute_pseudo): Likewise. If true, use
16665 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
16667 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
16669 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
16671 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
16673 PR rtl-optimization/81443
16674 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
16675 from inner REGs to paradoxical SUBREGs.
16677 2018-02-16 Richard Biener <rguenther@suse.de>
16679 PR tree-optimization/84399
16680 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
16681 For operands we can analyze at their definition make sure we can
16682 analyze them at each use as well.
16684 2018-02-16 Richard Biener <rguenther@suse.de>
16686 PR tree-optimization/84190
16687 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
16688 volatile accesses if the decl isn't volatile.
16690 2018-02-15 Jason Merrill <jason@redhat.com>
16692 PR c++/84314 - ICE with templates and fastcall attribute.
16693 * attribs.c (build_type_attribute_qual_variant): Don't clobber
16694 TYPE_CANONICAL on an existing type.
16696 2018-02-15 Jakub Jelinek <jakub@redhat.com>
16698 PR tree-optimization/84383
16699 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
16700 dstoff nor call operand_equal_p if dstbase is NULL.
16702 PR tree-optimization/84334
16703 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
16704 also a CONSTANT_CLASS_P, punt.
16706 2018-02-14 Jim Wilson <jimw@sifive.com>
16708 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
16709 first SMALL_OPERAND check. New local min_second_step. Move assert
16710 to where locals are set. Add TARGET_RVC support.
16711 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
16713 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
16715 * doc/invoke.texi: Correct -Wformat-overflow code sample.
16717 2018-02-14 Martin Sebor <msebor@redhat.com>
16719 PR tree-optimization/83698
16720 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
16721 arrays constrain the offset range to their bounds.
16722 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
16723 (builtin_access::overlap): Avoid setting the size of overlap if it's
16725 (maybe_diag_overlap): Also consider arrays when deciding what values
16726 of offsets to include in diagnostics.
16728 2018-02-14 Martin Sebor <msebor@redhat.com>
16731 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
16732 that correspond to the kind of a declaration.
16734 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
16737 * config/pa/pa.md: Load address of PIC label using the linkage table
16738 if the label is nonlocal.
16740 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
16742 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
16743 warning message if user requests -maltivec=be.
16744 * doc/invoke.texi: Document deprecation of -maltivec=be.
16746 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
16749 * config/rs6000/rs6000-c.c: Update definitions for
16750 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
16751 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
16753 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16756 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
16757 add _get_ssp intrinsics. Remove argument from
16758 __builtin_ia32_rdssp[d|q].
16759 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
16760 * config/i386/i386-builtin.def: Remove argument from
16761 __builtin_ia32_rdssp[d|q].
16762 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
16763 ix86_expand_special_args_builtin for _rdssp[d|q].
16764 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
16765 Clear register before usage.
16766 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
16767 Add documentation for new _get_ssp and _inc_ssp intrinsics.
16769 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
16771 PR tree-optimization/84357
16772 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
16773 operand 1 of an ARRAY_REF too.
16775 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
16778 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
16779 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16781 (set_of_reg): New struct.
16782 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
16783 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
16784 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16786 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
16787 Split into bitclr, bitset, bitinvert patterns if appropriate.
16788 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
16789 use rx_fuse_in_memory_bitop.
16790 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
16791 to named insn, correct maximum insn length.
16793 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
16796 * machmode.def: Define a complex mode for PARTIAL_INT.
16797 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
16799 * doc/rtl.texi: Document CSPImode.
16800 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
16802 (msp430_hard_regno_nregs_with_padding): Likewise.
16804 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
16807 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
16809 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
16811 PR rtl-optimization/84169
16812 * combine.c (try_combine): New variable split_i2i3. Set it to true if
16813 we generated a parallel as new i3 and we split that to new i2 and i3
16814 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
16815 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
16816 those to i2, not i1. Partially rewrite this scan code.
16818 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16821 * stor-layout.c (place_field): For variable length fields, adjust
16822 offset_align afterwards not just based on the field's alignment,
16823 but also on the size.
16825 PR middle-end/84309
16826 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
16827 of exps and logs in the use_exp2 case.
16829 2018-02-13 Jeff Law <law@redhat.com>
16831 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
16832 entry for "vector".
16834 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
16837 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
16841 * common.opt (gas-loc-support, gas-locview-support): New.
16842 (ginline-points, ginternal-reset-location-views): New.
16843 * doc/invoke.texi: Document them. Use @itemx where intended.
16844 (gvariable-location-views): Adjust.
16845 * target.def (reset_location_view): New.
16846 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
16847 (TARGET_RESET_LOCATION_VIEW): New.
16848 * doc/tm.texi: Rebuilt.
16849 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
16850 (dwarf2out_default_as_locview_support): New.
16851 (output_asm_line_debug_info): Use option variables.
16852 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
16853 (output_loc_list): Likewise.
16854 (add_high_low_attributes): Check option variables.
16855 Don't output entry view attribute in strict mode.
16856 (gen_inlined_subroutine_die): Check option variables.
16857 (dwarf2out_inline_entry): Likewise.
16858 (init_sections_and_labels): Likewise.
16859 (dwarf2out_early_finish): Likewise.
16860 (maybe_reset_location_view): New, from...
16861 (dwarf2out_var_location): ... here. Call it.
16862 * debug.h (dwarf2out_default_as_loc_support): Declare.
16863 (dwarf2out_default_as_locview_support): Declare.
16864 * hooks.c (hook_int_rtx_insn_0): New.
16865 * hooks.h (hook_int_rtx_insn_0): Declare.
16866 * toplev.c (process_options): Take -gas-loc-support and
16867 -gas-locview-support from dwarf2out. Enable
16868 -gvariable-location-views by default only with locview
16869 assembler support. Enable -ginternal-reset-location-views by
16870 default only if the target defines the corresponding hook.
16871 Enable -ginline-points by default if location views are
16872 enabled; force it disabled if statement frontiers are
16874 * tree-inline.c (expand_call_inline): Check option variables.
16875 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
16877 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
16879 PR tree-optimization/84321
16880 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
16881 handling. Also check whether the anti-range contains any values
16882 that satisfy the mask; switch to a VR_RANGE if not.
16884 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
16887 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
16889 2018-02-13 Martin Jambor <mjambor@suse.cz>
16892 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
16893 of call statements, also set location of a load to a temporary.
16895 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
16897 * config/rl78/rl78.c (add_vector_labels): New function.
16898 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
16899 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
16900 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
16901 which checks that no arguments are passed.
16902 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
16903 * doc/extend.texi: Documentation for the new attribute.
16905 2018-02-13 Andreas Schwab <schwab@suse.de>
16907 * config/riscv/linux.h (CPP_SPEC): Define.
16909 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16912 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
16913 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
16914 OPTION_MASK_ISA_AES as first argument to def_builtin_const
16915 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
16916 instead of OPTION_MASK_ISA_PCLMUL as first argument to
16917 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
16918 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
16919 temporarily for AES and PCLMUL builtins.
16921 PR tree-optimization/84339
16922 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
16923 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
16926 PR middle-end/84309
16927 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
16928 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
16929 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
16931 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
16933 * omp-simd-clone.h: New file.
16934 * omp-simd-clone.c: Include omp-simd-clone.h.
16935 (expand_simd_clones): No longer static.
16936 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
16937 cgraph.h and omp-simd-clone.h.
16938 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
16939 (vect_recog_widen_shift_pattern): Formatting fix.
16940 (vect_pattern_recog_1): Don't check optab for calls.
16943 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
16944 operands[2] into a REG before using gen_lowpart on it.
16946 2018-02-12 Jeff Law <law@redhat.com>
16949 * config/sh/sh.c (find_barrier): Consider a sibling call
16952 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
16953 successfully back substituting a reg.
16955 2018-02-12 Richard Biener <rguenther@suse.de>
16957 PR tree-optimization/84037
16958 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
16959 parameter, move visited init to caller.
16960 (vect_slp_analyze_operations): Separate cost from validity
16961 check, initialize visited once for all instances.
16962 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
16964 * tree-vect-stmts.c (vect_model_simple_cost): Make early
16966 (vect_model_promotion_demotion_cost): Likewise.
16967 (vectorizable_bswap): Guard cost modeling with !slp_node
16968 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
16970 (vectorizable_call): Likewise.
16971 (vectorizable_conversion): Likewise.
16972 (vectorizable_assignment): Likewise.
16973 (vectorizable_shift): Likewise.
16974 (vectorizable_operation): Likewise.
16975 (vectorizable_store): Likewise.
16976 (vectorizable_load): Likewise.
16977 (vectorizable_condition): Likewise.
16978 (vectorizable_comparison): Likewise.
16980 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
16983 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
16984 (ASAN_MARK): Fix fnspec to account for return value, change pointer
16985 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
16987 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16989 PR middle-end/83665
16990 * params.def (inline-min-speedup): Increase from 8 to 15.
16991 (max-inline-insns-auto): Decrease from 40 to 30.
16992 * ipa-split.c (consider_split): Add some buffer for function to
16993 be considered inlining candidate.
16994 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
16997 2018-02-12 Richard Biener <rguenther@suse.de>
16999 PR tree-optimization/84037
17000 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
17001 matched stmts if we cannot swap the non-matched ones.
17003 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
17005 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
17006 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
17007 _mm_maskz_scalef_round_ss): New intrinsics.
17008 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
17009 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
17010 __builtin_ia32_scalefss_round): Remove.
17011 (__builtin_ia32_scalefsd_mask_round,
17012 __builtin_ia32_scalefss_mask_round): New intrinsics.
17013 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
17014 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
17015 ((match_operand:VF_128 2 "<round_nimm_predicate>"
17016 "<round_constraint>")): Changed to ...
17017 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
17018 "<round_scalar_constraint>")): ... this.
17019 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
17020 %0, %1, %2<round_op3>}"): Changed to ...
17021 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
17022 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
17023 %2<round_scalar_mask_op3>}"): ... this.
17024 * config/i386/subst.md (round_scalar_nimm_predicate): New.
17026 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
17028 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
17029 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
17030 (_mm_maskz_sqrt_round_ss): New intrinsics.
17031 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
17032 (__builtin_ia32_sqrtsd_mask_round)
17033 (__builtin_ia32_sqrtss_mask_round): New builtins.
17034 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
17035 (__builtin_ia32_sqrtss_round): Remove.
17036 (__builtin_ia32_sqrtsd_mask_round)
17037 (__builtin_ia32_sqrtss_mask_round): New builtins.
17038 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
17039 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
17040 ((match_operand:VF_128 1 "vector_operand"
17041 "xBm,<round_constraint>")): Changed to ...
17042 ((match_operand:VF_128 1 "vector_operand"
17043 "xBm,<round_scalar_constraint>")): ... this.
17044 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
17045 %0, %2, %<iptr>1<round_op3>}): Changed to ...
17046 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
17047 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
17048 %<iptr>1<round_scalar_mask_op3>}): ... this.
17049 ((set_attr "prefix" "<round_prefix>")): Changed to ...
17050 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
17052 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
17055 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
17056 Cast vec_cmpeq result to correct type.
17057 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
17058 Cast vec_cmpgt result to correct type.
17060 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
17062 * final.c (final_scan_insn_1): Renamed from...
17063 (final_scan_insn): ... this. New wrapper, to recover
17064 seen from the outermost call in recursive ones.
17065 * config/sparc/sparc.c (output_return): Drop seen from call.
17066 (output_sibcall): Likewise.
17067 * config/visium/visium.c (output_branch): Likewise.
17069 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
17071 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
17074 2018-02-10 Alan Modra <amodra@gmail.com>
17077 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
17078 Specify LR as an input.
17080 2018-02-10 Jakub Jelinek <jakub@redhat.com>
17083 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
17084 remove_member_access_dummy_vars): New functions.
17085 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
17086 lower_omp_1, execute_lower_omp): Use them.
17088 PR rtl-optimization/84308
17089 * shrink-wrap.c (spread_components): Release todo vector.
17091 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
17093 PR rtl-optimization/57193
17094 * ira-color.c (struct allocno_color_data): Add member
17095 conflict_allocno_hard_prefs.
17096 (update_conflict_allocno_hard_prefs): New.
17097 (bucket_allocno_compare_func): Add a preference based on
17098 conflict_allocno_hard_prefs.
17099 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
17100 (color_allocnos): Remove a dead code. Initiate
17101 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
17103 2018-02-09 Jakub Jelinek <jakub@redhat.com>
17106 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
17107 constraint from =wa to wa. Avoid a subreg on the output operand,
17108 instead use a pseudo and subreg it in a move.
17109 (p9_xxbrd_<mode>): Changed to ...
17110 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
17111 (p9_xxbrd_v2df): New expander.
17112 (p9_xxbrw_<mode>): Changed to ...
17113 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
17114 (p9_xxbrw_v4sf): New expander.
17116 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
17118 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
17120 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
17123 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
17124 multiply in 32-bit mode.
17125 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
17126 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
17129 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
17131 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
17132 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
17133 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
17134 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
17136 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
17139 * dwarf2out.c (is_trivial_indirect_ref): New function.
17140 (dwarf2out_late_global_decl): Do not generate a location
17141 attribute for variables that have a non-trivial DECL_VALUE_EXPR
17142 and that are not defined in the current unit.
17144 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
17146 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
17147 instead of a libcall for UNORDERED.
17149 2018-02-09 Tamar Christina <tamar.christina@arm.com>
17152 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
17153 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
17155 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17158 * config/s390/s390.c (s390_set_current_function): Invoke
17159 s390_indirect_branch_settings also if fndecl didn't change.
17161 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
17163 * config/rs6000/rs6000.md (blockage): Set length to zero.
17165 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
17167 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
17169 2018-02-09 Jakub Jelinek <jakub@redhat.com>
17172 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
17173 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
17177 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
17178 PARALLEL incoming that failed vt_get_decl_and_offset check.
17180 PR middle-end/84237
17181 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
17182 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
17184 (get_variable_section): For decls in named .bss* sections pass true as
17185 second argument to bss_initializer_p.
17187 2018-02-09 Marek Polacek <polacek@redhat.com>
17188 Jakub Jelinek <jakub@redhat.com>
17191 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
17192 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
17193 Sync some changes from cxx_fold_indirect_ref.
17195 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
17197 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
17199 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
17200 (BLOCK_INLINE_ENTRY_LABEL): New.
17201 (dwarf2out_var_location): Disregard inline entry markers.
17202 (inline_entry_data): New struct.
17203 (inline_entry_data_hasher): New hashtable type.
17204 (inline_entry_data_hasher::hash): New.
17205 (inline_entry_data_hasher::equal): New.
17206 (inline_entry_data_table): New variable.
17207 (add_high_low_attributes): Add DW_AT_entry_pc and
17208 DW_AT_GNU_entry_view attributes if a pending entry is found
17209 in inline_entry_data_table. Add old entry_pc attribute only
17210 if debug nonbinding markers are disabled.
17211 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
17212 markers are enabled.
17213 (block_within_block_p, dwarf2out_inline_entry): New.
17214 (dwarf2out_finish): Check that no entries remained in
17215 inline_entry_data_table.
17216 * final.c (reemit_insn_block_notes): Handle inline entry notes.
17217 (final_scan_insn, notice_source_line): Likewise.
17218 (rest_of_clean_state): Skip inline entry markers.
17219 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
17221 * gimple.c (gimple_build_debug_inline_entry): New.
17222 * gimple.h (enum gimple_debug_subcode): Add
17223 GIMPLE_DEBUG_INLINE_ENTRY.
17224 (gimple_build_debug_inline_entry): Declare.
17225 (gimple_debug_inline_entry_p): New.
17226 (gimple_debug_nonbind_marker_p): Adjust.
17227 * insn-notes.def (INLINE_ENTRY): New.
17228 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
17229 inline entry marker notes.
17230 (print_insn): Likewise.
17231 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
17232 (INSN_DEBUG_MARKER_KIND): Likewise.
17233 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
17234 * tree-inline.c (expand_call_inline): Build and insert
17235 debug_inline_entry stmt.
17236 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
17237 inline entry blocks early, if nonbind markers are enabled.
17238 (dump_scope_block): Dump fragment info.
17239 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
17240 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
17241 (gimple_build_debug_inline_entry): New.
17242 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
17243 Enable/disable inline entry points too.
17244 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
17245 (DEBUG_INSN): Describe inline entry markers.
17247 * common.opt (gvariable-location-views): New.
17248 (gvariable-location-views=incompat5): New.
17249 * config.in: Rebuilt.
17250 * configure: Rebuilt.
17251 * configure.ac: Test assembler for view support.
17252 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
17253 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
17254 * dwarf2out.c (var_loc_view): New typedef.
17255 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
17256 (dwarf2out_locviews_in_attribute): New.
17257 (dwarf2out_locviews_in_loclist): New.
17258 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
17259 (enum dw_line_info_opcode): Add LI_adv_address.
17260 (struct dw_line_info_table): Add view.
17261 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
17262 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
17263 (zero_view_p): New variable.
17264 (ZERO_VIEW_P): New macro.
17265 (output_asm_line_debug_info): New.
17266 (struct var_loc_node): Add view.
17267 (add_AT_view_list, AT_loc_list): New.
17268 (add_var_loc_to_decl): Add view param. Test it against last.
17269 (new_loc_list): Add view params. Record them.
17270 (AT_loc_list_ptr): Handle loc and view lists.
17271 (view_list_to_loc_list_val_node): New.
17272 (print_dw_val): Handle dw_val_class_view_list.
17273 (size_of_die): Likewise.
17274 (value_format): Likewise.
17275 (loc_list_has_views): New.
17276 (gen_llsym): Set vl_symbol too.
17277 (maybe_gen_llsym, skip_loc_list_entry): New.
17278 (dwarf2out_maybe_output_loclist_view_pair): New.
17279 (output_loc_list): Output view list or entries too.
17280 (output_view_list_offset): New.
17281 (output_die): Handle dw_val_class_view_list.
17282 (output_dwarf_version): New.
17283 (output_compilation_unit_header): Use it.
17284 (output_skeleton_debug_sections): Likewise.
17285 (output_rnglists, output_line_info): Likewise.
17286 (output_pubnames, output_aranges): Update version comments.
17287 (output_one_line_info_table): Output view numbers in asm comments.
17288 (dw_loc_list): Determine current endview, pass it to new_loc_list.
17289 Call maybe_gen_llsym.
17290 (loc_list_from_tree_1): Adjust.
17291 (add_AT_location_description): Create view list attribute if
17292 needed, check it's absent otherwise.
17293 (convert_cfa_to_fb_loc_list): Adjust.
17294 (maybe_emit_file): Call output_asm_line_debug_info for test.
17295 (dwarf2out_var_location): Reset views as needed. Precompute
17296 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
17297 attribute. Set view.
17298 (new_line_info_table): Reset next view.
17299 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
17300 (dwarf2out_source_line): Likewise. Output view resets and labels to
17301 the assembler, or select appropriate line info opcodes.
17302 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
17303 (optimize_string_length): Catch it. Adjust.
17304 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
17305 dw_val_class_view_list, and remove it if no longer needed.
17306 (hash_loc_list): Hash view numbers.
17307 (loc_list_hasher::equal): Compare them.
17308 (optimize_location_lists): Check whether a view list symbol is
17309 needed, and whether the locview attribute is present, and
17310 whether they match. Remove the locview attribute if no longer
17312 (index_location_lists): Call skip_loc_list_entry for test.
17313 (dwarf2out_finish): Call output_asm_line_debug_info for test.
17314 Use output_dwarf_version.
17315 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
17316 (struct dw_val_node): Add val_view_list.
17317 * final.c (SEEN_NEXT_VIEW): New.
17318 (set_next_view_needed): New.
17319 (clear_next_view_needed): New.
17320 (maybe_output_next_view): New.
17321 (final_start_function): Rename to...
17322 (final_start_function_1): ... this. Take pointer to FIRST,
17323 add SEEN parameter. Emit param bindings in the initial view.
17324 (final_start_function): Reintroduce SEEN-less interface.
17325 (final): Rename to...
17326 (final_1): ... this. Take SEEN parameter. Output final pending
17327 next view at the end.
17328 (final): Reintroduce seen-less interface.
17329 (final_scan_insn): Output pending next view before switching
17330 sections or ending a block. Mark the next view as needed when
17331 outputting variable locations. Notify debug backend of section
17332 changes, and of location view changes.
17333 (rest_of_handle_final): Adjust.
17334 * toplev.c (process_options): Autodetect value for debug variable
17335 location views option. Warn on incompat5 without -gdwarf-5.
17336 * doc/invoke.texi (gvariable-location-views): New.
17337 (gvariable-location-views=incompat5): New.
17338 (gno-variable-location-views): New.
17340 2018-02-08 David Malcolm <dmalcolm@redhat.com>
17342 PR tree-optimization/84136
17343 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
17344 that the result of find_edge is non-NULL.
17346 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
17349 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
17350 storing integer register in SImode. Fix cost of 256 and 512
17351 byte aligned SSE register store.
17353 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
17355 * config/i386/i386.c (ix86_multiplication_cost): Fix
17356 multiplication cost for TARGET_AVX512DQ.
17358 2018-02-08 Marek Polacek <polacek@redhat.com>
17360 PR tree-optimization/84238
17361 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
17364 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17366 PR tree-optimization/84265
17367 * tree-vect-stmts.c (vectorizable_store): Don't treat
17368 VMAT_CONTIGUOUS accesses as grouped.
17369 (vectorizable_load): Likewise.
17371 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17373 PR tree-optimization/81635
17374 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
17375 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
17376 (test_round_for_mask): New functions.
17377 (wide_int_cc_tests): Call test_round_for_mask.
17378 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
17379 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
17380 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
17381 range returned by get_range_info.
17383 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
17386 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
17387 * symtab.c: Include builtins.h
17388 (symtab_node::output_to_lto_symbol_table_p): Move here
17389 from lto-streamer-out.c:output_symbol_p.
17390 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
17391 (output_symbol_p): Move all logic to symtab.c
17392 (produce_symtab): Update.
17394 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17396 * config/s390/s390-opts.h (enum indirect_branch): Define.
17397 * config/s390/s390-protos.h (s390_return_addr_from_memory)
17398 (s390_indirect_branch_via_thunk)
17399 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
17400 (enum s390_indirect_branch_type): Define.
17401 * config/s390/s390.c (struct s390_frame_layout, struct
17402 machine_function): Remove.
17403 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
17404 (indirect_branch_table_label_no, indirect_branch_table_name):
17406 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
17407 (enum s390_indirect_branch_option): Define.
17408 (s390_return_addr_from_memory): New function.
17409 (s390_handle_string_attribute): New function.
17410 (s390_attribute_table): Add new attribute handler.
17411 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
17412 (s390_indirect_branch_via_thunk): New function.
17413 (s390_indirect_branch_via_inline_thunk): New function.
17414 (s390_function_ok_for_sibcall): When jumping via thunk disallow
17415 sibling call optimization for non z10 compiles.
17416 (s390_emit_call): Force indirect branch target to be a single
17417 register. Add r1 clobber for non-z10 compiles.
17418 (s390_emit_epilogue): Emit return jump via return_use expander.
17419 (s390_reorg): Handle JUMP_INSNs as execute targets.
17420 (s390_option_override_internal): Perform validity checks for the
17421 new command line options.
17422 (s390_indirect_branch_attrvalue): New function.
17423 (s390_indirect_branch_settings): New function.
17424 (s390_set_current_function): Invoke s390_indirect_branch_settings.
17425 (s390_output_indirect_thunk_function): New function.
17426 (s390_code_end): Implement target hook.
17427 (s390_case_values_threshold): Implement target hook.
17428 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
17430 * config/s390/s390.h (struct s390_frame_layout)
17431 (struct machine_function): Move here from s390.c.
17432 (TARGET_INDIRECT_BRANCH_NOBP_RET)
17433 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
17434 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
17435 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
17436 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
17437 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
17438 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
17439 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
17440 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
17441 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
17442 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
17443 (mnemonic attribute): Add values which aren't recognized
17445 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
17446 pattern for branch conversion. Fix mnemonic attribute.
17447 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
17448 indirect branch via thunk if requested.
17449 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
17450 ("*indirect_jump"): Disable for branch conversion using out of
17452 ("indirect_jump_via_thunk<mode>_z10")
17453 ("indirect_jump_via_thunk<mode>")
17454 ("indirect_jump_via_inlinethunk<mode>_z10")
17455 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
17456 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
17457 ("casesi_jump_via_inlinethunk<mode>_z10")
17458 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
17459 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
17460 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
17461 ("*indirect2_jump"): Disable for branch conversion.
17462 ("casesi_jump"): Turn into expander and expand patterns for branch
17464 ("return_use"): New expander.
17465 ("*return"): Emit return via thunk and rename it to ...
17466 ("*return<mode>"): ... this one.
17467 * config/s390/s390.opt: Add new options and and enum for the
17470 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17472 * lra-constraints.c (match_reload): Unconditionally use
17473 gen_lowpart_SUBREG, rather than selecting between that
17474 and equivalent gen_rtx_SUBREG code.
17476 2018-02-08 Richard Biener <rguenther@suse.de>
17478 PR tree-optimization/84233
17479 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
17480 changed flag instead of boguously re-using phi_inserted.
17482 2018-02-08 Martin Jambor <mjambor@suse.cz>
17484 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
17485 static local variables.
17487 2018-02-08 Richard Biener <rguenther@suse.de>
17489 PR tree-optimization/84278
17490 * tree-vect-stmts.c (vectorizable_store): When looking for
17491 smaller vector types to perform grouped strided loads/stores
17492 make sure the mode is supported by the target.
17493 (vectorizable_load): Likewise.
17495 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
17497 * config/aarch64/aarch64.c (aarch64_components_for_bb):
17498 Increase LDP/STP opportunities by adding adjacent callee-saves.
17500 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
17502 PR rtl-optimization/84068
17503 PR rtl-optimization/83459
17504 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
17506 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
17508 PR tree-optimization/84224
17509 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
17510 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
17511 non-zero arguments.
17513 2018-02-07 Iain Sandoe <iain@codesourcery.com>
17516 * config/rs6000/altivec.md (*restore_world): Remove LR use.
17517 * config/rs6000/predicates.md (restore_world_operation): Adjust op
17518 count, remove one USE.
17520 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
17522 * doc/install.texi (Configuration): Document the
17523 --with-long-double-format={ibm,ieee} PowerPC configuration
17527 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
17528 Convert from define_expand to be define_insn_and_split. Rework
17529 float/double/_Float128 conversions to QI/HI/SImode to work with
17530 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
17531 conversions to QI/HImode types did a store and then a load to
17532 truncate the value. For conversions to VSX registers, don't split
17533 the insn, instead emit the code directly. Use the code iterator
17534 any_fix to combine signed and unsigned conversions.
17535 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
17536 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
17537 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17538 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
17539 (fix_<mode>di2_hw): Likewise.
17540 (fixuns_<mode>di2_hw): Likewise.
17541 (fix_<mode>si2_hw): Likewise.
17542 (fixuns_<mode>si2_hw): Likewise.
17543 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
17544 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17545 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
17546 fix<uns>_trunc<SFDF:mode>si2_p8.
17547 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
17548 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
17549 (fix<uns>_<mode>_mem): Likewise.
17550 (fctiw<u>z_<mode>_mem): Likewise.
17551 (fix<uns>_<mode>_mem): Likewise.
17552 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
17553 the register allocator from doing a direct move to the GPRs to do
17554 a store, and instead use the ISA 3.0 store byte/half-word from
17555 vector register instruction. For IEEE 128-bit floating point,
17556 also optimize stores of 32-bit ints.
17557 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
17559 2018-02-07 Alan Hayward <alan.hayward@arm.com>
17561 * genextract.c (push_pathstr_operand): New function to support
17563 (walk_rtx): Call push_pathstr_operand.
17564 (print_path): Support [a-zA-Z].
17566 2018-02-07 Richard Biener <rguenther@suse.de>
17568 PR tree-optimization/84037
17569 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
17570 (cse_and_gimplify_to_preheader): Declare.
17571 (vect_get_place_in_interleaving_chain): Likewise.
17572 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17574 (_loop_vec_info::~_loop_vec_info): Delete it.
17575 (cse_and_gimplify_to_preheader): New function.
17576 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
17577 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
17578 (vectorizable_load): Likewise. For grouped stores always base
17579 the IV on the first element.
17580 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
17581 condition before gimplifying.
17583 2018-02-07 Jakub Jelinek <jakub@redhat.com>
17585 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
17586 *DIV_EXPR and *MOD_EXPR.
17588 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
17591 * config/i386/i386.c (ix86_option_override_internal): Mask out
17592 the CF_SET bit when checking -fcf-protection.
17594 2018-02-07 Tom de Vries <tom@codesourcery.com>
17597 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
17600 2018-02-07 Richard Biener <rguenther@suse.de>
17602 PR tree-optimization/84204
17603 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
17606 PR tree-optimization/84205
17607 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
17608 special-case isl_ast_op_zdiv_r.
17610 PR tree-optimization/84223
17611 * graphite-scop-detection.c (gather_bbs::before_dom_children):
17612 Only add conditions from within the region.
17613 (gather_bbs::after_dom_children): Adjust.
17615 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
17618 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
17619 * config/avr/avr.md: Only post-reload split REG-REG moves if
17620 either register is GENERAL_REG_P.
17622 2018-02-07 Jakub Jelinek <jakub@redhat.com>
17624 PR tree-optimization/84235
17625 * tree-ssa-scopedtables.c
17626 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
17627 if the subtraction is performed in floating point type where NaNs are
17628 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
17629 build 1. Formatting fix.
17631 2018-02-06 Jakub Jelinek <jakub@redhat.com>
17634 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
17635 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
17636 and skip it regardless of bb boundaries. Use CALL_P macro,
17637 don't test INSN_P (insn) together with CALL_P or JUMP_P check
17638 unnecessarily, formatting fix.
17640 2018-02-06 Michael Collison <michael.collison@arm.com>
17642 * config/arm/thumb2.md:
17643 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
17644 (*thumb_mov_notscc): Ditto.
17646 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
17649 * config/rs6000/rs6000.md (su code attribute): Use "u" for
17650 unsigned_fix, not "s".
17652 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17654 * configure.ac (gcc_fn_eh_frame_ro): New function.
17655 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
17656 correct .eh_frame permissions.
17657 * configure: Regenerate.
17659 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
17661 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
17662 irrelevant options.
17664 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17666 * config/rs6000/rs6000.c (rs6000_option_override_internal):
17667 Display warning message for -mno-speculate-indirect-jumps.
17669 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
17671 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
17673 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
17675 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
17677 PR tree-optimization/84225
17678 * tree-eh.c (find_trapping_overflow): Only call
17679 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
17681 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17684 * config/i386/i386.c: Reimplement the check of possible options
17685 -mibt/-mshstk conbination. Change error messages.
17686 * doc/invoke.texi: Fix a typo: remove extra '='.
17688 2018-02-06 Marek Polacek <polacek@redhat.com>
17690 PR tree-optimization/84228
17691 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
17693 2018-02-06 Tamar Christina <tamar.christina@arm.com>
17696 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
17697 emitted arch directives.
17698 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
17699 __ARM_FEATURE_COPROC before changing architectures.
17701 2018-02-06 Richard Biener <rguenther@suse.de>
17703 * config/i386/i386.c (print_reg): Fix typo.
17704 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
17706 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
17708 * configure: Regenerate.
17710 2018-02-05 Martin Sebor <msebor@redhat.com>
17712 PR tree-optimization/83369
17713 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
17716 2018-02-05 Martin Liska <mliska@suse.cz>
17718 * doc/invoke.texi: Cherry-pick upstream r323995.
17720 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
17722 * ira.c (ira_init_register_move_cost): Adjust comment.
17724 2018-02-05 Martin Liska <mliska@suse.cz>
17726 PR gcov-profile/84137
17727 * doc/gcov.texi: Fix typo in documentation.
17729 2018-02-05 Martin Liska <mliska@suse.cz>
17731 PR gcov-profile/83879
17732 * doc/gcov.texi: Document necessity of --dynamic-list-data when
17733 using dlopen functionality.
17735 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
17737 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
17738 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
17739 _mm_maskz_range_ss, _mm_mask_range_round_ss,
17740 _mm_maskz_range_round_ss): New intrinsics.
17741 (__builtin_ia32_rangesd128_round)
17742 (__builtin_ia32_rangess128_round): Remove.
17743 (__builtin_ia32_rangesd128_mask_round,
17744 __builtin_ia32_rangess128_mask_round): New builtins.
17745 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
17746 __builtin_ia32_rangess128_round): Remove.
17747 (__builtin_ia32_rangesd128_mask_round,
17748 __builtin_ia32_rangess128_mask_round): New builtins.
17749 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
17750 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
17751 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
17752 "<round_saeonly_constraint>")): Changed to ...
17753 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
17754 "<round_saeonly_scalar_constraint>")): ... this.
17755 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
17756 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
17757 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
17758 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
17759 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
17761 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
17763 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
17765 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
17766 Remove all values except native, 8540 and 8548.
17768 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
17770 * config/i386/i386.c (ix86_output_function_return): Pass
17771 INVALID_REGNUM, instead of -1, as invalid register number to
17772 indirect_thunk_name and output_indirect_thunk.
17774 2018-02-02 Julia Koval <julia.koval@intel.com>
17776 * config.gcc: Add -march=icelake.
17777 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
17778 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
17779 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
17780 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
17781 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
17782 (processor_target_table): Add icelake.
17783 (ix86_option_override_internal): Handle new PTAs.
17784 (get_builtin_code_for_version): Handle icelake.
17785 (M_INTEL_COREI7_ICELAKE): New.
17786 (fold_builtin_cpu): Handle icelake.
17787 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
17788 * doc/invoke.texi: Add -march=icelake.
17790 2018-02-02 Julia Koval <julia.koval@intel.com>
17792 * config/i386/i386.c (ix86_option_override_internal): Change flags type
17793 to wide_int_bitmask.
17794 * wide-int-bitmask.h: New.
17796 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17799 * config/i386/i386.md: Replace Pmode with word_mode in
17800 builtin_setjmp_setup and builtin_longjmp to support x32.
17802 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
17806 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
17808 (rs6000_supported_cpu_names): New static variable.
17809 (linux_cpu_translation_table): Likewise.
17810 (elf_platform) <cpu>: Define new static variable and use it.
17811 Translate kernel AT_PLATFORM name to canonical name if needed.
17812 Error if platform name is unknown.
17814 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
17817 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
17819 2018-02-01 Jeff Law <law@redhat.com>
17822 * config/i386/i386.c (release_scratch_register_on_entry): Add new
17823 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
17824 the scratch if RELEASE_VIA_POP is false.
17825 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
17826 If we have to save a temporary register, decrement SIZE appropriately.
17827 Pass new arguments to release_scratch_register_on_entry.
17828 (ix86_adjust_stack_and_probe): Likewise.
17829 (ix86_emit_probe_stack_range): Pass new arguments to
17830 release_scratch_register_on_entry.
17832 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
17834 PR rtl-optimization/84157
17835 * combine.c (change_zero_ext): Use REG_P predicate in
17836 front of HARD_REGISTER_P predicate.
17838 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
17840 * config/avr/avr.c (avr_option_override): Move disabling of
17841 -fdelete-null-pointer-checks to...
17842 * common/config/avr/avr-common.c (avr_option_optimization_table):
17845 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17847 PR tree-optimization/81635
17848 * tree-data-ref.c (split_constant_offset_1): For types that
17849 wrap on overflow, try to use range info to prove that wrapping
17852 2018-02-01 Renlin Li <renlin.li@arm.com>
17855 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
17856 TAILCALL_ADDR_REGS.
17857 (aarch64_register_move_cost): Likewise.
17858 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
17859 TAILCALL_ADDR_REGS.
17860 (REG_CLASS_NAMES): Likewise.
17861 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
17862 TAILCALL_ADDR_REGS. Remove IP registers.
17863 * config/aarch64/aarch64.md (Ucs): Update register constraint.
17865 2018-02-01 Richard Biener <rguenther@suse.de>
17867 * domwalk.h (dom_walker::dom_walker): Add additional constructor
17868 for specifying RPO order and allow NULL for that.
17869 * domwalk.c (dom_walker::dom_walker): Likewise.
17870 (dom_walker::walk): Handle NULL RPO order.
17871 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
17873 (rewrite_update_dom_walker): Likewise.
17874 (mark_def_dom_walker): Likewise.
17876 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17878 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
17879 (aarch64_maybe_expand_sve_subreg_move): Declare.
17880 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
17881 * config/aarch64/predicates.md (aarch64_any_register_operand): New
17883 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
17884 that are semantically a reverse operation.
17885 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
17886 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
17887 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
17889 (aarch64_can_change_mode_class): For big-endian, forbid changes
17890 between two SVE modes if they have different element sizes.
17892 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17894 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
17895 the TImode handling for big-endian targets.
17897 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17899 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
17900 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
17902 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
17903 Remove BSWAP handing for big-endian targets and use the form of
17904 LD1RQ appropariate for the mode.
17906 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17908 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
17909 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
17910 duplicated element.
17912 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17915 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
17916 check for operands that need to go through aarch64_sve_reload_be.
17918 2018-02-01 Jakub Jelinek <jakub@redhat.com>
17920 PR tree-optimization/81661
17921 PR tree-optimization/84117
17922 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
17923 * tree-eh.c: Include gimplify.h.
17924 (find_trapping_overflow, replace_trapping_overflow,
17925 rewrite_to_non_trapping_overflow): New functions.
17926 * tree-vect-loop.c: Include tree-eh.h.
17927 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
17928 * tree-data-ref.c: Include tree-eh.h.
17929 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
17931 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
17933 PR rtl-optimization/84123
17934 * combine.c (change_zero_ext): Check if hard register satisfies
17935 can_change_dest_mode before calling gen_lowpart_SUBREG.
17937 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
17940 * ira.c (ira_init_register_move_cost): Remove assert.
17942 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17944 PR rtl-optimization/84071
17945 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
17946 * doc/tm.texi: Regenerate.
17948 2018-01-31 Richard Biener <rguenther@suse.de>
17950 PR tree-optimization/84132
17951 * tree-data-ref.c (analyze_miv_subscript): Properly
17952 check whether evolution_function_is_affine_multivariate_p
17953 before calling gcd_of_steps_may_divide_p.
17955 2018-01-31 Julia Koval <julia.koval@intel.com>
17958 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
17959 * config/i386/i386.md (rdpid_rex64) New.
17960 (rdpid): Make 32bit only.
17962 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
17965 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
17966 an IDENTIFIER_NODE for FUNCTION_TYPE's.
17968 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17971 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17973 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17975 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17977 PR rtl-optimization/84071
17978 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
17979 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
17981 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17983 * config/arc/arc.c (arc_handle_aux_attribute): New function.
17984 (arc_attribute_table): Add 'aux' attribute.
17985 (arc_in_small_data_p): Consider aux like variables.
17986 (arc_is_aux_reg_p): New function.
17987 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
17988 (arc_get_aux_arg): New function.
17989 (prepare_move_operands): Handle aux-register access.
17990 (arc_handle_aux_attribute): New function.
17991 * doc/extend.texi (ARC Variable attributes): Add subsection.
17993 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17995 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
17996 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
17997 (arc_attribute_table): Add 'uncached' attribute.
17998 (arc_print_operand): Print '.di' flag for uncached memory
18000 (arc_in_small_data_p): Do not consider for small data the uncached
18002 (arc_is_uncached_mem_p): New function.
18003 * config/arc/predicates.md (compact_store_memory_operand): Check
18004 for uncached memory accesses.
18005 (nonvol_nonimm_operand): Likewise.
18006 * doc/extend.texi (ARC Type Attribute): New subsection.
18008 2018-01-31 Jakub Jelinek <jakub@redhat.com>
18011 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
18012 falign-loops=): Add Optimization flag.
18014 2018-01-30 Jeff Law <law@redhat.com>
18017 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
18018 INT_REGISTERS_SAVED. Check it prior to calling
18019 get_scratch_register_on_entry.
18020 (ix86_adjust_stack_and_probe): Similarly.
18021 (ix86_emit_probe_stack_range): Similarly.
18022 (ix86_expand_prologue): Corresponding changes.
18024 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18027 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
18028 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
18030 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
18033 * lra-constraints.c (curr_insn_transform): Process AND in the
18036 2018-01-30 Jakub Jelinek <jakub@redhat.com>
18038 PR rtl-optimization/83986
18039 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
18040 dependence against last_pending_memory_flush in addition to
18041 pending_jump_insns.
18043 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
18045 PR tree-optimization/81611
18046 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
18049 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18052 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
18055 2018-01-30 Richard Biener <rguenther@suse.de>
18056 Jakub Jelinek <jakub@redhat.com>
18058 PR tree-optimization/84111
18059 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
18060 inner loops added during recursion, as they don't have up-to-date
18063 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
18066 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
18067 (can_inline_edge_by_limits_p): ... here.
18068 (can_early_inline_edge_p, check_callers,
18069 update_caller_keys, update_callee_keys, recursive_inlining,
18070 add_new_edges_to_heap, speculation_useful_p,
18071 inline_small_functions,
18072 inline_small_functions, flatten_function,
18073 inline_to_all_callers_1): Update.
18075 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
18077 * profile-count.c (profile_count::combine_with_ipa_count): Handle
18080 2018-01-30 Richard Biener <rguenther@suse.de>
18082 PR tree-optimization/83008
18083 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
18084 invariant and constant vector uses in stmts when they need
18085 more than one stmt.
18087 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18090 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
18091 * configure: Regenerate.
18093 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
18095 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
18097 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
18098 Use gen_rtx_REG rather than gen_lowpart.
18100 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
18102 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
18103 rather than 0 when creating partial subregs.
18105 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
18107 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
18110 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
18113 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
18114 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
18115 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
18116 flags. This restores the settings used before the 2017-07-24.
18117 Turning off pre increment/decrement/modify allows IVOPTS to
18118 optimize DF/SF loops where the index is an int.
18120 2018-01-29 Richard Biener <rguenther@suse.de>
18121 Kelvin Nilsen <kelvin@gcc.gnu.org>
18124 * tree-vect-stmts.c (vectorizable_call): Don't call
18125 targetm.vectorize_builtin_md_vectorized_function if callee is
18128 2018-01-22 Carl Love <cel@us.ibm.com>
18130 * doc/extend.tex: Fix typo in second arg in
18131 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
18133 2018-01-29 Richard Biener <rguenther@suse.de>
18135 PR tree-optimization/84086
18136 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
18137 (flush_ssaname_freelist): When SSA names were released reset
18138 the SCEV hash table.
18140 2018-01-29 Richard Biener <rguenther@suse.de>
18142 PR tree-optimization/84057
18143 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
18144 removed paths when removing edges.
18146 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
18148 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
18149 -mfunction-return=@var{choice}.
18151 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
18153 PR diagnostic/84034
18154 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
18155 Handle CR like TAB.
18156 (layout::print_source_line): Likewise.
18157 (test_get_line_width_without_trailing_whitespace): Add test cases.
18159 2018-01-27 Jakub Jelinek <jakub@redhat.com>
18161 PR middle-end/84040
18162 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
18165 2018-01-26 Jim Wilson <jimw@sifive.com>
18167 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
18169 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
18172 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18174 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
18175 and CMP + SUB-immediate -> SUBS.
18177 2018-01-26 Martin Sebor <msebor@redhat.com>
18179 PR tree-optimization/83896
18180 * tree-ssa-strlen.c (get_string_len): Rename...
18181 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
18182 Avoid assuming length is constant.
18183 (handle_char_store): Use HOST_WIDE_INT for string length.
18185 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
18188 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
18189 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
18191 2018-01-26 Richard Biener <rguenther@suse.de>
18193 PR rtl-optimization/84003
18194 * dse.c (record_store): Only record redundant stores when
18195 the earlier store aliases at least all accesses the later one does.
18197 2018-01-26 Jakub Jelinek <jakub@redhat.com>
18199 PR rtl-optimization/83985
18200 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
18201 REG_CFA_RESTORE insns.
18202 (delete_unmarked_insns): Don't ignore separate shrink wrapping
18203 REG_CFA_RESTORE insns here.
18206 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
18207 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
18209 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18211 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
18212 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
18213 (arc_init): Likewise.
18214 (arc_override_options): Likewise.
18215 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
18217 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
18219 * config/arc/arc.h (TARGET_DBNZ): Define.
18220 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
18221 properly set the tune attribute.
18222 (dbnz): Use TARGET_DBNZ guard.
18223 * config/arc/arc.opt (mtune): Add core3 option.
18225 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18227 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
18228 recognize new pic like addresses.
18229 (arc_delegitimize_address): Clean up.
18231 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18233 * config/arc/arc-arches.def: Option mrf16 valid for all
18235 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
18236 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
18237 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
18238 * config/arc/arc-tables.opt: Regenerate.
18239 * config/arc/arc.c (arc_conditional_register_usage): Handle
18240 reduced register file case.
18241 (arc_file_start): Set must have build attributes.
18242 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
18243 mrf16 option value.
18244 * config/arc/arc.opt (mrf16): Add new option.
18245 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
18246 * config/arc/genmultilib.awk: Handle new mrf16 option.
18247 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
18248 * config/arc/t-multilib: Regenerate.
18249 * doc/invoke.texi (ARC Options): Document mrf16 option.
18251 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18253 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
18254 * config/arc/arc.c (arc_handle_secure_attribute): New function.
18255 (arc_attribute_table): Add 'secure_call' attribute.
18256 (arc_print_operand): Print secure call operand.
18257 (arc_function_ok_for_sibcall): Don't optimize tail calls when
18259 (arc_is_secure_call_p): New function. * config/arc/arc.md
18260 (call_i): Add support for sjli instruction.
18261 (call_value_i): Likewise.
18262 * config/arc/constraints.md (Csc): New constraint.
18264 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18265 John Eric Martin <John.Martin@emmicro-us.com>
18267 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
18268 * config/arc/arc.c (_arc_jli_section): New struct.
18269 (arc_jli_section): New type.
18270 (rc_jli_sections): New static variable.
18271 (arc_handle_jli_attribute): New function.
18272 (arc_attribute_table): Add jli_always and jli_fixed attribute.
18273 (arc_file_end): New function.
18274 (TARGET_ASM_FILE_END): Define.
18275 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
18276 (arc_add_jli_section): New function.
18277 (jli_call_scan): Likewise.
18278 (arc_reorg): Call jli_call_scan.
18279 (arc_output_addsi): Remove 'S' from printing asm operand.
18280 (arc_is_jli_call_p): New function.
18281 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
18283 (movhi_insn): Likewise.
18284 (movsi_insn): Likewise.
18285 (movsi_set_cc_insn): Likewise.
18286 (loadqi_update): Likewise.
18287 (load_zeroextendqisi_update): Likewise.
18288 (load_signextendqisi_update): Likewise.
18289 (loadhi_update): Likewise.
18290 (load_zeroextendhisi_update): Likewise.
18291 (load_signextendhisi_update): Likewise.
18292 (loadsi_update): Likewise.
18293 (loadsf_update): Likewise.
18294 (movsicc_insn): Likewise.
18295 (bset_insn): Likewise.
18296 (bxor_insn): Likewise.
18297 (bclr_insn): Likewise.
18298 (bmsk_insn): Likewise.
18299 (bicsi3_insn): Likewise.
18300 (cmpsi_cc_c_insn): Likewise.
18301 (movsi_ne): Likewise.
18302 (movsi_cond_exec): Likewise.
18303 (clrsbsi2): Likewise.
18304 (norm_f): Likewise.
18314 (call_i): Remove 'S' asm letter, add jli instruction.
18315 (call_value_i): Likewise.
18316 * config/arc/arc.op (mjli-always): New option.
18317 * config/arc/constraints.md (Cji): New constraint.
18318 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
18320 (subsf3_fpx): Likewise.
18321 (mulsf3_fpx): Likewise.
18322 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
18324 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
18325 function attrbutes.
18326 * doc/invoke.texi (ARC): Document mjli-always option.
18328 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
18330 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
18331 avoid addition with 0 and use incw and decw where possible.
18333 2018-01-26 Richard Biener <rguenther@suse.de>
18335 PR tree-optimization/81082
18336 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
18337 association if it requires casting to unsigned.
18338 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
18339 from fold_plusminus_mult_expr to catch important cases late when
18340 range info is available.
18342 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18344 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
18345 * configure.ac (hidden_linkonce): New test.
18346 * configure: Regenerate.
18347 * config.in: Regenerate.
18349 2018-01-26 Julia Koval <julia.koval@intel.com>
18351 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
18352 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
18353 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
18354 _mm_mask_bitshuffle_epi64_mask): Fix type.
18355 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
18356 USI_FTYPE_V4DI_V4DI_USI): Remove.
18357 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
18358 __builtin_ia32_vpshufbitqmb256_mask,
18359 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
18360 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
18361 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
18363 2018-01-26 Alan Modra <amodra@gmail.com>
18366 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
18367 UNSPEC_VBPERMQ. Sort other unspecs.
18369 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
18371 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
18373 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
18375 PR middle-end/83055
18376 * predict.c (drop_profile): Do not push/pop cfun; update also
18378 (handle_missing_profiles): Fix logic looking for zero profiles.
18380 2018-01-25 Jakub Jelinek <jakub@redhat.com>
18382 PR middle-end/83977
18383 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
18384 on functions with #pragma omp declare simd or functions with simd
18386 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
18387 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
18388 Remove trailing \n from warning_at calls.
18390 2018-01-25 Tom de Vries <tom@codesourcery.com>
18393 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
18394 for neutered workers.
18396 2018-01-24 Joseph Myers <joseph@codesourcery.com>
18399 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
18400 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
18402 2018-01-24 Jeff Law <law@redhat.com>
18405 * i386.c (get_probe_interval): Move to earlier point.
18406 (ix86_compute_frame_layout): If -fstack-clash-protection and
18407 the frame is larger than the probe interval, then use pushes
18408 to save registers rather than reg->mem moves.
18409 (ix86_expand_prologue): Remove conditional for int_registers_saved
18412 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
18415 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
18416 min/max for never referenced object.
18418 2018-01-24 Jakub Jelinek <jakub@redhat.com>
18420 PR middle-end/83977
18421 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
18423 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
18424 attributes from DECL_ATTRIBUTES (decl) without affecting
18425 DECL_ATTRIBUTES (current_function_decl).
18426 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
18427 functions with non-NULL DECL_ABSTRACT_ORIGIN.
18429 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
18431 PR tree-optimization/83979
18432 * fold-const.c (fold_comparison): Use constant_boolean_node
18433 instead of boolean_{true,false}_node.
18435 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
18437 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
18440 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18442 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18443 Simplify the clause that sets the length attribute.
18444 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18445 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
18446 clause that sets the length attribute.
18447 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18449 2018-01-24 Tom de Vries <tom@codesourcery.com>
18452 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
18453 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
18454 Add strict parameter.
18455 (prevent_branch_around_nothing): Insert dummy insn between branch to
18456 label and label with no ptx insn inbetween.
18457 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
18459 2018-01-24 Tom de Vries <tom@codesourcery.com>
18462 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
18463 for neutered threads in warp.
18464 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
18466 2018-01-24 Richard Biener <rguenther@suse.de>
18468 PR tree-optimization/83176
18469 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
18472 2018-01-24 Richard Biener <rguenther@suse.de>
18474 PR tree-optimization/82819
18475 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
18476 code generating pluses that are no-ops in the target precision.
18478 2018-01-24 Richard Biener <rguenther@suse.de>
18480 PR middle-end/84000
18481 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
18483 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18485 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
18486 to merge probabilities.
18487 * predict.c (probably_never_executed): Also mark as cold functions
18488 with global 0 profile and guessed local profile.
18489 * profile-count.c (profile_probability::combine_with_count): New
18491 * profile-count.h (profile_probability::operator*,
18492 profile_probability::operator*=, profile_probability::operator/,
18493 profile_probability::operator/=): Reduce precision to adjusted
18494 and set value to guessed on contradictory divisions.
18495 (profile_probability::combine_with_freq): Remove.
18496 (profile_probability::combine_wiht_count): Declare.
18497 (profile_count::force_nonzero):: Set to adjusted.
18498 (profile_count::probability_in):: Set quality to adjusted.
18499 * tree-ssa-tail-merge.c (replace_block_by): Use
18500 combine_with_count.
18502 2018-01-23 Andrew Waterman <andrew@sifive.com>
18503 Jim Wilson <jimw@sifive.com>
18505 * config/riscv/riscv.c (riscv_stack_boundary): New.
18506 (riscv_option_override): Set riscv_stack_boundary. Handle
18507 riscv_preferred_stack_boundary_arg.
18508 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
18509 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
18510 (STACK_BOUNDARY): Set to riscv_stack_boundary.
18511 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
18512 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
18513 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
18515 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
18518 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
18519 of struct ix86_frame.
18520 (ix86_expand_epilogue): Likewise. Add a local variable for
18521 the reg_save_offset field in struct ix86_frame.
18523 2018-01-23 Bin Cheng <bin.cheng@arm.com>
18525 PR tree-optimization/82604
18526 * tree-loop-distribution.c (enum partition_kind): New enum item
18527 PKIND_PARTIAL_MEMSET.
18528 (partition_builtin_p): Support above new enum item.
18529 (generate_code_for_partition): Ditto.
18530 (compute_access_range): Differentiate cases that equality can be
18531 proven at all loops, the innermost loops or no loops.
18532 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
18533 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
18534 (finalize_partitions, distribute_loop): Don't fuse partition of
18535 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
18536 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
18537 parloop is enabled.
18539 2018-01-23 Martin Liska <mliska@suse.cz>
18541 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
18542 order to ignore the predictor.
18543 (PRED_POLYMORPHIC_CALL): Likewise.
18544 (PRED_RECURSIVE_CALL): Likewise.
18546 2018-01-23 Martin Liska <mliska@suse.cz>
18548 * tree-profile.c (tree_profiling): Print function header to
18549 aware reader which function we are working on.
18550 * value-prof.c (gimple_find_values_to_profile): Do not print
18551 not interesting value histograms.
18553 2018-01-23 Martin Liska <mliska@suse.cz>
18555 * profile-count.h (enum profile_quality): Add
18556 profile_uninitialized as the first value. Do not number values
18557 as they are zero based.
18558 (profile_count::verify): Update sanity check.
18559 (profile_probability::verify): Likewise.
18561 2018-01-23 Nathan Sidwell <nathan@acm.org>
18563 * doc/invoke.texi (ffor-scope): Deprecate.
18565 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18567 PR tree-optimization/83510
18568 * domwalk.c (set_all_edges_as_executable): New function.
18569 (dom_walker::dom_walker): Convert bool param
18570 "skip_unreachable_blocks" to enum reachability. Move setup of
18571 edge flags to set_all_edges_as_executable and only do it when
18572 reachability is REACHABLE_BLOCKS.
18573 * domwalk.h (enum dom_walker::reachability): New enum.
18574 (dom_walker::dom_walker): Convert bool param
18575 "skip_unreachable_blocks" to enum reachability.
18576 (set_all_edges_as_executable): New decl.
18577 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
18578 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
18580 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
18581 but converting true to REACHABLE_BLOCKS.
18582 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
18584 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
18585 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
18586 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
18588 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
18589 if check_all_array_refs will be called.
18591 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18593 * tree.c (selftest::test_location_wrappers): Add more test
18596 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18598 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
18599 (selftest::test_bit_in_range): Likewise.
18601 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
18604 * doc/sourcebuild.texi (vect_float): Say that the selector
18605 only describes the situation when -funsafe-math-optimizations is on.
18606 (vect_float_strict): Document.
18608 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
18610 PR tree-optimization/83965
18611 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
18612 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
18613 instead of checking only for a reduction.
18614 (vect_recog_widen_sum_pattern): Likewise.
18616 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18618 * predict.c (probably_never_executed): Only use precise profile info.
18619 (compute_function_frequency): Skip after inlining hack since we now
18620 have quality checking.
18622 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18624 * profile-count.h (profile_probability::very_unlikely,
18625 profile_probability::unlikely, profile_probability::even): Set
18626 precision to guessed.
18628 2018-01-23 Richard Biener <rguenther@suse.de>
18630 PR tree-optimization/83963
18631 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
18632 Properly terminate dominator walk when crossing the exit edge not
18633 when visiting its source block.
18635 2018-01-23 Jakub Jelinek <jakub@redhat.com>
18638 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
18639 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
18641 2018-01-22 Jakub Jelinek <jakub@redhat.com>
18643 PR tree-optimization/83957
18644 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
18645 semicolon after for body surrounded by braces.
18647 PR tree-optimization/83081
18648 * profile-count.h (profile_probability::split): New method.
18649 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
18650 Use profile_probability::split.
18651 (do_compare_rtx_and_jump): Fix adjustment of probabilities
18652 when splitting a single conditional jump into 2.
18654 2018-01-22 David Malcolm <dmalcolm@redhat.com>
18656 PR tree-optimization/69452
18657 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
18660 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18662 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
18663 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
18664 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
18666 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18668 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
18670 * config/rl78/rl78.md (movdi): New define_expand.
18671 * config/rl78/rl78.c (rl78_split_movdi): New function.
18673 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
18676 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
18678 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
18679 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
18680 128-bit to produce an UNSPEC move to get the double word with the
18681 signbit and then a shift directly to do signbit.
18682 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
18683 implementation with a new version that just does either a direct
18684 move or a regular move. Move memory interface to separate insns.
18685 Move insns so they are next to the expander.
18686 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
18687 with signbit move. Split big and little endian case.
18688 (signbit<mode>2_dm_mem_le): Likewise.
18689 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
18690 (signbit<mode>2_dm2): Likewise.
18692 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18694 * config/rl78/rl78.md (anddi3): New define_expand.
18696 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18698 * config/rl78/rl78.md (umindi3): New define_expand.
18700 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18702 * config/rl78/rl78.md (smindi3): New define_expand.
18704 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18706 * config/rl78/rl78.md (smaxdi3): New define_expand.
18708 2018-01-22 Carl Love <cel@us.ibm.com>
18710 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
18711 LVX_V1TI): Add macro expansion.
18712 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
18713 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
18714 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
18715 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
18716 Change check to determine if the instruction is a byte reversing
18717 entry. Fix typo in comment.
18718 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
18719 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
18720 Add def_builtin calls for new builtins.
18721 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
18722 Add define_insn expansion.
18724 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18726 * config/rl78/rl78.md (umaxdi3): New define_expand.
18728 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18730 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
18731 for non-QImode registers.
18733 2018-01-22 Richard Biener <rguenther@suse.de>
18735 PR tree-optimization/83963
18736 * graphite-scop-detection.c (scop_detection::get_sese): Delay
18737 including the loop exit block.
18738 (scop_detection::merge_sese): Likewise.
18739 (scop_detection::add_scop): Do it here instead.
18741 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18743 * doc/sourcebuild.texi (arm_softfloat): Document.
18745 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
18748 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
18749 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
18750 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
18752 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18753 David Edelsohn <dje.gcc@gmail.com>
18756 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18757 Change "crset eq" to "crset 2".
18758 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18759 (*call_indirect_aix<mode>_nospec): Likewise.
18760 (*call_value_indirect_aix<mode>_nospec): Likewise.
18761 (*call_indirect_elfv2<mode>_nospec): Likewise.
18762 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
18763 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
18764 change assembly output from . to $.
18765 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18766 (indirect_jump<mode>_nospec): Change assembly output from . to $.
18767 (*tablejump<mode>_internal1_nospec): Likewise.
18769 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
18772 * config/sh/sh_optimize_sett_clrt.cc:
18773 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
18775 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
18777 PR tree-optimization/83940
18778 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
18779 offset_dt to vect_constant_def rather than vect_unknown_def_type.
18780 (vect_check_load_store_mask): Add a mask_dt_out parameter and
18781 use it to pass back the definition type.
18782 (vect_check_store_rhs): Likewise rhs_dt_out.
18783 (vect_build_gather_load_calls): Add a mask_dt argument and use
18784 it instead of a call to vect_is_simple_use.
18785 (vectorizable_store): Update calls to vect_check_load_store_mask
18786 and vect_check_store_rhs. Use the dt returned by the latter instead
18787 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
18788 instead of calls to vect_is_simple_use. Pass the scalar rather
18789 than the vector operand to vect_is_simple_use when handling
18790 second and subsequent copies of an rhs value.
18791 (vectorizable_load): Update calls to vect_check_load_store_mask
18792 and vect_build_gather_load_calls. Use the cached mask_dt and
18793 gs_info.offset_dt instead of calls to vect_is_simple_use.
18795 2018-01-20 Jakub Jelinek <jakub@redhat.com>
18797 PR middle-end/83945
18798 * tree-emutls.c: Include gimplify.h.
18799 (lower_emutls_2): New function.
18800 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
18801 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
18802 it before further processing.
18805 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
18806 UINTVAL (trueop1) instead of INTVAL (op1).
18808 2018-01-19 Jakub Jelinek <jakub@redhat.com>
18812 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
18813 INCOMING_FRAME_SP_OFFSET if not defined.
18814 (scan_trace): Add ENTRY argument. If true and
18815 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
18816 emit a note to adjust the CFA offset.
18817 (create_cfi_notes): Adjust scan_trace callers.
18818 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
18819 INCOMING_FRAME_SP_OFFSET in the CIE.
18820 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
18821 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
18823 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
18824 * doc/tm.texi: Regenerated.
18826 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18828 PR rtl-optimization/83147
18829 * lra-constraints.c (remove_inheritance_pseudos): Use
18830 lra_substitute_pseudo_within_insn.
18832 2018-01-19 Tom de Vries <tom@codesourcery.com>
18833 Cesar Philippidis <cesar@codesourcery.com>
18836 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
18838 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
18841 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
18842 spaces for function labels.
18844 2018-01-19 Martin Liska <mliska@suse.cz>
18846 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
18847 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
18848 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
18849 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
18850 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
18851 (PRED_CONST_RETURN): Change from 69 to 65.
18852 (PRED_NULL_RETURN): Change from 91 to 71.
18853 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
18854 (PRED_LOOP_GUARD): Change from 66 to 73.
18856 2018-01-19 Martin Liska <mliska@suse.cz>
18858 * predict.c (predict_insn_def): Add new assert.
18859 (struct branch_predictor): Change type to signed integer.
18860 (test_prediction_value_range): Amend test to cover
18861 PROB_UNINITIALIZED.
18862 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
18863 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
18864 (PRED_LOOP_ITERATIONS_MAX): Likewise.
18865 (PRED_LOOP_IV_COMPARE): Likewise.
18866 * predict.h (PROB_UNINITIALIZED): Define new constant.
18868 2018-01-19 Martin Liska <mliska@suse.cz>
18870 * predict.c (dump_prediction): Add new format for
18871 analyze_brprob.py script which is enabled with -details
18873 * profile-count.h (precise_p): New function.
18875 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18877 PR tree-optimization/83922
18878 * tree-vect-loop.c (vect_verify_full_masking): Return false if
18879 there are no statements that need masking.
18880 (vect_active_double_reduction_p): New function.
18881 (vect_analyze_loop_operations): Use it when handling phis that
18882 are not in the loop header.
18884 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18886 PR tree-optimization/83914
18887 * tree-vect-loop.c (vectorizable_induction): Don't convert
18888 init_expr or apply the peeling adjustment for inductions
18889 that are nested within the vectorized loop.
18891 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18893 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
18896 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18900 * function.h (gimplify_parameters): Add gimple_seq * argument.
18901 * function.c: Include gimple.h and options.h.
18902 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
18903 for the added local temporaries if needed.
18904 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
18905 if there are any parameter cleanups, wrap whole body into a
18906 try/finally with the cleanups.
18908 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
18911 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
18912 Use GET_MODE_CLASS for scalar floating point.
18914 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18918 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
18919 Fix call of call_cgraph_insertion_hooks.
18921 2018-01-18 Martin Sebor <msebor@redhat.com>
18923 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
18925 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18928 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
18931 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
18934 * common.opt: (-ffile-prefix-map): New option.
18935 * opts.c (common_handle_option): Defer it.
18936 * opts-global.c (handle_common_deferred_options): Handle it.
18937 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
18938 * file-prefix-map.h: New file.
18939 (remap_debug_filename, add_debug_prefix_map): ...here.
18940 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
18941 * final.c (debug_prefix_map, add_debug_prefix_map
18942 remap_debug_filename): Move to...
18943 * file-prefix-map.c: New file.
18944 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
18945 generalize, get rid of alloca(), use strrchr() instead of strchr().
18946 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
18947 Implement in terms of add_prefix_map().
18948 (remap_macro_filename, remap_debug_filename): Implement in term of
18950 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
18951 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
18952 * dbxout.c: Include file-prefix-map.h.
18953 * varasm.c: Likewise.
18954 * vmsdbgout.c: Likewise.
18955 * xcoffout.c: Likewise.
18956 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
18957 * doc/cppopts.texi (-fmacro-prefix-map): Document.
18958 * doc/invoke.texi (-ffile-prefix-map): Document.
18959 (-fdebug-prefix-map): Update description.
18961 2018-01-18 Martin Liska <mliska@suse.cz>
18963 * config/i386/i386.c (indirect_thunk_name): Document that also
18965 (output_indirect_thunk): Document why both instructions
18966 (pause and lfence) are generated.
18968 2018-01-18 Richard Biener <rguenther@suse.de>
18970 PR tree-optimization/83887
18971 * graphite-scop-detection.c
18972 (scop_detection::get_nearest_dom_with_single_entry): Remove.
18973 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
18974 (scop_detection::merge_sese): Re-implement with a flood-fill
18975 algorithm that properly finds a SESE region if it exists.
18977 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18980 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
18981 pointer_diff optimizations use view_convert instead of convert.
18983 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18985 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18986 Generate different code for -mno-speculate-indirect-jumps.
18987 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18988 (*call_indirect_aix<mode>): Disable for
18989 -mno-speculate-indirect-jumps.
18990 (*call_indirect_aix<mode>_nospec): New define_insn.
18991 (*call_value_indirect_aix<mode>): Disable for
18992 -mno-speculate-indirect-jumps.
18993 (*call_value_indirect_aix<mode>_nospec): New define_insn.
18994 (*sibcall_nonlocal_sysv<mode>): Generate different code for
18995 -mno-speculate-indirect-jumps.
18996 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18998 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
19000 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
19001 long double type, set the flags for noting the default long double
19002 type, even if we don't pass or return a long double type.
19004 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
19007 * ipa-inline.c (flatten_function): Do not overwrite final inlining
19010 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
19012 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
19013 support for merge[hl].
19014 (fold_mergehl_helper): New helper function.
19015 (tree-vector-builder.h): New #include for tree_vector_builder usage.
19016 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
19017 (altivec_vmrglw_direct): Add xxmrglw insn.
19019 2018-01-17 Andrew Waterman <andrew@sifive.com>
19021 * config/riscv/riscv.c (riscv_conditional_register_usage): If
19022 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
19024 2018-01-17 David Malcolm <dmalcolm@redhat.com>
19027 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
19028 call the lto_location_cache before reading the
19029 DECL_SOURCE_LOCATION of the types.
19031 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
19032 Richard Sandiford <richard.sandiford@linaro.org>
19034 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
19035 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
19036 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
19037 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
19038 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
19040 * config/aarch64/constraints.md (aarch64_movti_operand):
19042 * config/aarch64/predicates.md (Uti): Add new constraint.
19044 2018-01-17 Carl Love <cel@us.ibm.com>
19046 * config/rs6000/vsx.md (define_expand xl_len_r,
19047 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
19048 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
19050 (define_expand, define_insn): Move the shift left from the
19051 define_insn to the define_expand for lxvl and stxvl instructions.
19052 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
19053 and XL_LEN_R definitions to PURE.
19055 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
19057 * config/i386/i386.c (indirect_thunk_name): Declare regno
19058 as unsigned int. Compare regno with INVALID_REGNUM.
19059 (output_indirect_thunk): Ditto.
19060 (output_indirect_thunk_function): Ditto.
19061 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
19062 in the call to output_indirect_thunk_function.
19064 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
19066 PR middle-end/83884
19067 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
19068 rather than the size of inner_type to determine the stack slot size
19069 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
19071 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
19074 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
19077 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
19079 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
19080 endian Linux systems to optionally enable multilibs for selecting
19081 the long double type if the user configured an explicit type.
19082 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
19083 have no long double multilibs if not defined.
19084 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
19085 warn if the user used -mabi={ieee,ibm}longdouble and we built
19086 multilibs for long double.
19087 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
19088 appropriate multilib option.
19089 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
19091 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
19092 for building long double multilibs.
19093 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
19095 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
19097 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
19100 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
19102 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
19105 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
19108 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
19111 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
19113 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
19114 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
19116 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
19118 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
19119 different rtl trees depending on TARGET_64BIT.
19120 (rs6000_gen_lvx): Likewise.
19122 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
19124 * config/visium/visium.md (nop): Tweak comment.
19125 (hazard_nop): Likewise.
19127 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19129 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
19130 -mspeculate-indirect-jumps.
19131 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
19132 for -mno-speculate-indirect-jumps.
19133 (*call_indirect_elfv2<mode>_nospec): New define_insn.
19134 (*call_value_indirect_elfv2<mode>): Disable for
19135 -mno-speculate-indirect-jumps.
19136 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
19137 (indirect_jump): Emit different RTL for
19138 -mno-speculate-indirect-jumps.
19139 (*indirect_jump<mode>): Disable for
19140 -mno-speculate-indirect-jumps.
19141 (*indirect_jump<mode>_nospec): New define_insn.
19142 (tablejump): Emit different RTL for
19143 -mno-speculate-indirect-jumps.
19144 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
19145 (tablejumpsi_nospec): New define_expand.
19146 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
19147 (tablejumpdi_nospec): New define_expand.
19148 (*tablejump<mode>_internal1): Disable for
19149 -mno-speculate-indirect-jumps.
19150 (*tablejump<mode>_internal1_nospec): New define_insn.
19151 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
19154 2018-01-16 Artyom Skrobov tyomitch@gmail.com
19156 * caller-save.c (insert_save): Drop unnecessary parameter. All
19159 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19160 Richard Biener <rguenth@suse.de>
19163 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
19164 return early, inline manually is_gimple_sizepos. Make sure if we
19165 call gimplify_expr we don't end up with a gimple constant.
19166 * tree.c (variably_modified_type_p): Don't return true for
19167 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
19168 * gimplify.h (is_gimple_sizepos): Remove.
19170 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19172 PR tree-optimization/83857
19173 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
19174 vectorizable_live_operation for pure SLP statements.
19175 (vectorizable_live_operation): Handle PHIs.
19177 2018-01-16 Richard Biener <rguenther@suse.de>
19179 PR tree-optimization/83867
19180 * tree-vect-stmts.c (vect_transform_stmt): Precompute
19181 nested_in_vect_loop_p since the scalar stmt may get invalidated.
19183 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19186 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
19187 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
19188 If off is not INTEGER_CST, issue a may not be aligned warning
19189 rather than isn't aligned. Use isn%'t rather than isn't.
19190 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
19192 <case MULT_EXPR>: Improve the case when bottom and one of the
19193 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
19194 operand, in that case check if the other operand is multiple of
19195 bottom divided by the INTEGER_CST operand.
19197 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19200 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
19201 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
19202 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
19203 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
19204 * config/pa/pa.c (pa_function_arg_advance): Likewise.
19205 (pa_function_arg, pa_arg_partial_bytes): Likewise.
19206 (pa_function_arg_size): New function.
19208 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19210 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
19211 in a separate statement.
19213 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19215 PR tree-optimization/83847
19216 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
19217 group gathers and scatters.
19219 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19221 PR rtl-optimization/86620
19222 * params.def (max-sched-ready-insns): Bump minimum value to 1.
19224 PR rtl-optimization/83213
19225 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
19226 to last if both are JUMP_INSNs.
19228 PR tree-optimization/83843
19229 * gimple-ssa-store-merging.c
19230 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
19231 store_immediate_info for bswap/nop orig_stores.
19233 2018-01-15 Andrew Waterman <andrew@sifive.com>
19235 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
19237 <UDIV>: Increase cost if !TARGET_DIV.
19239 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
19241 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
19242 (define_attr "cr_logical_3op"): New.
19243 (cceq_ior_compare): Adjust.
19244 (cceq_ior_compare_complement): Adjust.
19245 (*cceq_rev_compare): Adjust.
19246 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
19247 (is_cracked_insn): Adjust.
19248 (insn_must_be_first_in_group): Adjust.
19249 * config/rs6000/40x.md: Adjust.
19250 * config/rs6000/440.md: Adjust.
19251 * config/rs6000/476.md: Adjust.
19252 * config/rs6000/601.md: Adjust.
19253 * config/rs6000/603.md: Adjust.
19254 * config/rs6000/6xx.md: Adjust.
19255 * config/rs6000/7450.md: Adjust.
19256 * config/rs6000/7xx.md: Adjust.
19257 * config/rs6000/8540.md: Adjust.
19258 * config/rs6000/cell.md: Adjust.
19259 * config/rs6000/e300c2c3.md: Adjust.
19260 * config/rs6000/e500mc.md: Adjust.
19261 * config/rs6000/e500mc64.md: Adjust.
19262 * config/rs6000/e5500.md: Adjust.
19263 * config/rs6000/e6500.md: Adjust.
19264 * config/rs6000/mpc.md: Adjust.
19265 * config/rs6000/power4.md: Adjust.
19266 * config/rs6000/power5.md: Adjust.
19267 * config/rs6000/power6.md: Adjust.
19268 * config/rs6000/power7.md: Adjust.
19269 * config/rs6000/power8.md: Adjust.
19270 * config/rs6000/power9.md: Adjust.
19271 * config/rs6000/rs64.md: Adjust.
19272 * config/rs6000/titan.md: Adjust.
19274 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19276 * config/i386/predicates.md (indirect_branch_operand): Rewrite
19277 ix86_indirect_branch_register logic.
19279 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19281 * config/i386/constraints.md (Bs): Update
19282 ix86_indirect_branch_register check. Don't check
19283 ix86_indirect_branch_register with GOT_memory_operand.
19285 * config/i386/predicates.md (GOT_memory_operand): Don't check
19286 ix86_indirect_branch_register here.
19287 (GOT32_symbol_operand): Likewise.
19289 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19291 * config/i386/predicates.md (constant_call_address_operand):
19292 Rewrite ix86_indirect_branch_register logic.
19293 (sibcall_insn_operand): Likewise.
19295 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19297 * config/i386/constraints.md (Bs): Replace
19298 ix86_indirect_branch_thunk_register with
19299 ix86_indirect_branch_register.
19301 * config/i386/i386.md (indirect_jump): Likewise.
19302 (tablejump): Likewise.
19303 (*sibcall_memory): Likewise.
19304 (*sibcall_value_memory): Likewise.
19305 Peepholes of indirect call and jump via memory: Likewise.
19306 * config/i386/i386.opt: Likewise.
19307 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19308 (GOT_memory_operand): Likewise.
19309 (call_insn_operand): Likewise.
19310 (sibcall_insn_operand): Likewise.
19311 (GOT32_symbol_operand): Likewise.
19313 2018-01-15 Jakub Jelinek <jakub@redhat.com>
19315 PR middle-end/83837
19316 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
19317 type rather than type addr's type points to.
19318 (expand_omp_atomic_mutex): Likewise.
19319 (expand_omp_atomic): Likewise.
19321 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19324 * config/i386/i386.c (output_indirect_thunk_function): Use
19325 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
19326 for __x86_return_thunk.
19328 2018-01-15 Richard Biener <rguenther@suse.de>
19330 PR middle-end/83850
19331 * expmed.c (extract_bit_field_1): Fix typo.
19333 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19336 * config/arm/iterators.md (VF): New mode iterator.
19337 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
19338 Remove integer-related logic from pattern.
19339 (neon_vabd<mode>_3): Likewise.
19341 2018-01-15 Jakub Jelinek <jakub@redhat.com>
19343 PR middle-end/82694
19344 * common.opt (fstrict-overflow): No longer an alias.
19345 (fwrapv-pointer): New option.
19346 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
19347 also for pointer types based on flag_wrapv_pointer.
19348 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
19349 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
19350 opts->x_flag_wrapv got set.
19351 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
19352 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
19353 POINTER_TYPE_OVERFLOW_UNDEFINED.
19354 * match.pd: Likewise in address comparison pattern.
19355 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
19357 2018-01-15 Richard Biener <rguenther@suse.de>
19360 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
19361 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
19362 Reset type names to their identifier if their TYPE_DECL doesn't
19363 have linkage (and thus is used for ODR and devirt).
19364 (save_debug_info_for_decl): Remove.
19365 (save_debug_info_for_type): Likewise.
19366 (add_tree_to_fld_list): Adjust.
19367 * tree-pretty-print.c (dump_generic_node): Make dumping of
19368 type names more robust.
19370 2018-01-15 Richard Biener <rguenther@suse.de>
19372 * BASE-VER: Bump to 8.0.1.
19374 2018-01-14 Martin Sebor <msebor@redhat.com>
19377 * builtins.c (check_access): Avoid warning when the no-warning bit
19380 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
19382 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
19383 * ira-color (allocno_hard_regs_compare): Likewise.
19385 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
19388 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
19389 Use .pushsection/.popsection.
19391 2018-01-14 Martin Sebor <msebor@redhat.com>
19394 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
19396 2018-01-14 Jakub Jelinek <jakub@redhat.com>
19398 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
19399 entry from extra_headers.
19400 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
19401 extra_headers, make the list bitwise identical to the i?86-*-* one.
19403 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19405 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
19406 -mcmodel=large with -mindirect-branch=thunk,
19407 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
19408 -mfunction-return=thunk-extern.
19409 * doc/invoke.texi: Document -mcmodel=large is incompatible with
19410 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
19411 -mfunction-return=thunk and -mfunction-return=thunk-extern.
19413 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19415 * config/i386/i386.c (print_reg): Print the name of the full
19416 integer register without '%'.
19417 (ix86_print_operand): Handle 'V'.
19418 * doc/extend.texi: Document 'V' modifier.
19420 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19422 * config/i386/constraints.md (Bs): Disallow memory operand for
19423 -mindirect-branch-register.
19425 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19426 (GOT_memory_operand): Likewise.
19427 (call_insn_operand): Likewise.
19428 (sibcall_insn_operand): Likewise.
19429 (GOT32_symbol_operand): Likewise.
19430 * config/i386/i386.md (indirect_jump): Call convert_memory_address
19431 for -mindirect-branch-register.
19432 (tablejump): Likewise.
19433 (*sibcall_memory): Likewise.
19434 (*sibcall_value_memory): Likewise.
19435 Disallow peepholes of indirect call and jump via memory for
19436 -mindirect-branch-register.
19437 (*call_pop): Replace m with Bw.
19438 (*call_value_pop): Likewise.
19439 (*sibcall_pop_memory): Replace m with Bs.
19440 * config/i386/i386.opt (mindirect-branch-register): New option.
19441 * doc/invoke.texi: Document -mindirect-branch-register option.
19443 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19445 * config/i386/i386-protos.h (ix86_output_function_return): New.
19446 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
19447 set function_return_type.
19448 (indirect_thunk_name): Add ret_p to indicate thunk for function
19450 (output_indirect_thunk_function): Pass false to
19451 indirect_thunk_name.
19452 (ix86_output_indirect_branch_via_reg): Likewise.
19453 (ix86_output_indirect_branch_via_push): Likewise.
19454 (output_indirect_thunk_function): Create alias for function
19455 return thunk if regno < 0.
19456 (ix86_output_function_return): New function.
19457 (ix86_handle_fndecl_attribute): Handle function_return.
19458 (ix86_attribute_table): Add function_return.
19459 * config/i386/i386.h (machine_function): Add
19460 function_return_type.
19461 * config/i386/i386.md (simple_return_internal): Use
19462 ix86_output_function_return.
19463 (simple_return_internal_long): Likewise.
19464 * config/i386/i386.opt (mfunction-return=): New option.
19465 (indirect_branch): Mention -mfunction-return=.
19466 * doc/extend.texi: Document function_return function attribute.
19467 * doc/invoke.texi: Document -mfunction-return= option.
19469 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19471 * config/i386/i386-opts.h (indirect_branch): New.
19472 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
19473 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
19474 with local indirect jump when converting indirect call and jump.
19475 (ix86_set_indirect_branch_type): New.
19476 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
19477 (indirectlabelno): New.
19478 (indirect_thunk_needed): Likewise.
19479 (indirect_thunk_bnd_needed): Likewise.
19480 (indirect_thunks_used): Likewise.
19481 (indirect_thunks_bnd_used): Likewise.
19482 (INDIRECT_LABEL): Likewise.
19483 (indirect_thunk_name): Likewise.
19484 (output_indirect_thunk): Likewise.
19485 (output_indirect_thunk_function): Likewise.
19486 (ix86_output_indirect_branch_via_reg): Likewise.
19487 (ix86_output_indirect_branch_via_push): Likewise.
19488 (ix86_output_indirect_branch): Likewise.
19489 (ix86_output_indirect_jmp): Likewise.
19490 (ix86_code_end): Call output_indirect_thunk_function if needed.
19491 (ix86_output_call_insn): Call ix86_output_indirect_branch if
19493 (ix86_handle_fndecl_attribute): Handle indirect_branch.
19494 (ix86_attribute_table): Add indirect_branch.
19495 * config/i386/i386.h (machine_function): Add indirect_branch_type
19496 and has_local_indirect_jump.
19497 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
19499 (tablejump): Likewise.
19500 (*indirect_jump): Use ix86_output_indirect_jmp.
19501 (*tablejump_1): Likewise.
19502 (simple_return_indirect_internal): Likewise.
19503 * config/i386/i386.opt (mindirect-branch=): New option.
19504 (indirect_branch): New.
19507 (thunk-inline): Likewise.
19508 (thunk-extern): Likewise.
19509 * doc/extend.texi: Document indirect_branch function attribute.
19510 * doc/invoke.texi: Document -mindirect-branch= option.
19512 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
19515 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
19517 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
19519 * ipa-inline.c (want_inline_small_function_p): Return false if
19520 inlining has already failed with CIF_FINAL_ERROR.
19521 (update_caller_keys): Call want_inline_small_function_p before
19523 (update_callee_keys): Likewise.
19525 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
19527 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
19529 (rs6000_quadword_masked_address_p): Likewise.
19530 (quad_aligned_load_p): Likewise.
19531 (quad_aligned_store_p): Likewise.
19532 (const_load_sequence_p): Add comment to describe the outer-most loop.
19533 (mimic_memory_attributes_and_flags): New function.
19534 (rs6000_gen_stvx): Likewise.
19535 (replace_swapped_aligned_store): Likewise.
19536 (rs6000_gen_lvx): Likewise.
19537 (replace_swapped_aligned_load): Likewise.
19538 (replace_swapped_load_constant): Capitalize argument name in
19539 comment describing this function.
19540 (rs6000_analyze_swaps): Add a third pass to search for vector loads
19541 and stores that access quad-word aligned addresses and replace
19542 with stvx or lvx instructions when appropriate.
19543 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
19544 New function prototype.
19545 (rs6000_quadword_masked_address_p): Likewise.
19546 (rs6000_gen_lvx): Likewise.
19547 (rs6000_gen_stvx): Likewise.
19548 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
19549 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
19550 when memory address is aligned.
19551 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
19552 this split to select lvx instruction when memory address is aligned.
19553 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
19554 instruction when memory address is aligned.
19555 (*vsx_le_perm_load_v16qi): Likewise.
19556 (four unnamed splitters): Modify to select the stvx instruction
19557 when memory is aligned.
19559 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
19561 * predict.c (determine_unlikely_bbs): Handle correctly BBs
19562 which appears in the queue multiple times.
19564 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19565 Alan Hayward <alan.hayward@arm.com>
19566 David Sherwood <david.sherwood@arm.com>
19568 * tree-vectorizer.h (vec_lower_bound): New structure.
19569 (_loop_vec_info): Add check_nonzero and lower_bounds.
19570 (LOOP_VINFO_CHECK_NONZERO): New macro.
19571 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
19572 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
19573 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
19574 fields. Make seg_len the distance travelled, not including the
19576 (dr_direction_indicator): Declare.
19577 (dr_zero_step_indicator): Likewise.
19578 (dr_known_forward_stride_p): Likewise.
19579 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
19581 (runtime_alias_check_p): Allow runtime alias checks with
19583 (operator ==): Compare access_size and align.
19584 (prune_runtime_alias_test_list): Rework for new distinction between
19585 the access_size and seg_len.
19586 (create_intersect_range_checks_index): Likewise. Cope with polynomial
19588 (get_segment_min_max): New function.
19589 (create_intersect_range_checks): Use it.
19590 (dr_step_indicator): New function.
19591 (dr_direction_indicator): Likewise.
19592 (dr_zero_step_indicator): Likewise.
19593 (dr_known_forward_stride_p): Likewise.
19594 * tree-loop-distribution.c (data_ref_segment_size): Return
19595 DR_STEP * (niters - 1).
19596 (compute_alias_check_pairs): Update call to the dr_with_seg_len
19598 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
19599 (vect_preserves_scalar_order_p): New function, split out from...
19600 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
19601 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
19602 (vect_vfa_access_size): New function.
19603 (vect_vfa_align): Likewise.
19604 (vect_compile_time_alias): Take access_size_a and access_b arguments.
19605 (dump_lower_bound): New function.
19606 (vect_check_lower_bound): Likewise.
19607 (vect_small_gap_p): Likewise.
19608 (vectorizable_with_step_bound_p): Likewise.
19609 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
19610 depencies if the vectorization factor is 1. Convert the checks
19611 for nonzero steps into checks on the bounds of DR_STEP. Try using
19612 a bunds check for variable steps if the minimum required step is
19613 relatively small. Update calls to the dr_with_seg_len
19614 constructor and to vect_compile_time_alias.
19615 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
19617 (vect_loop_versioning): Call it.
19618 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
19620 (vect_estimate_min_profitable_iters): Account for any bounds checks.
19622 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19623 Alan Hayward <alan.hayward@arm.com>
19624 David Sherwood <david.sherwood@arm.com>
19626 * doc/sourcebuild.texi (vect_scatter_store): Document.
19627 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
19629 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
19631 * genopinit.c (main): Add supports_vec_scatter_store and
19632 supports_vec_scatter_store_cached to target_optabs.
19633 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
19634 IFN_MASK_SCATTER_STORE.
19635 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
19637 * internal-fn.h (internal_store_fn_p): Declare.
19638 (internal_fn_stored_value_index): Likewise.
19639 * internal-fn.c (scatter_store_direct): New macro.
19640 (expand_scatter_store_optab_fn): New function.
19641 (direct_scatter_store_optab_supported_p): New macro.
19642 (internal_store_fn_p): New function.
19643 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
19644 IFN_MASK_SCATTER_STORE.
19645 (internal_fn_mask_index): Likewise.
19646 (internal_fn_stored_value_index): New function.
19647 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
19648 for scatter stores.
19649 * optabs-query.h (supports_vec_scatter_store_p): Declare.
19650 * optabs-query.c (supports_vec_scatter_store_p): New function.
19651 * tree-vectorizer.h (vect_get_store_rhs): Declare.
19652 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
19653 true for scatter stores.
19654 (vect_gather_scatter_fn_p): Handle scatter stores too.
19655 (vect_check_gather_scatter): Consider using scatter stores if
19656 supports_vec_scatter_store_p.
19657 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
19658 scatter stores too.
19659 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19660 internal_fn_stored_value_index.
19661 (check_load_store_masking): Handle scatter stores too.
19662 (vect_get_store_rhs): Make public.
19663 (vectorizable_call): Use internal_store_fn_p.
19664 (vectorizable_store): Handle scatter store internal functions.
19665 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
19666 when deciding whether the end of the group has been reached.
19667 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
19668 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
19669 (mask_scatter_store<mode>): New insns.
19671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19672 Alan Hayward <alan.hayward@arm.com>
19673 David Sherwood <david.sherwood@arm.com>
19675 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
19676 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
19677 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
19679 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
19680 Use vect_truncate_gather_scatter_offset if we can't treat the
19681 operation as a normal gather load or scatter store.
19682 (get_group_load_store_type): Take the gather_scatter_info
19683 as argument. Try using a gather load or scatter store for
19684 single-element groups.
19685 (get_load_store_type): Update calls to get_group_load_store_type
19686 and vect_use_strided_gather_scatters_p.
19688 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19689 Alan Hayward <alan.hayward@arm.com>
19690 David Sherwood <david.sherwood@arm.com>
19692 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
19693 optional tree argument.
19694 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
19696 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
19697 but continue to use the current value as a fallback.
19698 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
19699 to compare the updates.
19700 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
19701 (get_load_store_type): Use it when handling a strided access.
19702 (vect_get_strided_load_store_ops): New function.
19703 (vect_get_data_ptr_increment): Likewise.
19704 (vectorizable_load): Handle strided gather loads. Always pass
19705 a step to vect_create_data_ref_ptr and bump_vector_ptr.
19707 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19708 Alan Hayward <alan.hayward@arm.com>
19709 David Sherwood <david.sherwood@arm.com>
19711 * doc/md.texi (gather_load@var{m}): Document.
19712 (mask_gather_load@var{m}): Likewise.
19713 * genopinit.c (main): Add supports_vec_gather_load and
19714 supports_vec_gather_load_cached to target_optabs.
19715 * optabs-tree.c (init_tree_optimization_optabs): Use
19716 ggc_cleared_alloc to allocate target_optabs.
19717 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
19718 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
19720 * internal-fn.h (internal_load_fn_p): Declare.
19721 (internal_gather_scatter_fn_p): Likewise.
19722 (internal_fn_mask_index): Likewise.
19723 (internal_gather_scatter_fn_supported_p): Likewise.
19724 * internal-fn.c (gather_load_direct): New macro.
19725 (expand_gather_load_optab_fn): New function.
19726 (direct_gather_load_optab_supported_p): New macro.
19727 (direct_internal_fn_optab): New function.
19728 (internal_load_fn_p): Likewise.
19729 (internal_gather_scatter_fn_p): Likewise.
19730 (internal_fn_mask_index): Likewise.
19731 (internal_gather_scatter_fn_supported_p): Likewise.
19732 * optabs-query.c (supports_at_least_one_mode_p): New function.
19733 (supports_vec_gather_load_p): Likewise.
19734 * optabs-query.h (supports_vec_gather_load_p): Declare.
19735 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
19736 and memory_type field.
19737 (NUM_PATTERNS): Bump to 15.
19738 * tree-vect-data-refs.c: Include internal-fn.h.
19739 (vect_gather_scatter_fn_p): New function.
19740 (vect_describe_gather_scatter_call): Likewise.
19741 (vect_check_gather_scatter): Try using internal functions for
19742 gather loads. Recognize existing calls to a gather load function.
19743 (vect_analyze_data_refs): Consider using gather loads if
19744 supports_vec_gather_load_p.
19745 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
19746 (vect_get_gather_scatter_offset_type): Likewise.
19747 (vect_convert_mask_for_vectype): Likewise.
19748 (vect_add_conversion_to_patterm): Likewise.
19749 (vect_try_gather_scatter_pattern): Likewise.
19750 (vect_recog_gather_scatter_pattern): New pattern recognizer.
19751 (vect_vect_recog_func_ptrs): Add it.
19752 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19753 internal_fn_mask_index and internal_gather_scatter_fn_p.
19754 (check_load_store_masking): Take the gather_scatter_info as an
19755 argument and handle gather loads.
19756 (vect_get_gather_scatter_ops): New function.
19757 (vectorizable_call): Check internal_load_fn_p.
19758 (vectorizable_load): Likewise. Handle gather load internal
19760 (vectorizable_store): Update call to check_load_store_masking.
19761 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
19762 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
19763 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
19764 (aarch64_gather_scale_operand_d): New predicates.
19765 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
19766 (mask_gather_load<mode>): New insns.
19768 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19769 Alan Hayward <alan.hayward@arm.com>
19770 David Sherwood <david.sherwood@arm.com>
19772 * optabs.def (fold_left_plus_optab): New optab.
19773 * doc/md.texi (fold_left_plus_@var{m}): Document.
19774 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
19775 * internal-fn.c (fold_left_direct): Define.
19776 (expand_fold_left_optab_fn): Likewise.
19777 (direct_fold_left_optab_supported_p): Likewise.
19778 * fold-const-call.c (fold_const_fold_left): New function.
19779 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
19780 * tree-parloops.c (valid_reduction_p): New function.
19781 (gather_scalar_reductions): Use it.
19782 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
19783 (vect_finish_replace_stmt): Declare.
19784 * tree-vect-loop.c (fold_left_reduction_fn): New function.
19785 (needs_fold_left_reduction_p): New function, split out from...
19786 (vect_is_simple_reduction): ...here. Accept reductions that
19787 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
19788 (vect_force_simple_reduction): Also store the reduction type in
19789 the assignment's STMT_VINFO_REDUC_TYPE.
19790 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
19791 (merge_with_identity): New function.
19792 (vect_expand_fold_left): Likewise.
19793 (vectorize_fold_left_reduction): Likewise.
19794 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
19795 scalar phi in place for it. Check for target support and reject
19796 cases that would reassociate the operation. Defer the transform
19797 phase to vectorize_fold_left_reduction.
19798 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
19799 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
19800 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
19802 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19804 * tree-if-conv.c (predicate_mem_writes): Remove redundant
19805 call to ifc_temp_var.
19807 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19808 Alan Hayward <alan.hayward@arm.com>
19809 David Sherwood <david.sherwood@arm.com>
19811 * target.def (legitimize_address_displacement): Take the original
19812 offset as a poly_int.
19813 * targhooks.h (default_legitimize_address_displacement): Update
19815 * targhooks.c (default_legitimize_address_displacement): Likewise.
19816 * doc/tm.texi: Regenerate.
19817 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
19818 as an argument, moving assert of ad->disp == ad->disp_term to...
19819 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
19820 Try calling targetm.legitimize_address_displacement before expanding
19821 the address rather than afterwards, and adjust for the new interface.
19822 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
19823 Match the new hook interface. Handle SVE addresses.
19824 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
19825 new hook interface.
19827 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19829 * Makefile.in (OBJS): Add early-remat.o.
19830 * target.def (select_early_remat_modes): New hook.
19831 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
19832 * doc/tm.texi: Regenerate.
19833 * targhooks.h (default_select_early_remat_modes): Declare.
19834 * targhooks.c (default_select_early_remat_modes): New function.
19835 * timevar.def (TV_EARLY_REMAT): New timevar.
19836 * passes.def (pass_early_remat): New pass.
19837 * tree-pass.h (make_pass_early_remat): Declare.
19838 * early-remat.c: New file.
19839 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
19841 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
19843 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19844 Alan Hayward <alan.hayward@arm.com>
19845 David Sherwood <david.sherwood@arm.com>
19847 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
19848 vfm1 with a bound_epilog parameter.
19849 (vect_do_peeling): Update calls accordingly, and move the prologue
19850 call earlier in the function. Treat the base bound_epilog as 0 for
19851 fully-masked loops and retain vf - 1 for other loops. Add 1 to
19852 this base when peeling for gaps.
19853 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
19854 with fully-masked loops.
19855 (vect_estimate_min_profitable_iters): Handle the single peeled
19856 iteration in that case.
19858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19859 Alan Hayward <alan.hayward@arm.com>
19860 David Sherwood <david.sherwood@arm.com>
19862 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
19863 single-element interleaving even if the size is not a power of 2.
19864 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
19865 accesses for single-element interleaving if the group size is
19868 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19869 Alan Hayward <alan.hayward@arm.com>
19870 David Sherwood <david.sherwood@arm.com>
19872 * doc/md.texi (fold_extract_last_@var{m}): Document.
19873 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
19874 * optabs.def (fold_extract_last_optab): New optab.
19875 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
19876 * internal-fn.c (fold_extract_direct): New macro.
19877 (expand_fold_extract_optab_fn): Likewise.
19878 (direct_fold_extract_optab_supported_p): Likewise.
19879 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
19880 * tree-vect-loop.c (vect_model_reduction_cost): Handle
19881 EXTRACT_LAST_REDUCTION.
19882 (get_initial_def_for_reduction): Do not create an initial vector
19883 for EXTRACT_LAST_REDUCTION reductions.
19884 (vectorizable_reduction): Leave the scalar phi in place for
19885 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
19886 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
19887 epilogue code for EXTRACT_LAST_REDUCTION and defer the
19888 transform phase to vectorizable_condition.
19889 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
19891 (vect_finish_stmt_generation): ...here.
19892 (vect_finish_replace_stmt): New function.
19893 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
19894 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
19896 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
19898 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19899 Alan Hayward <alan.hayward@arm.com>
19900 David Sherwood <david.sherwood@arm.com>
19902 * doc/md.texi (extract_last_@var{m}): Document.
19903 * optabs.def (extract_last_optab): New optab.
19904 * internal-fn.def (EXTRACT_LAST): New internal function.
19905 * internal-fn.c (cond_unary_direct): New macro.
19906 (expand_cond_unary_optab_fn): Likewise.
19907 (direct_cond_unary_optab_supported_p): Likewise.
19908 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
19909 loops using EXTRACT_LAST.
19910 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
19911 (extract_last_<mode>): ...this optab.
19912 (vec_extract<mode><Vel>): Update accordingly.
19914 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19915 Alan Hayward <alan.hayward@arm.com>
19916 David Sherwood <david.sherwood@arm.com>
19918 * target.def (empty_mask_is_expensive): New hook.
19919 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
19920 * doc/tm.texi: Regenerate.
19921 * targhooks.h (default_empty_mask_is_expensive): Declare.
19922 * targhooks.c (default_empty_mask_is_expensive): New function.
19923 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
19924 if the target says that empty masks are expensive.
19925 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
19927 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
19929 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19930 Alan Hayward <alan.hayward@arm.com>
19931 David Sherwood <david.sherwood@arm.com>
19933 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
19934 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
19935 (vect_use_loop_mask_for_alignment_p): New function.
19936 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
19937 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
19938 niters_skip argument. Make sure that the first niters_skip elements
19939 of the first iteration are inactive.
19940 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
19941 Update call to vect_set_loop_masks_directly.
19942 (get_misalign_in_elems): New function, split out from...
19943 (vect_gen_prolog_loop_niters): ...here.
19944 (vect_update_init_of_dr): Take a code argument that specifies whether
19945 the adjustment should be added or subtracted.
19946 (vect_update_init_of_drs): Likewise.
19947 (vect_prepare_for_masked_peels): New function.
19948 (vect_do_peeling): Skip prologue peeling if we're using a mask
19949 instead. Update call to vect_update_inits_of_drs.
19950 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19952 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
19953 alignment. Do not include the number of peeled iterations in
19954 the minimum threshold in that case.
19955 (vectorizable_induction): Adjust the start value down by
19956 LOOP_VINFO_MASK_SKIP_NITERS iterations.
19957 (vect_transform_loop): Call vect_prepare_for_masked_peels.
19958 Take the number of skipped iterations into account when calculating
19960 * tree-vect-stmts.c (vect_gen_while_not): New function.
19962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19963 Alan Hayward <alan.hayward@arm.com>
19964 David Sherwood <david.sherwood@arm.com>
19966 * doc/sourcebuild.texi (vect_fully_masked): Document.
19967 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
19968 default value to 0.
19969 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
19971 (vect_analyze_loop_2): ...here. Don't check the vectorization
19972 factor against the number of loop iterations if the loop is
19975 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19976 Alan Hayward <alan.hayward@arm.com>
19977 David Sherwood <david.sherwood@arm.com>
19979 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
19980 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
19981 (dump_groups): Update accordingly.
19982 (iv_use::mem_type): New member variable.
19983 (address_p): New function.
19984 (record_use): Add a mem_type argument and initialize the new
19986 (record_group_use): Add a mem_type argument. Use address_p.
19987 Remove obsolete null checks of base_object. Update call to record_use.
19988 (find_interesting_uses_op): Update call to record_group_use.
19989 (find_interesting_uses_cond): Likewise.
19990 (find_interesting_uses_address): Likewise.
19991 (get_mem_type_for_internal_fn): New function.
19992 (find_address_like_use): Likewise.
19993 (find_interesting_uses_stmt): Try find_address_like_use before
19994 calling find_interesting_uses_op.
19995 (addr_offset_valid_p): Use the iv mem_type field as the type
19996 of the addressed memory.
19997 (add_autoinc_candidates): Likewise.
19998 (get_address_cost): Likewise.
19999 (split_small_address_groups_p): Use address_p.
20000 (split_address_groups): Likewise.
20001 (add_iv_candidate_for_use): Likewise.
20002 (autoinc_possible_for_pair): Likewise.
20003 (rewrite_groups): Likewise.
20004 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
20005 (determine_group_iv_cost): Update after split of USE_ADDRESS.
20006 (get_alias_ptr_type_for_ptr_address): New function.
20007 (rewrite_use_address): Rewrite address uses in calls that were
20008 identified by find_address_like_use.
20010 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20011 Alan Hayward <alan.hayward@arm.com>
20012 David Sherwood <david.sherwood@arm.com>
20014 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
20016 * gimple-expr.h (is_gimple_addressable: Likewise.
20017 * gimple-expr.c (is_gimple_address): Likewise.
20018 * internal-fn.c (expand_call_mem_ref): New function.
20019 (expand_mask_load_optab_fn): Use it.
20020 (expand_mask_store_optab_fn): Likewise.
20022 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20023 Alan Hayward <alan.hayward@arm.com>
20024 David Sherwood <david.sherwood@arm.com>
20026 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
20027 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
20028 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
20029 (cond_umax@var{mode}): Document.
20030 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
20031 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
20032 (cond_umin_optab, cond_umax_optab): New optabs.
20033 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
20034 (COND_IOR, COND_XOR): New internal functions.
20035 * internal-fn.h (get_conditional_internal_fn): Declare.
20036 * internal-fn.c (cond_binary_direct): New macro.
20037 (expand_cond_binary_optab_fn): Likewise.
20038 (direct_cond_binary_optab_supported_p): Likewise.
20039 (get_conditional_internal_fn): New function.
20040 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
20041 Cope with reduction statements that are vectorized as calls rather
20043 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
20044 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
20045 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
20046 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
20047 (UNSPEC_COND_EOR): New unspecs.
20048 (optab): Add mappings for them.
20049 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
20050 (sve_int_op, sve_fp_op): New int attributes.
20052 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20053 Alan Hayward <alan.hayward@arm.com>
20054 David Sherwood <david.sherwood@arm.com>
20056 * optabs.def (while_ult_optab): New optab.
20057 * doc/md.texi (while_ult@var{m}@var{n}): Document.
20058 * internal-fn.def (WHILE_ULT): New internal function.
20059 * internal-fn.h (direct_internal_fn_supported_p): New override
20060 that takes two types as argument.
20061 * internal-fn.c (while_direct): New macro.
20062 (expand_while_optab_fn): New function.
20063 (convert_optab_supported_p): Likewise.
20064 (direct_while_optab_supported_p): New macro.
20065 * wide-int.h (wi::udiv_ceil): New function.
20066 * tree-vectorizer.h (rgroup_masks): New structure.
20067 (vec_loop_masks): New typedef.
20068 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
20069 and fully_masked_p.
20070 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
20071 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
20072 (vect_max_vf): New function.
20073 (slpeel_make_loop_iterate_ntimes): Delete.
20074 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
20075 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
20076 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
20077 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
20078 internal-fn.h, stor-layout.h and optabs-query.h.
20079 (vect_set_loop_mask): New function.
20080 (add_preheader_seq): Likewise.
20081 (add_header_seq): Likewise.
20082 (interleave_supported_p): Likewise.
20083 (vect_maybe_permute_loop_masks): Likewise.
20084 (vect_set_loop_masks_directly): Likewise.
20085 (vect_set_loop_condition_masked): Likewise.
20086 (vect_set_loop_condition_unmasked): New function, split out from
20087 slpeel_make_loop_iterate_ntimes.
20088 (slpeel_make_loop_iterate_ntimes): Rename to..
20089 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
20090 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
20091 (vect_do_peeling): Update call accordingly.
20092 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
20094 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
20095 mask_compare_type, can_fully_mask_p and fully_masked_p.
20096 (release_vec_loop_masks): New function.
20097 (_loop_vec_info): Use it to free the loop masks.
20098 (can_produce_all_loop_masks_p): New function.
20099 (vect_get_max_nscalars_per_iter): Likewise.
20100 (vect_verify_full_masking): Likewise.
20101 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
20102 retries, and free the mask rgroups before retrying. Check loop-wide
20103 reasons for disallowing fully-masked loops. Make the final decision
20104 about whether use a fully-masked loop or not.
20105 (vect_estimate_min_profitable_iters): Do not assume that peeling
20106 for the number of iterations will be needed for fully-masked loops.
20107 (vectorizable_reduction): Disable fully-masked loops.
20108 (vectorizable_live_operation): Likewise.
20109 (vect_halve_mask_nunits): New function.
20110 (vect_double_mask_nunits): Likewise.
20111 (vect_record_loop_mask): Likewise.
20112 (vect_get_loop_mask): Likewise.
20113 (vect_transform_loop): Handle the case in which the final loop
20114 iteration might handle a partial vector. Call vect_set_loop_condition
20115 instead of slpeel_make_loop_iterate_ntimes.
20116 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
20117 (check_load_store_masking): New function.
20118 (prepare_load_store_mask): Likewise.
20119 (vectorizable_store): Handle fully-masked loops.
20120 (vectorizable_load): Likewise.
20121 (supportable_widening_operation): Use vect_halve_mask_nunits for
20123 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
20124 (vect_gen_while): New function.
20125 * config/aarch64/aarch64.md (umax<mode>3): New expander.
20126 (aarch64_uqdec<mode>): New insn.
20128 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20129 Alan Hayward <alan.hayward@arm.com>
20130 David Sherwood <david.sherwood@arm.com>
20132 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
20133 (reduc_xor_scal_optab): New optabs.
20134 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
20135 (reduc_xor_scal_@var{m}): Document.
20136 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
20137 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
20138 internal functions.
20139 * fold-const-call.c (fold_const_call): Handle them.
20140 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
20141 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
20142 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
20143 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
20144 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
20145 (UNSPEC_XORV): New unspecs.
20146 (optab): Add entries for them.
20147 (BITWISEV): New int iterator.
20148 (bit_reduc_op): New int attributes.
20150 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20151 Alan Hayward <alan.hayward@arm.com>
20152 David Sherwood <david.sherwood@arm.com>
20154 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
20155 * internal-fn.def (VEC_SHL_INSERT): New internal function.
20156 * optabs.def (vec_shl_insert_optab): New optab.
20157 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
20158 (duplicate_and_interleave): Likewise.
20159 * tree-vect-loop.c: Include internal-fn.h.
20160 (neutral_op_for_slp_reduction): New function, split out from
20161 get_initial_defs_for_reduction.
20162 (get_initial_def_for_reduction): Handle option 2 for variable-length
20163 vectors by loading the neutral value into a vector and then shifting
20164 the initial value into element 0.
20165 (get_initial_defs_for_reduction): Replace the code argument with
20166 the neutral value calculated by neutral_op_for_slp_reduction.
20167 Use gimple_build_vector for constant-length vectors.
20168 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
20169 but the first group_size elements have a neutral value.
20170 Use duplicate_and_interleave otherwise.
20171 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
20172 Update call to get_initial_defs_for_reduction. Handle SLP
20173 reductions for variable-length vectors by creating one vector
20174 result for each scalar result, with the elements associated
20175 with other scalar results stubbed out with the neutral value.
20176 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
20177 Require IFN_VEC_SHL_INSERT for double reductions on
20178 variable-length vectors, or SLP reductions that have
20179 a neutral value. Require can_duplicate_and_interleave_p
20180 support for variable-length unchained SLP reductions if there
20181 is no neutral value, such as for MIN/MAX reductions. Also require
20182 the number of vector elements to be a multiple of the number of
20183 SLP statements when doing variable-length unchained SLP reductions.
20184 Update call to vect_create_epilog_for_reduction.
20185 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
20186 and remove initial values.
20187 (duplicate_and_interleave): Make public.
20188 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
20189 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
20191 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20192 Alan Hayward <alan.hayward@arm.com>
20193 David Sherwood <david.sherwood@arm.com>
20195 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
20196 (can_duplicate_and_interleave_p): New function.
20197 (vect_get_and_check_slp_defs): Take the vector of statements
20198 rather than just the current one. Remove excess parentheses.
20199 Restriction rejectinon of vect_constant_def and vect_external_def
20200 for variable-length vectors to boolean types, or types for which
20201 can_duplicate_and_interleave_p is false.
20202 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
20203 (duplicate_and_interleave): New function.
20204 (vect_get_constant_vectors): Use gimple_build_vector for
20205 constant-length vectors and suitable variable-length constant
20206 vectors. Use duplicate_and_interleave for other variable-length
20207 vectors. Don't defer the update when inserting new statements.
20209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20210 Alan Hayward <alan.hayward@arm.com>
20211 David Sherwood <david.sherwood@arm.com>
20213 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
20214 min_profitable_iters doesn't go negative.
20216 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20217 Alan Hayward <alan.hayward@arm.com>
20218 David Sherwood <david.sherwood@arm.com>
20220 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
20221 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
20222 * optabs.def (vec_mask_load_lanes_optab): New optab.
20223 (vec_mask_store_lanes_optab): Likewise.
20224 * internal-fn.def (MASK_LOAD_LANES): New internal function.
20225 (MASK_STORE_LANES): Likewise.
20226 * internal-fn.c (mask_load_lanes_direct): New macro.
20227 (mask_store_lanes_direct): Likewise.
20228 (expand_mask_load_optab_fn): Handle masked operations.
20229 (expand_mask_load_lanes_optab_fn): New macro.
20230 (expand_mask_store_optab_fn): Handle masked operations.
20231 (expand_mask_store_lanes_optab_fn): New macro.
20232 (direct_mask_load_lanes_optab_supported_p): Likewise.
20233 (direct_mask_store_lanes_optab_supported_p): Likewise.
20234 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
20236 (vect_load_lanes_supported): Likewise.
20237 * tree-vect-data-refs.c (strip_conversion): New function.
20238 (can_group_stmts_p): Likewise.
20239 (vect_analyze_data_ref_accesses): Use it instead of checking
20240 for a pair of assignments.
20241 (vect_store_lanes_supported): Take a masked_p parameter.
20242 (vect_load_lanes_supported): Likewise.
20243 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
20244 vect_store_lanes_supported and vect_load_lanes_supported.
20245 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
20246 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
20247 parameter. Don't allow gaps for masked accesses.
20248 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
20249 and vect_load_lanes_supported.
20250 (get_load_store_type): Take a masked_p parameter and update
20251 call to get_group_load_store_type.
20252 (vectorizable_store): Update call to get_load_store_type.
20253 Handle IFN_MASK_STORE_LANES.
20254 (vectorizable_load): Update call to get_load_store_type.
20255 Handle IFN_MASK_LOAD_LANES.
20257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20258 Alan Hayward <alan.hayward@arm.com>
20259 David Sherwood <david.sherwood@arm.com>
20261 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
20263 * config/aarch64/aarch64-protos.h
20264 (aarch64_sve_struct_memory_operand_p): Declare.
20265 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
20266 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
20267 (VPRED, vpred): Handle SVE structure modes.
20268 * config/aarch64/constraints.md (Utx): New constraint.
20269 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
20270 (aarch64_sve_struct_nonimmediate_operand): New predicates.
20271 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
20272 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
20273 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
20274 structure modes. Split into pieces after RA.
20275 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
20276 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
20278 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
20279 SVE structure modes.
20280 (aarch64_classify_address): Likewise.
20281 (sizetochar): Move earlier in file.
20282 (aarch64_print_operand): Handle SVE register lists.
20283 (aarch64_array_mode): New function.
20284 (aarch64_sve_struct_memory_operand_p): Likewise.
20285 (TARGET_ARRAY_MODE): Redefine.
20287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20288 Alan Hayward <alan.hayward@arm.com>
20289 David Sherwood <david.sherwood@arm.com>
20291 * target.def (array_mode): New target hook.
20292 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
20293 * doc/tm.texi: Regenerate.
20294 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
20295 * hooks.c (hook_optmode_mode_uhwi_none): New function.
20296 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
20297 targetm.array_mode.
20298 * stor-layout.c (mode_for_array): Likewise. Support polynomial
20301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20302 Alan Hayward <alan.hayward@arm.com>
20303 David Sherwood <david.sherwood@arm.com>
20305 * fold-const.c (fold_binary_loc): Check the argument types
20306 rather than the result type when testing for a vector operation.
20308 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20310 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
20311 * doc/tm.texi: Regenerate.
20313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20314 Alan Hayward <alan.hayward@arm.com>
20315 David Sherwood <david.sherwood@arm.com>
20317 * doc/invoke.texi (-msve-vector-bits=): Document new option.
20318 (sve): Document new AArch64 extension.
20319 * doc/md.texi (w): Extend the description of the AArch64
20320 constraint to include SVE vectors.
20321 (Upl, Upa): Document new AArch64 predicate constraints.
20322 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
20324 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
20325 (msve-vector-bits=): New option.
20326 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
20327 SVE when these are disabled.
20328 (sve): New extension.
20329 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
20330 modes. Adjust their number of units based on aarch64_sve_vg.
20331 (MAX_BITSIZE_MODE_ANY_MODE): Define.
20332 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
20333 aarch64_addr_query_type.
20334 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
20335 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
20336 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
20337 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
20338 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
20339 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
20340 (aarch64_simd_imm_zero_p): Delete.
20341 (aarch64_check_zero_based_sve_index_immediate): Declare.
20342 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
20343 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
20344 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
20345 (aarch64_sve_float_mul_immediate_p): Likewise.
20346 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
20347 rather than an rtx.
20348 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
20349 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
20350 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
20351 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
20352 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
20353 (aarch64_regmode_natural_size): Likewise.
20354 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
20355 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
20357 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
20358 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
20359 for VG and the SVE predicate registers.
20360 (V_ALIASES): Add a "z"-prefixed alias.
20361 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
20362 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
20363 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
20364 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
20365 (REG_CLASS_NAMES): Add entries for them.
20366 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
20367 and the predicate registers.
20368 (aarch64_sve_vg): Declare.
20369 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
20370 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
20371 (REGMODE_NATURAL_SIZE): Define.
20372 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
20374 * config/aarch64/aarch64.c: Include cfgrtl.h.
20375 (simd_immediate_info): Add a constructor for series vectors,
20376 and an associated step field.
20377 (aarch64_sve_vg): New variable.
20378 (aarch64_dbx_register_number): Handle VG and the predicate registers.
20379 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
20380 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
20381 (VEC_ANY_DATA, VEC_STRUCT): New constants.
20382 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
20383 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
20384 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
20385 (aarch64_get_mask_mode): New functions.
20386 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
20387 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
20388 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
20389 predicate modes and predicate registers. Explicitly restrict
20390 GPRs to modes of 16 bytes or smaller. Only allow FP registers
20391 to store a vector mode if it is recognized by
20392 aarch64_classify_vector_mode.
20393 (aarch64_regmode_natural_size): New function.
20394 (aarch64_hard_regno_caller_save_mode): Return the original mode
20396 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
20397 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
20398 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
20399 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
20401 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
20402 does not overlap dest if the function is frame-related. Handle
20404 (aarch64_split_add_offset): New function.
20405 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
20406 them aarch64_add_offset.
20407 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
20408 and update call to aarch64_sub_sp.
20409 (aarch64_add_cfa_expression): New function.
20410 (aarch64_expand_prologue): Pass extra temporary registers to the
20411 functions above. Handle the case in which we need to emit new
20412 DW_CFA_expressions for registers that were originally saved
20413 relative to the stack pointer, but now have to be expressed
20414 relative to the frame pointer.
20415 (aarch64_output_mi_thunk): Pass extra temporary registers to the
20417 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
20418 IP0 and IP1 values for SVE frames.
20419 (aarch64_expand_vec_series): New function.
20420 (aarch64_expand_sve_widened_duplicate): Likewise.
20421 (aarch64_expand_sve_const_vector): Likewise.
20422 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
20423 Handle SVE constants. Use emit_move_insn to move a force_const_mem
20424 into the register, rather than emitting a SET directly.
20425 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
20426 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
20427 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
20428 (offset_9bit_signed_scaled_p): New functions.
20429 (aarch64_replicate_bitmask_imm): New function.
20430 (aarch64_bitmask_imm): Use it.
20431 (aarch64_cannot_force_const_mem): Reject expressions involving
20432 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
20433 (aarch64_classify_index): Handle SVE indices, by requiring
20434 a plain register index with a scale that matches the element size.
20435 (aarch64_classify_address): Handle SVE addresses. Assert that
20436 the mode of the address is VOIDmode or an integer mode.
20437 Update call to aarch64_classify_symbol.
20438 (aarch64_classify_symbolic_expression): Update call to
20439 aarch64_classify_symbol.
20440 (aarch64_const_vec_all_in_range_p): New function.
20441 (aarch64_print_vector_float_operand): Likewise.
20442 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
20443 "vN" for FP registers with SVE modes. Handle (const ...) vectors
20444 and the FP immediates 1.0 and 0.5.
20445 (aarch64_print_address_internal): Handle SVE addresses.
20446 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
20447 (aarch64_regno_regclass): Handle predicate registers.
20448 (aarch64_secondary_reload): Handle big-endian reloads of SVE
20450 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
20451 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
20452 (aarch64_convert_sve_vector_bits): New function.
20453 (aarch64_override_options): Use it to handle -msve-vector-bits=.
20454 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
20455 rather than an rtx.
20456 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
20457 Handle SVE vector and predicate modes. Accept VL-based constants
20458 that need only one temporary register, and VL offsets that require
20459 no temporary registers.
20460 (aarch64_conditional_register_usage): Mark the predicate registers
20461 as fixed if SVE isn't available.
20462 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
20463 Return true for SVE vector and predicate modes.
20464 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
20465 rather than an unsigned int. Handle SVE modes.
20466 (aarch64_preferred_simd_mode): Update call accordingly. Handle
20468 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
20470 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
20471 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
20472 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
20473 (aarch64_sve_float_mul_immediate_p): New functions.
20474 (aarch64_sve_valid_immediate): New function.
20475 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
20476 Explicitly reject structure modes. Check for INDEX constants.
20477 Handle PTRUE and PFALSE constants.
20478 (aarch64_check_zero_based_sve_index_immediate): New function.
20479 (aarch64_simd_imm_zero_p): Delete.
20480 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
20481 vector modes. Accept constants in the range of CNT[BHWD].
20482 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
20483 ask for an Advanced SIMD mode.
20484 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
20485 (aarch64_simd_vector_alignment): Handle SVE predicates.
20486 (aarch64_vectorize_preferred_vector_alignment): New function.
20487 (aarch64_simd_vector_alignment_reachable): Use it instead of
20489 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
20490 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
20492 (MAX_VECT_LEN): Delete.
20493 (expand_vec_perm_d): Add a vec_flags field.
20494 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
20495 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
20496 (aarch64_evpc_ext): Don't apply a big-endian lane correction
20498 (aarch64_evpc_rev): Rename to...
20499 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
20500 (aarch64_evpc_rev_global): New function.
20501 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
20502 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
20504 (aarch64_evpc_sve_tbl): New function.
20505 (aarch64_expand_vec_perm_const_1): Update after rename of
20506 aarch64_evpc_rev. Handle SVE permutes too, trying
20507 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
20508 than aarch64_evpc_tbl.
20509 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
20510 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
20511 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
20512 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
20513 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
20514 (aarch64_expand_sve_vcond): New functions.
20515 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
20516 of aarch64_vector_mode_p.
20517 (aarch64_dwarf_poly_indeterminate_value): New function.
20518 (aarch64_compute_pressure_classes): Likewise.
20519 (aarch64_can_change_mode_class): Likewise.
20520 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
20521 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
20522 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
20523 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
20524 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
20525 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
20526 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
20527 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
20529 (Dn, Dl, Dr): Accept const as well as const_vector.
20530 (Dz): Likewise. Compare against CONST0_RTX.
20531 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
20532 of "vector" where appropriate.
20533 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
20534 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
20535 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
20536 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
20537 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
20538 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
20539 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
20540 (v_int_equiv): Extend to SVE modes.
20541 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
20543 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
20544 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
20545 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
20546 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
20547 (SVE_COND_FP_CMP): New int iterators.
20548 (perm_hilo): Handle the new unpack unspecs.
20549 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
20551 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
20552 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
20553 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
20554 (aarch64_equality_operator, aarch64_constant_vector_operand)
20555 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
20556 (aarch64_sve_nonimmediate_operand): Likewise.
20557 (aarch64_sve_general_operand): Likewise.
20558 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
20559 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
20560 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
20561 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
20562 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
20563 (aarch64_sve_float_arith_immediate): Likewise.
20564 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
20565 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
20566 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
20567 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
20568 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
20569 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
20570 (aarch64_sve_float_arith_operand): Likewise.
20571 (aarch64_sve_float_arith_with_sub_operand): Likewise.
20572 (aarch64_sve_float_mul_operand): Likewise.
20573 (aarch64_sve_vec_perm_operand): Likewise.
20574 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
20575 (aarch64_mov_operand): Accept const_poly_int and const_vector.
20576 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
20577 as well as const_vector.
20578 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
20579 in file. Use CONST0_RTX and CONSTM1_RTX.
20580 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
20581 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
20582 Use aarch64_simd_imm_zero.
20583 * config/aarch64/aarch64-sve.md: New file.
20584 * config/aarch64/aarch64.md: Include it.
20585 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
20586 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
20587 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
20588 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
20589 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
20590 (sve): New attribute.
20591 (enabled): Disable instructions with the sve attribute unless
20593 (movqi, movhi): Pass CONST_POLY_INT operaneds through
20594 aarch64_expand_mov_immediate.
20595 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
20596 CNT[BHSD] immediates.
20597 (movti): Split CONST_POLY_INT moves into two halves.
20598 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
20599 Split additions that need a temporary here if the destination
20600 is the stack pointer.
20601 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
20602 (*add<mode>3_poly_1): New instruction.
20603 (set_clobber_cc): New expander.
20605 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20607 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
20608 parameter and use it instead of GET_MODE_SIZE (innermode). Use
20609 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
20610 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
20611 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
20612 Change innermode from fixed_mode_size to machine_mode.
20613 (simplify_subreg): Update call accordingly. Handle a constant-sized
20614 subreg of a variable-length CONST_VECTOR.
20616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20617 Alan Hayward <alan.hayward@arm.com>
20618 David Sherwood <david.sherwood@arm.com>
20620 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
20621 (add_offset_to_base): New function, split out from...
20622 (create_mem_ref): ...here. When handling a scale other than 1,
20623 check first whether the address is valid without the offset.
20624 Add it into the base if so, leaving the index and scale as-is.
20626 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20629 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
20630 fold_for_warn before checking if arg2 is INTEGER_CST.
20632 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
20634 * config/rs6000/predicates.md (load_multiple_operation): Delete.
20635 (store_multiple_operation): Delete.
20636 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
20637 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
20638 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
20639 guarded by TARGET_STRING.
20640 (rs6000_output_load_multiple): Delete.
20641 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
20642 OPTION_MASK_STRING / TARGET_STRING handling.
20643 (print_operand) <'N', 'O'>: Add comment that these are unused now.
20644 (const rs6000_opt_masks) <"string">: Change mask to 0.
20645 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
20646 (MASK_STRING): Delete.
20647 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
20649 (load_multiple): Delete.
20656 (store_multiple): Delete.
20663 (movmemsi_8reg): Delete.
20664 (corresponding unnamed define_insn): Delete.
20665 (movmemsi_6reg): Delete.
20666 (corresponding unnamed define_insn): Delete.
20667 (movmemsi_4reg): Delete.
20668 (corresponding unnamed define_insn): Delete.
20669 (movmemsi_2reg): Delete.
20670 (corresponding unnamed define_insn): Delete.
20671 (movmemsi_1reg): Delete.
20672 (corresponding unnamed define_insn): Delete.
20673 * config/rs6000/rs6000.opt (mno-string): New.
20674 (mstring): Replace by deprecation warning stub.
20675 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
20677 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20679 * regrename.c (regrename_do_replace): If replacing the same
20680 reg multiple times, try to reuse last created gen_raw_REG.
20683 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
20684 main to workaround a bug in GDB.
20686 2018-01-12 Tom de Vries <tom@codesourcery.com>
20689 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
20691 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
20693 PR rtl-optimization/80481
20694 * ira-color.c (get_cap_member): New function.
20695 (allocnos_conflict_by_live_ranges_p): Use it.
20696 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
20697 (setup_slot_coalesced_allocno_live_ranges): Ditto.
20699 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
20702 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
20703 (*saddl_se_1): Ditto.
20704 (*ssubsi_1): Ditto.
20705 (*ssubl_se_1): Ditto.
20707 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20709 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
20710 rather than wi::to_widest for DR_INITs.
20711 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
20712 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
20713 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
20715 (vect_analyze_group_access_1): Note that here.
20717 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20719 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
20720 polynomial type sizes.
20722 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20724 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
20725 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
20726 (gimple_add_tmp_var): Likewise.
20728 2018-01-12 Martin Liska <mliska@suse.cz>
20730 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
20731 (gimple_alloc_sizes): Likewise.
20732 (dump_gimple_statistics): Use PRIu64 in printf format.
20733 * gimple.h: Change uint64_t to int.
20735 2018-01-12 Martin Liska <mliska@suse.cz>
20737 * tree-core.h: Use uint64_t instead of int.
20738 * tree.c (tree_node_counts): Likewise.
20739 (tree_node_sizes): Likewise.
20740 (dump_tree_statistics): Use PRIu64 in printf format.
20742 2018-01-12 Martin Liska <mliska@suse.cz>
20744 * Makefile.in: As qsort_chk is implemented in vec.c, add
20745 vec.o to linkage of gencfn-macros.
20746 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
20747 passing the info to record_node_allocation_statistics.
20748 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
20750 * ggc-common.c (struct ggc_usage): Add operator== and use
20751 it in operator< and compare function.
20752 * mem-stats.h (struct mem_usage): Likewise.
20753 * vec.c (struct vec_usage): Remove operator< and compare
20754 function. Can be simply inherited.
20756 2018-01-12 Martin Jambor <mjambor@suse.cz>
20759 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
20760 * tree-ssa-math-opts.c: Include domwalk.h.
20761 (convert_mult_to_fma_1): New function.
20762 (fma_transformation_info): New type.
20763 (fma_deferring_state): Likewise.
20764 (cancel_fma_deferring): New function.
20765 (result_of_phi): Likewise.
20766 (last_fma_candidate_feeds_initial_phi): Likewise.
20767 (convert_mult_to_fma): Added deferring logic, split actual
20768 transformation to convert_mult_to_fma_1.
20769 (math_opts_dom_walker): New type.
20770 (math_opts_dom_walker::after_dom_children): New method, body moved
20771 here from pass_optimize_widening_mul::execute, added deferring logic
20773 (pass_optimize_widening_mul::execute): Moved most of code to
20774 math_opts_dom_walker::after_dom_children.
20775 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
20776 * config/i386/i386.c (ix86_option_override_internal): Added
20777 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
20779 2018-01-12 Richard Biener <rguenther@suse.de>
20782 * dwarf2out.c (gen_variable_die): Do not reset old_die for
20783 inline instance vars.
20785 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
20788 * config/rx/rx.c (rx_is_restricted_memory_address):
20789 Handle SUBREG case.
20791 2018-01-12 Richard Biener <rguenther@suse.de>
20793 PR tree-optimization/80846
20794 * target.def (split_reduction): New target hook.
20795 * targhooks.c (default_split_reduction): New function.
20796 * targhooks.h (default_split_reduction): Declare.
20797 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
20798 target requests first reduce vectors by combining low and high
20800 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
20801 (get_vectype_for_scalar_type_and_size): Export.
20802 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
20803 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
20804 * doc/tm.texi: Regenerate.
20805 * config/i386/i386.c (ix86_split_reduction): Implement
20806 TARGET_VECTORIZE_SPLIT_REDUCTION.
20808 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20811 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
20812 in PIC mode except for TARGET_VXWORKS_RTP.
20813 * config/sparc/sparc.c: Include cfgrtl.h.
20814 (TARGET_INIT_PIC_REG): Define.
20815 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
20816 (sparc_pic_register_p): New predicate.
20817 (sparc_legitimate_address_p): Use it.
20818 (sparc_legitimize_pic_address): Likewise.
20819 (sparc_delegitimize_address): Likewise.
20820 (sparc_mode_dependent_address_p): Likewise.
20821 (gen_load_pcrel_sym): Remove 4th parameter.
20822 (load_got_register): Adjust call to above. Remove obsolete stuff.
20823 (sparc_expand_prologue): Do not call load_got_register here.
20824 (sparc_flat_expand_prologue): Likewise.
20825 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
20826 (sparc_use_pseudo_pic_reg): New function.
20827 (sparc_init_pic_reg): Likewise.
20828 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20829 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
20831 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
20833 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
20834 Add item for branch_cost.
20836 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20838 PR rtl-optimization/83565
20839 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
20840 not extend the result to a larger mode for rotate operations.
20841 (num_sign_bit_copies1): Likewise.
20843 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20846 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
20848 Use values-Xc.o for -pedantic.
20849 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
20851 2018-01-12 Martin Liska <mliska@suse.cz>
20854 * ipa-devirt.c (final_warning_record::grow_type_warnings):
20856 (possible_polymorphic_call_targets): Use it.
20857 (ipa_devirt): Likewise.
20859 2018-01-12 Martin Liska <mliska@suse.cz>
20861 * profile-count.h (enum profile_quality): Use 0 as invalid
20862 enum value of profile_quality.
20864 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
20866 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
20867 -mext-string options.
20869 2018-01-12 Richard Biener <rguenther@suse.de>
20871 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
20872 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
20873 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
20875 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
20877 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
20879 * configure.ac (--with-long-double-format): Add support for the
20880 configuration option to change the default long double format on
20882 * config.gcc (powerpc*-linux*-*): Likewise.
20883 * configure: Regenerate.
20884 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
20885 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
20886 used without modification.
20888 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20890 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
20891 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
20892 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
20893 MISC_BUILTIN_SPEC_BARRIER.
20894 (rs6000_init_builtins): Likewise.
20895 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
20897 (speculation_barrier): New define_insn.
20898 * doc/extend.texi: Document __builtin_speculation_barrier.
20900 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20903 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
20904 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
20905 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
20907 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
20908 integral modes instead of "ss" and "sd".
20909 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
20910 vectors with 32-bit and 64-bit elements.
20911 (vecdupssescalarmodesuffix): New mode attribute.
20912 (vec_dup<mode>): Use it.
20914 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
20917 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
20918 frame if argument is passed on stack.
20920 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20923 * ree.c (combine_reaching_defs): Optimize also
20924 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
20925 reg2=any_extend(exp); reg1=reg2;, formatting fix.
20927 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20929 PR middle-end/83189
20930 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
20932 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20934 PR middle-end/83718
20935 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
20936 after they are computed.
20938 2018-01-11 Bin Cheng <bin.cheng@arm.com>
20940 PR tree-optimization/83695
20941 * gimple-loop-linterchange.cc
20942 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
20943 reset cached scev information after interchange.
20944 (pass_linterchange::execute): Remove call to scev_reset_htab.
20946 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20948 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
20949 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
20950 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
20951 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
20952 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
20953 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
20954 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
20955 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
20956 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
20957 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
20958 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
20959 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
20960 (V_lane_reg): Likewise.
20961 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
20963 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
20964 (vfmal_lane_low<mode>_intrinsic,
20965 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
20966 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
20967 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
20968 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
20969 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
20970 vfmsl_lane_high<mode>_intrinsic): New define_insns.
20972 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20974 * config/arm/arm-cpus.in (fp16fml): New feature.
20975 (ALL_SIMD): Add fp16fml.
20976 (armv8.2-a): Add fp16fml as an option.
20977 (armv8.3-a): Likewise.
20978 (armv8.4-a): Add fp16fml as part of fp16.
20979 * config/arm/arm.h (TARGET_FP16FML): Define.
20980 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
20982 * config/arm/arm-modes.def (V2HF): Define.
20983 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
20984 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
20985 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
20986 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
20987 vfmsl_low, vfmsl_high): New set of builtins.
20988 * config/arm/iterators.md (PLUSMINUS): New code iterator.
20989 (vfml_op): New code attribute.
20990 (VFMLHALVES): New int iterator.
20991 (VFML, VFMLSEL): New mode attributes.
20992 (V_reg): Define mapping for V2HF.
20993 (V_hi, V_lo): New mode attributes.
20994 (VF_constraint): Likewise.
20995 (vfml_half, vfml_half_selector): New int attributes.
20996 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
20998 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
20999 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
21001 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
21002 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
21003 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
21004 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
21006 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
21007 Document new effective target and option set.
21009 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21011 * config/arm/arm-cpus.in (armv8_4): New feature.
21012 (ARMv8_4a): New fgroup.
21013 (armv8.4-a): New arch.
21014 * config/arm/arm-tables.opt: Regenerate.
21015 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
21016 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
21017 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
21018 Add matching rules for -march=armv8.4-a and extensions.
21019 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
21021 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
21024 * config/rx/rx.md (BW): New mode attribute.
21025 (sync_lock_test_and_setsi): Add mode suffix to insn output.
21027 2018-01-11 Richard Biener <rguenther@suse.de>
21029 PR tree-optimization/83435
21030 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
21031 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
21032 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
21034 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21035 Alan Hayward <alan.hayward@arm.com>
21036 David Sherwood <david.sherwood@arm.com>
21038 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
21040 (aarch64_classify_address): Initialize it. Track polynomial offsets.
21041 (aarch64_print_address_internal): Use it to check for a zero offset.
21043 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21044 Alan Hayward <alan.hayward@arm.com>
21045 David Sherwood <david.sherwood@arm.com>
21047 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
21048 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
21049 Return a poly_int64 rather than a HOST_WIDE_INT.
21050 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
21051 rather than a HOST_WIDE_INT.
21052 * config/aarch64/aarch64.h (aarch64_frame): Protect with
21053 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
21054 hard_fp_offset, frame_size, initial_adjust, callee_offset and
21055 final_offset from HOST_WIDE_INT to poly_int64.
21056 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
21057 to_constant when getting the number of units in an Advanced SIMD
21059 (aarch64_builtin_vectorized_function): Check for a constant number
21061 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
21063 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
21064 attribute instead of GET_MODE_NUNITS.
21065 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
21066 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
21067 GET_MODE_SIZE for fixed-size registers.
21068 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
21069 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
21070 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
21071 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
21072 (aarch64_print_operand, aarch64_print_address_internal)
21073 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
21074 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
21075 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
21076 Handle polynomial GET_MODE_SIZE.
21077 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
21078 wider than SImode without modification.
21079 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
21080 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
21081 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
21082 passing and returning SVE modes.
21083 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
21084 rather than GEN_INT.
21085 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
21086 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
21087 (aarch64_allocate_and_probe_stack_space): Likewise.
21088 (aarch64_layout_frame): Cope with polynomial offsets.
21089 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
21090 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
21091 polynomial offsets.
21092 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
21093 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
21094 poly_int64 rather than a HOST_WIDE_INT.
21095 (aarch64_get_separate_components, aarch64_process_components)
21096 (aarch64_expand_prologue, aarch64_expand_epilogue)
21097 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
21098 (aarch64_anchor_offset): New function, split out from...
21099 (aarch64_legitimize_address): ...here.
21100 (aarch64_builtin_vectorization_cost): Handle polynomial
21101 TYPE_VECTOR_SUBPARTS.
21102 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
21104 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
21105 number of elements from the PARALLEL rather than the mode.
21106 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
21107 rather than GET_MODE_BITSIZE.
21108 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
21109 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
21110 (aarch64_expand_vec_perm_const_1): Handle polynomial
21111 d->perm.length () and d->perm elements.
21112 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
21113 Apply to_constant to d->perm elements.
21114 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
21115 polynomial CONST_VECTOR_NUNITS.
21116 (aarch64_move_pointer): Take amount as a poly_int64 rather
21118 (aarch64_progress_pointer): Avoid temporary variable.
21119 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
21120 the mode attribute instead of GET_MODE.
21122 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21123 Alan Hayward <alan.hayward@arm.com>
21124 David Sherwood <david.sherwood@arm.com>
21126 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
21127 x exists before using it.
21128 (aarch64_add_constant_internal): Rename to...
21129 (aarch64_add_offset_1): ...this. Replace regnum with separate
21130 src and dest rtxes. Handle the case in which they're different,
21131 including when the offset is zero. Replace scratchreg with an rtx.
21132 Use 2 additions if there is no spare register into which we can
21133 move a 16-bit constant.
21134 (aarch64_add_constant): Delete.
21135 (aarch64_add_offset): Replace reg with separate src and dest
21136 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
21137 Use aarch64_add_offset_1.
21138 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
21139 an rtx rather than an int. Take the delta as a poly_int64
21140 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
21141 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
21142 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
21143 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
21144 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
21145 and aarch64_add_sp.
21146 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
21147 aarch64_add_constant.
21149 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21151 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
21152 Use scalar_float_mode.
21154 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21156 * config/aarch64/aarch64-simd.md
21157 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
21158 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
21159 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
21160 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
21161 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
21162 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
21163 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
21164 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
21165 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
21166 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
21168 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21171 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
21172 targ_options->x_arm_arch_string is non NULL.
21174 2018-01-11 Tamar Christina <tamar.christina@arm.com>
21176 * config/aarch64/aarch64.h
21177 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
21179 2018-01-11 Sudakshina Das <sudi.das@arm.com>
21182 * expmed.c (emit_store_flag_force): Swap if const op0
21183 and change VOIDmode to mode of op0.
21185 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21187 PR rtl-optimization/83761
21188 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
21189 than bytes to mode_for_size.
21191 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21193 PR middle-end/83189
21194 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
21195 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
21198 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21200 PR middle-end/83575
21201 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
21202 when in layout mode.
21203 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
21204 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
21207 2018-01-10 Michael Collison <michael.collison@arm.com>
21209 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
21210 * config/aarch64/aarch64-option-extension.def: Add
21211 AARCH64_OPT_EXTENSION of 'fp16fml'.
21212 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21213 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
21214 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
21215 * config/aarch64/constraints.md (Ui7): New constraint.
21216 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
21217 (VFMLA_SEL_W): Ditto.
21220 (VFMLA16_LOW): New int iterator.
21221 (VFMLA16_HIGH): Ditto.
21222 (UNSPEC_FMLAL): New unspec.
21223 (UNSPEC_FMLSL): Ditto.
21224 (UNSPEC_FMLAL2): Ditto.
21225 (UNSPEC_FMLSL2): Ditto.
21226 (f16mac): New code attribute.
21227 * config/aarch64/aarch64-simd-builtins.def
21228 (aarch64_fmlal_lowv2sf): Ditto.
21229 (aarch64_fmlsl_lowv2sf): Ditto.
21230 (aarch64_fmlalq_lowv4sf): Ditto.
21231 (aarch64_fmlslq_lowv4sf): Ditto.
21232 (aarch64_fmlal_highv2sf): Ditto.
21233 (aarch64_fmlsl_highv2sf): Ditto.
21234 (aarch64_fmlalq_highv4sf): Ditto.
21235 (aarch64_fmlslq_highv4sf): Ditto.
21236 (aarch64_fmlal_lane_lowv2sf): Ditto.
21237 (aarch64_fmlsl_lane_lowv2sf): Ditto.
21238 (aarch64_fmlal_laneq_lowv2sf): Ditto.
21239 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
21240 (aarch64_fmlalq_lane_lowv4sf): Ditto.
21241 (aarch64_fmlsl_lane_lowv4sf): Ditto.
21242 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
21243 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
21244 (aarch64_fmlal_lane_highv2sf): Ditto.
21245 (aarch64_fmlsl_lane_highv2sf): Ditto.
21246 (aarch64_fmlal_laneq_highv2sf): Ditto.
21247 (aarch64_fmlsl_laneq_highv2sf): Ditto.
21248 (aarch64_fmlalq_lane_highv4sf): Ditto.
21249 (aarch64_fmlsl_lane_highv4sf): Ditto.
21250 (aarch64_fmlalq_laneq_highv4sf): Ditto.
21251 (aarch64_fmlsl_laneq_highv4sf): Ditto.
21252 * config/aarch64/aarch64-simd.md:
21253 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
21254 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
21255 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
21256 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
21257 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
21258 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
21259 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
21260 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
21261 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
21262 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
21263 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
21264 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
21265 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
21266 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
21267 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
21268 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
21269 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
21270 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
21271 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
21272 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
21273 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
21274 (vfmlsl_low_u32): Ditto.
21275 (vfmlalq_low_u32): Ditto.
21276 (vfmlslq_low_u32): Ditto.
21277 (vfmlal_high_u32): Ditto.
21278 (vfmlsl_high_u32): Ditto.
21279 (vfmlalq_high_u32): Ditto.
21280 (vfmlslq_high_u32): Ditto.
21281 (vfmlal_lane_low_u32): Ditto.
21282 (vfmlsl_lane_low_u32): Ditto.
21283 (vfmlal_laneq_low_u32): Ditto.
21284 (vfmlsl_laneq_low_u32): Ditto.
21285 (vfmlalq_lane_low_u32): Ditto.
21286 (vfmlslq_lane_low_u32): Ditto.
21287 (vfmlalq_laneq_low_u32): Ditto.
21288 (vfmlslq_laneq_low_u32): Ditto.
21289 (vfmlal_lane_high_u32): Ditto.
21290 (vfmlsl_lane_high_u32): Ditto.
21291 (vfmlal_laneq_high_u32): Ditto.
21292 (vfmlsl_laneq_high_u32): Ditto.
21293 (vfmlalq_lane_high_u32): Ditto.
21294 (vfmlslq_lane_high_u32): Ditto.
21295 (vfmlalq_laneq_high_u32): Ditto.
21296 (vfmlslq_laneq_high_u32): Ditto.
21297 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
21298 (AARCH64_FL_FOR_ARCH8_4): New.
21299 (AARCH64_ISA_F16FML): New ISA flag.
21300 (TARGET_F16FML): New feature flag for fp16fml.
21301 (doc/invoke.texi): Document new fp16fml option.
21303 2018-01-10 Michael Collison <michael.collison@arm.com>
21305 * config/aarch64/aarch64-builtins.c:
21306 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
21307 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21308 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
21309 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
21310 (AARCH64_ISA_SHA3): New ISA flag.
21311 (TARGET_SHA3): New feature flag for sha3.
21312 * config/aarch64/iterators.md (sha512_op): New int attribute.
21313 (CRYPTO_SHA512): New int iterator.
21314 (UNSPEC_SHA512H): New unspec.
21315 (UNSPEC_SHA512H2): Ditto.
21316 (UNSPEC_SHA512SU0): Ditto.
21317 (UNSPEC_SHA512SU1): Ditto.
21318 * config/aarch64/aarch64-simd-builtins.def
21319 (aarch64_crypto_sha512hqv2di): New builtin.
21320 (aarch64_crypto_sha512h2qv2di): Ditto.
21321 (aarch64_crypto_sha512su0qv2di): Ditto.
21322 (aarch64_crypto_sha512su1qv2di): Ditto.
21323 (aarch64_eor3qv8hi): Ditto.
21324 (aarch64_rax1qv2di): Ditto.
21325 (aarch64_xarqv2di): Ditto.
21326 (aarch64_bcaxqv8hi): Ditto.
21327 * config/aarch64/aarch64-simd.md:
21328 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
21329 (aarch64_crypto_sha512su0qv2di): Ditto.
21330 (aarch64_crypto_sha512su1qv2di): Ditto.
21331 (aarch64_eor3qv8hi): Ditto.
21332 (aarch64_rax1qv2di): Ditto.
21333 (aarch64_xarqv2di): Ditto.
21334 (aarch64_bcaxqv8hi): Ditto.
21335 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
21336 (vsha512h2q_u64): Ditto.
21337 (vsha512su0q_u64): Ditto.
21338 (vsha512su1q_u64): Ditto.
21339 (veor3q_u16): Ditto.
21340 (vrax1q_u64): Ditto.
21341 (vxarq_u64): Ditto.
21342 (vbcaxq_u16): Ditto.
21343 * config/arm/types.md (crypto_sha512): New type attribute.
21344 (crypto_sha3): Ditto.
21345 (doc/invoke.texi): Document new sha3 option.
21347 2018-01-10 Michael Collison <michael.collison@arm.com>
21349 * config/aarch64/aarch64-builtins.c:
21350 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
21351 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21352 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
21353 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
21354 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
21355 (AARCH64_ISA_SM4): New ISA flag.
21356 (TARGET_SM4): New feature flag for sm4.
21357 * config/aarch64/aarch64-simd-builtins.def
21358 (aarch64_sm3ss1qv4si): Ditto.
21359 (aarch64_sm3tt1aq4si): Ditto.
21360 (aarch64_sm3tt1bq4si): Ditto.
21361 (aarch64_sm3tt2aq4si): Ditto.
21362 (aarch64_sm3tt2bq4si): Ditto.
21363 (aarch64_sm3partw1qv4si): Ditto.
21364 (aarch64_sm3partw2qv4si): Ditto.
21365 (aarch64_sm4eqv4si): Ditto.
21366 (aarch64_sm4ekeyqv4si): Ditto.
21367 * config/aarch64/aarch64-simd.md:
21368 (aarch64_sm3ss1qv4si): Ditto.
21369 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
21370 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
21371 (aarch64_sm4eqv4si): Ditto.
21372 (aarch64_sm4ekeyqv4si): Ditto.
21373 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
21374 (sm3part_op): Ditto.
21375 (CRYPTO_SM3TT): Ditto.
21376 (CRYPTO_SM3PART): Ditto.
21377 (UNSPEC_SM3SS1): New unspec.
21378 (UNSPEC_SM3TT1A): Ditto.
21379 (UNSPEC_SM3TT1B): Ditto.
21380 (UNSPEC_SM3TT2A): Ditto.
21381 (UNSPEC_SM3TT2B): Ditto.
21382 (UNSPEC_SM3PARTW1): Ditto.
21383 (UNSPEC_SM3PARTW2): Ditto.
21384 (UNSPEC_SM4E): Ditto.
21385 (UNSPEC_SM4EKEY): Ditto.
21386 * config/aarch64/constraints.md (Ui2): New constraint.
21387 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
21388 * config/arm/types.md (crypto_sm3): New type attribute.
21389 (crypto_sm4): Ditto.
21390 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
21391 (vsm3tt1aq_u32): Ditto.
21392 (vsm3tt1bq_u32): Ditto.
21393 (vsm3tt2aq_u32): Ditto.
21394 (vsm3tt2bq_u32): Ditto.
21395 (vsm3partw1q_u32): Ditto.
21396 (vsm3partw2q_u32): Ditto.
21397 (vsm4eq_u32): Ditto.
21398 (vsm4ekeyq_u32): Ditto.
21399 (doc/invoke.texi): Document new sm4 option.
21401 2018-01-10 Michael Collison <michael.collison@arm.com>
21403 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
21404 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
21405 (AARCH64_FL_FOR_ARCH8_4): New.
21406 (AARCH64_FL_V8_4): New flag.
21407 (doc/invoke.texi): Document new armv8.4-a option.
21409 2018-01-10 Michael Collison <michael.collison@arm.com>
21411 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21412 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
21413 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
21414 * config/aarch64/aarch64-option-extension.def: Add
21415 AARCH64_OPT_EXTENSION of 'sha2'.
21416 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
21417 (crypto): Disable sha2 and aes if crypto disabled.
21418 (crypto): Enable aes and sha2 if enabled.
21419 (simd): Disable sha2 and aes if simd disabled.
21420 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
21422 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
21423 (TARGET_SHA2): New feature flag for sha2.
21424 (TARGET_AES): New feature flag for aes.
21425 * config/aarch64/aarch64-simd.md:
21426 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
21427 conditional on TARGET_AES.
21428 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
21429 (aarch64_crypto_sha1hsi): Make pattern conditional
21431 (aarch64_crypto_sha1hv4si): Ditto.
21432 (aarch64_be_crypto_sha1hv4si): Ditto.
21433 (aarch64_crypto_sha1su1v4si): Ditto.
21434 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
21435 (aarch64_crypto_sha1su0v4si): Ditto.
21436 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
21437 (aarch64_crypto_sha256su0v4si): Ditto.
21438 (aarch64_crypto_sha256su1v4si): Ditto.
21439 (doc/invoke.texi): Document new aes and sha2 options.
21441 2018-01-10 Martin Sebor <msebor@redhat.com>
21443 PR tree-optimization/83781
21444 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
21447 2018-01-11 Martin Sebor <msebor@gmail.com>
21448 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21450 PR tree-optimization/83501
21451 PR tree-optimization/81703
21453 * tree-ssa-strlen.c (get_string_cst): Rename...
21454 (get_string_len): ...to this. Handle global constants.
21455 (handle_char_store): Adjust.
21457 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
21458 Jim Wilson <jimw@sifive.com>
21460 * config/riscv/riscv-protos.h (riscv_output_return): New.
21461 * config/riscv/riscv.c (struct machine_function): New naked_p field.
21462 (riscv_attribute_table, riscv_output_return),
21463 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
21464 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
21465 (riscv_compute_frame_info): Only compute frame->mask if not a naked
21467 (riscv_expand_prologue): Add early return for naked function.
21468 (riscv_expand_epilogue): Likewise.
21469 (riscv_function_ok_for_sibcall): Return false for naked function.
21470 (riscv_set_current_function): New.
21471 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
21472 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
21473 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
21474 * doc/extend.texi (RISC-V Function Attributes): New.
21476 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
21478 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
21479 check for 128-bit long double before checking TCmode.
21480 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
21481 128-bit long doubles before checking TFmode or TCmode.
21482 (FLOAT128_IBM_P): Likewise.
21484 2018-01-10 Martin Sebor <msebor@redhat.com>
21486 PR tree-optimization/83671
21487 * builtins.c (c_strlen): Unconditionally return zero for the empty
21489 Use -Warray-bounds for warnings.
21490 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
21491 for non-constant array indices with COMPONENT_REF, arrays of
21492 arrays, and pointers to arrays.
21493 (gimple_fold_builtin_strlen): Determine and set length range for
21494 non-constant character arrays.
21496 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
21498 PR middle-end/81897
21499 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
21502 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
21504 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
21506 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
21509 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
21510 VECTOR_MEM_ALTIVEC_OR_VSX_P.
21511 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
21512 indexed_or_indirect_operand predicate.
21513 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
21514 (*vsx_le_perm_load_v8hi): Likewise.
21515 (*vsx_le_perm_load_v16qi): Likewise.
21516 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
21517 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
21518 (*vsx_le_perm_store_v8hi): Likewise.
21519 (*vsx_le_perm_store_v16qi): Likewise.
21520 (eight unnamed splitters): Likewise.
21522 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
21524 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
21525 * config/rs6000/emmintrin.h: Likewise.
21526 * config/rs6000/mmintrin.h: Likewise.
21527 * config/rs6000/xmmintrin.h: Likewise.
21529 2018-01-10 David Malcolm <dmalcolm@redhat.com>
21532 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
21534 * tree.c (tree_nop_conversion): Return true for location wrapper
21536 (maybe_wrap_with_location): New function.
21537 (selftest::check_strip_nops): New function.
21538 (selftest::test_location_wrappers): New function.
21539 (selftest::tree_c_tests): Call it.
21540 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
21541 (maybe_wrap_with_location): New decl.
21542 (EXPR_LOCATION_WRAPPER_P): New macro.
21543 (location_wrapper_p): New inline function.
21544 (tree_strip_any_location_wrapper): New inline function.
21546 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
21549 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
21550 stack_realign_offset for the largest alignment of stack slot
21552 (ix86_find_max_used_stack_alignment): New function.
21553 (ix86_finalize_stack_frame_flags): Use it. Set
21554 max_used_stack_alignment if we don't realign stack.
21555 * config/i386/i386.h (machine_function): Add
21556 max_used_stack_alignment.
21558 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
21560 * config/arm/arm.opt (-mbranch-cost): New option.
21561 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
21564 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
21567 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
21568 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
21570 2018-01-10 Richard Biener <rguenther@suse.de>
21573 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
21574 early out so it also covers the case where we have a non-NULL
21577 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
21579 PR tree-optimization/83753
21580 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
21581 for non-strided grouped accesses if the number of elements is 1.
21583 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21586 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
21587 * i386.h (TARGET_USE_GATHER): Define.
21588 * x86-tune.def (X86_TUNE_USE_GATHER): New.
21590 2018-01-10 Martin Liska <mliska@suse.cz>
21593 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
21594 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
21596 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
21597 CLEANUP_NO_PARTITIONING is not set.
21599 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
21601 * doc/rtl.texi: Remove documentation of (const ...) wrappers
21602 for vectors, as a partial revert of r254296.
21603 * rtl.h (const_vec_p): Delete.
21604 (const_vec_duplicate_p): Don't test for vector CONSTs.
21605 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
21606 * expmed.c (make_tree): Likewise.
21609 * common.md (E, F): Use CONSTANT_P instead of checking for
21611 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
21612 checking for CONST_VECTOR.
21614 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21616 PR middle-end/83575
21617 * predict.c (force_edge_cold): Handle in more sane way edges
21618 with no prediction.
21620 2018-01-09 Carl Love <cel@us.ibm.com>
21622 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
21624 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
21625 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
21626 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
21627 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
21628 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
21629 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
21630 * config/rs6000/rs6000-protos.h: Add extern defition for
21631 rs6000_generate_float2_double_code.
21632 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
21634 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
21635 (float2_v2df): Add define_expand.
21637 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
21640 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
21641 op_mode in the force_to_mode call.
21643 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21645 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
21646 instead of checking each element individually.
21647 (aarch64_evpc_uzp): Likewise.
21648 (aarch64_evpc_zip): Likewise.
21649 (aarch64_evpc_ext): Likewise.
21650 (aarch64_evpc_rev): Likewise.
21651 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
21652 instead of checking each element individually. Return true without
21654 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
21655 whether all selected elements come from the same input, instead of
21656 checking each element individually. Remove calls to gen_rtx_REG,
21657 start_sequence and end_sequence and instead assert that no rtl is
21660 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21662 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
21663 order of HIGH and CONST checks.
21665 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21667 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
21668 if the destination isn't an SSA_NAME.
21670 2018-01-09 Richard Biener <rguenther@suse.de>
21672 PR tree-optimization/83668
21673 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
21675 (canonicalize_loop_form): ... here, renamed from ...
21676 (canonicalize_loop_closed_ssa_form): ... this and amended to
21677 swap successor edges for loop exit blocks to make us use
21678 the RPO order we need for initial schedule generation.
21680 2018-01-09 Joseph Myers <joseph@codesourcery.com>
21682 PR tree-optimization/64811
21683 * match.pd: When optimizing comparisons with Inf, avoid
21684 introducing or losing exceptions from comparisons with NaN.
21686 2018-01-09 Martin Liska <mliska@suse.cz>
21689 * asan.c (shadow_mem_size): Add gcc_assert.
21691 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
21693 Don't save registers in main().
21696 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
21697 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
21698 * config/avr/avr.c (avr_set_current_function): Don't error if
21699 naked, OS_task or OS_main are specified at the same time.
21700 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
21702 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
21704 * common/config/avr/avr-common.c (avr_option_optimization_table):
21705 Switch on -mmain-is-OS_task for optimizing compilations.
21707 2018-01-09 Richard Biener <rguenther@suse.de>
21709 PR tree-optimization/83572
21710 * graphite.c: Include cfganal.h.
21711 (graphite_transform_loops): Connect infinite loops to exit
21712 and remove fake edges at the end.
21714 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21716 * ipa-inline.c (edge_badness): Revert accidental checkin.
21718 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21721 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
21722 symbols; not inline clones.
21724 2018-01-09 Jakub Jelinek <jakub@redhat.com>
21727 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
21728 hard registers. Formatting fixes.
21730 PR preprocessor/83722
21731 * gcc.c (try_generate_repro): Pass
21732 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
21733 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
21736 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
21737 Kito Cheng <kito.cheng@gmail.com>
21739 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
21740 (riscv_leaf_function_p): Delete.
21741 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
21743 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21745 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
21747 (do_ifelse): New function.
21748 (do_isel): New function.
21749 (do_sub3): New function.
21750 (do_add3): New function.
21751 (do_load_mask_compare): New function.
21752 (do_overlap_load_compare): New function.
21753 (expand_compare_loop): New function.
21754 (expand_block_compare): Call expand_compare_loop() when appropriate.
21755 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
21756 option description.
21757 (-mblock-compare-inline-loop-limit): New option.
21759 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21762 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
21763 Reverse order of second and third operands in first alternative.
21764 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
21765 of first and second elements in UNSPEC_VPERMR vector.
21766 (altivec_expand_vec_perm_le): Likewise.
21768 2018-01-08 Jeff Law <law@redhat.com>
21770 PR rtl-optimizatin/81308
21771 * tree-switch-conversion.c (cfg_altered): New file scoped static.
21772 (process_switch): If group_case_labels makes a change, then set
21774 (pass_convert_switch::execute): If a switch is converted, then
21775 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
21777 PR rtl-optimization/81308
21778 * recog.c (split_all_insns): Conditionally cleanup the CFG after
21781 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
21783 PR target/83663 - Revert r255946
21784 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
21785 generation for cases where splatting a value is not useful.
21786 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
21787 across a vec_duplicate and a paradoxical subreg forming a vector
21788 mode to a vec_concat.
21790 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21792 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
21793 -march=armv8.3-a variants.
21794 * config/arm/t-multilib: Likewise.
21795 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
21797 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21799 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
21801 (cceq_ior_compare_complement): Give it a name so I can use it, and
21802 change boolean_or_operator predicate to boolean_operator so it can
21803 be used to generate a crand.
21804 (eqne): New code iterator.
21805 (bd/bd_neg): New code_attrs.
21806 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
21807 a single define_insn.
21808 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
21809 decrement (bdnzt/bdnzf/bdzt/bdzf).
21810 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
21811 with the new names of the branch decrement patterns, and added the
21812 names of the branch decrement conditional patterns.
21814 2018-01-08 Richard Biener <rguenther@suse.de>
21816 PR tree-optimization/83563
21817 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
21820 2018-01-08 Richard Biener <rguenther@suse.de>
21822 PR middle-end/83713
21823 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
21825 2018-01-08 Richard Biener <rguenther@suse.de>
21827 PR tree-optimization/83685
21828 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
21829 references to abnormals.
21831 2018-01-08 Richard Biener <rguenther@suse.de>
21834 * dwarf2out.c (output_indirect_strings): Handle empty
21835 skeleton_debug_str_hash.
21836 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
21838 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21840 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
21841 (emit_store_direct): Likewise.
21842 (arc_trampoline_adjust_address): Likewise.
21843 (arc_asm_trampoline_template): New function.
21844 (arc_initialize_trampoline): Use asm_trampoline_template.
21845 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
21846 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
21847 * config/arc/arc.md (flush_icache): Delete pattern.
21849 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21851 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
21852 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
21855 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21858 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
21859 by not USED_FOR_TARGET.
21860 (make_pass_resolve_sw_modes): Likewise.
21862 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21864 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
21867 2018-01-08 Richard Biener <rguenther@suse.de>
21869 PR middle-end/83580
21870 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
21872 2018-01-08 Richard Biener <rguenther@suse.de>
21874 PR middle-end/83517
21875 * match.pd ((t * 2) / 2) -> t): Add missing :c.
21877 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
21879 PR middle-end/81897
21880 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
21881 basic blocks with a small number of successors.
21882 (convert_control_dep_chain_into_preds): Improve handling of
21884 (dump_predicates): Split apart into...
21885 (dump_pred_chain): ...here...
21886 (dump_pred_info): ...and here.
21887 (can_one_predicate_be_invalidated_p): Add debugging printfs.
21888 (can_chain_union_be_invalidated_p): Improve check for invalidation
21890 (uninit_uses_cannot_happen): Avoid unnecessary if
21891 convert_control_dep_chain_into_preds yielded nothing.
21893 2018-01-06 Martin Sebor <msebor@redhat.com>
21895 PR tree-optimization/83640
21896 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
21897 subtracting negative offset from size.
21898 (builtin_access::overlap): Adjust offset bounds of the access to fall
21899 within the size of the object if possible.
21901 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
21903 PR rtl-optimization/83699
21904 * expmed.c (extract_bit_field_1): Restrict the vector usage of
21905 extract_bit_field_as_subreg to cases in which the extracted
21906 value is also a vector.
21908 * lra-constraints.c (process_alt_operands): Test for the equivalence
21909 substitutions when detecting a possible reload cycle.
21911 2018-01-06 Jakub Jelinek <jakub@redhat.com>
21914 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
21915 by default if flag_selective_schedling{,2}. Formatting fixes.
21917 PR rtl-optimization/83682
21918 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
21919 if it has non-VECTOR_MODE element mode.
21920 (vec_duplicate_p): Likewise.
21922 PR middle-end/83694
21923 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
21924 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
21926 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21929 * config/i386/i386-builtin.def
21930 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
21931 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
21932 Require also OPTION_MASK_ISA_AVX512F in addition to
21933 OPTION_MASK_ISA_GFNI.
21934 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
21935 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
21936 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
21937 to OPTION_MASK_ISA_GFNI.
21938 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
21939 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
21940 OPTION_MASK_ISA_AVX512BW.
21941 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
21942 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
21943 addition to OPTION_MASK_ISA_GFNI.
21944 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
21945 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
21946 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
21947 to OPTION_MASK_ISA_GFNI.
21948 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
21949 a requirement for all ISAs rather than any of them with a few
21951 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
21953 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
21954 bitmasks to be enabled with 3 exceptions, instead of requiring any
21955 enabled ISA with lots of exceptions.
21956 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
21957 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
21958 Change avx512bw in isa attribute to avx512f.
21959 * config/i386/sgxintrin.h: Add license boilerplate.
21960 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
21961 to __AVX512F__ and __AVX512VL to __AVX512VL__.
21962 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
21963 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
21965 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
21966 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
21967 temporarily sse2 rather than sse if not enabled already.
21970 * config/i386/sse.md (VI248_VLBW): Rename to ...
21971 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
21972 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
21973 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
21974 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
21975 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
21976 mode iterator instead of VI248_VLBW.
21978 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
21980 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
21981 (record_modified): Skip clobbers; add debug output.
21982 (param_change_prob): Use sreal frequencies.
21984 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21986 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
21987 punt for user-aligned variables.
21989 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21991 * tree-chrec.c (chrec_contains_symbols): Return true for
21994 2018-01-05 Sudakshina Das <sudi.das@arm.com>
21997 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
21998 of (x|y) == x for BICS pattern.
22000 2018-01-05 Jakub Jelinek <jakub@redhat.com>
22002 PR tree-optimization/83605
22003 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
22004 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
22007 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
22009 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
22010 * config/epiphany/rtems.h: New file.
22012 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22013 Uros Bizjak <ubizjak@gmail.com>
22016 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
22017 QIreg_operand instead of register_operand predicate.
22018 * config/i386/i386.c (ix86_rop_should_change_byte_p,
22019 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
22020 comments instead of -fmitigate[-_]rop.
22022 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22025 * cgraphunit.c (symbol_table::compile): Switch to text_section
22026 before calling assembly_start debug hook.
22027 * run-rtl-passes.c (run_rtl_passes): Likewise.
22030 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22032 * tree-vrp.c (extract_range_from_binary_expr_1): Check
22033 range_int_cst_p rather than !symbolic_range_p before calling
22034 extract_range_from_multiplicative_op_1.
22036 2018-01-04 Jeff Law <law@redhat.com>
22038 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
22039 redundant test in assertion.
22041 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22043 * doc/rtl.texi: Document machine_mode wrapper classes.
22045 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22047 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
22048 using tree_to_uhwi.
22050 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22052 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
22053 the VEC_PERM_EXPR fold to fail.
22055 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22058 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
22059 to switched_sections.
22061 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22064 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
22065 test for d.testing.
22067 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
22070 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
22071 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
22073 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22076 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
22077 is BLKmode and bitpos not zero or mode change is needed.
22079 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
22082 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
22085 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
22088 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
22089 instead of MULT rtx. Update all corresponding splitters.
22090 (*saddl_se): Ditto.
22091 (*ssub<modesuffix>): Ditto.
22092 (*ssubl_se): Ditto.
22093 (*cmp_sadd_di): Update split patterns.
22094 (*cmp_sadd_si): Ditto.
22095 (*cmp_sadd_sidi): Ditto.
22096 (*cmp_ssub_di): Ditto.
22097 (*cmp_ssub_si): Ditto.
22098 (*cmp_ssub_sidi): Ditto.
22099 * config/alpha/predicates.md (const23_operand): New predicate.
22100 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
22101 Look for ASHIFT, not MULT inner operand.
22102 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
22104 2018-01-04 Martin Liska <mliska@suse.cz>
22106 PR gcov-profile/83669
22107 * gcov.c (output_intermediate_file): Add version to intermediate
22109 * doc/gcov.texi: Document new field 'version' in intermediate
22110 file format. Fix location of '-k' option of gcov command.
22112 2018-01-04 Martin Liska <mliska@suse.cz>
22115 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
22117 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22119 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
22121 2018-01-03 Martin Sebor <msebor@redhat.com>
22123 PR tree-optimization/83655
22124 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
22125 checking calls with invalid arguments.
22127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22129 * tree-vect-stmts.c (vect_get_store_rhs): New function.
22130 (vectorizable_mask_load_store): Delete.
22131 (vectorizable_call): Return false for masked loads and stores.
22132 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
22133 instead of gimple_assign_rhs1.
22134 (vectorizable_load): Handle IFN_MASK_LOAD.
22135 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
22137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22139 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
22141 (vectorizable_mask_load_store): ...here.
22142 (vectorizable_load): ...and here.
22144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22146 * tree-vect-stmts.c (vect_build_all_ones_mask)
22147 (vect_build_zero_merge_argument): New functions, split out from...
22148 (vectorizable_load): ...here.
22150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22152 * tree-vect-stmts.c (vect_check_store_rhs): New function,
22154 (vectorizable_mask_load_store): ...here.
22155 (vectorizable_store): ...and here.
22157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22159 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
22161 (vectorizable_mask_load_store): ...here.
22163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22165 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
22166 (vect_model_store_cost): Take a vec_load_store_type instead of a
22168 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
22169 (vect_model_store_cost): Take a vec_load_store_type instead of a
22171 (vectorizable_mask_load_store): Update accordingly.
22172 (vectorizable_store): Likewise.
22173 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
22175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22177 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
22178 IFN_MASK_LOAD calls here rather than...
22179 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
22181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22182 Alan Hayward <alan.hayward@arm.com>
22183 David Sherwood <david.sherwood@arm.com>
22185 * expmed.c (extract_bit_field_1): For vector extracts,
22186 fall back to extract_bit_field_as_subreg if vec_extract
22189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22190 Alan Hayward <alan.hayward@arm.com>
22191 David Sherwood <david.sherwood@arm.com>
22193 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
22194 they are variable or constant sized.
22195 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
22196 slots for constant-sized data.
22198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22199 Alan Hayward <alan.hayward@arm.com>
22200 David Sherwood <david.sherwood@arm.com>
22202 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
22203 handling COND_EXPRs with boolean comparisons, try to find a better
22204 basis for the mask type than the boolean itself.
22206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22208 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
22209 is calculated and how it can be overridden.
22210 * genmodes.c (max_bitsize_mode_any_mode): New variable.
22211 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
22213 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
22216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22217 Alan Hayward <alan.hayward@arm.com>
22218 David Sherwood <david.sherwood@arm.com>
22220 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
22221 Remove the mode argument.
22222 (aarch64_simd_valid_immediate): Remove the mode and inverse
22224 * config/aarch64/iterators.md (bitsize): New iterator.
22225 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
22226 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
22227 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
22228 aarch64_simd_valid_immediate.
22229 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
22230 (aarch64_reg_or_bic_imm): Likewise.
22231 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
22232 with an insn_type enum and msl with a modifier_type enum.
22233 Replace element_width with a scalar_mode. Change the shift
22234 to unsigned int. Add constructors for scalar_float_mode and
22235 scalar_int_mode elements.
22236 (aarch64_vect_float_const_representable_p): Delete.
22237 (aarch64_can_const_movi_rtx_p)
22238 (aarch64_simd_scalar_immediate_valid_for_move)
22239 (aarch64_simd_make_constant): Update call to
22240 aarch64_simd_valid_immediate.
22241 (aarch64_advsimd_valid_immediate_hs): New function.
22242 (aarch64_advsimd_valid_immediate): Likewise.
22243 (aarch64_simd_valid_immediate): Remove mode and inverse
22244 arguments. Rewrite to use the above. Use const_vec_duplicate_p
22245 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
22246 and aarch64_float_const_representable_p on the result.
22247 (aarch64_output_simd_mov_immediate): Remove mode argument.
22248 Update call to aarch64_simd_valid_immediate and use of
22249 simd_immediate_info.
22250 (aarch64_output_scalar_simd_mov_immediate): Update call
22253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22254 Alan Hayward <alan.hayward@arm.com>
22255 David Sherwood <david.sherwood@arm.com>
22257 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
22258 (mode_nunits): Likewise CONST_MODE_NUNITS.
22259 * machmode.def (ADJUST_NUNITS): Document.
22260 * genmodes.c (mode_data::need_nunits_adj): New field.
22261 (blank_mode): Update accordingly.
22262 (adj_nunits): New variable.
22263 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
22265 (emit_mode_size_inline): Set need_bytesize_adj for all modes
22266 listed in adj_nunits.
22267 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
22268 listed in adj_nunits. Don't emit case statements for such modes.
22269 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
22270 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
22271 nothing if adj_nunits is nonnull.
22272 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
22273 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
22274 (emit_mode_fbit): Update use of print_maybe_const_decl.
22275 (emit_move_size): Likewise. Treat the array as non-const
22277 (emit_mode_adjustments): Handle adj_nunits.
22279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22281 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
22282 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
22283 (VECTOR_MODES): Use it.
22284 (make_vector_modes): Take the prefix as an argument.
22286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22287 Alan Hayward <alan.hayward@arm.com>
22288 David Sherwood <david.sherwood@arm.com>
22290 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
22291 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
22292 for MODE_VECTOR_BOOL.
22293 * machmode.def (VECTOR_BOOL_MODE): Document.
22294 * genmodes.c (VECTOR_BOOL_MODE): New macro.
22295 (make_vector_bool_mode): New function.
22296 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
22298 * lto-streamer-in.c (lto_input_mode_table): Likewise.
22299 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
22301 * stor-layout.c (int_mode_for_mode): Likewise.
22302 * tree.c (build_vector_type_for_mode): Likewise.
22303 * varasm.c (output_constant_pool_2): Likewise.
22304 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
22305 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
22306 for MODE_VECTOR_BOOL.
22307 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
22308 of mode class checks.
22309 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
22310 instead of a list of mode class checks.
22311 (expand_vector_scalar_condition): Likewise.
22312 (type_for_widest_vector_mode): Handle BImode as an inner mode.
22314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22315 Alan Hayward <alan.hayward@arm.com>
22316 David Sherwood <david.sherwood@arm.com>
22318 * machmode.h (mode_size): Change from unsigned short to
22320 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
22321 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
22322 or if measurement_type is not polynomial.
22323 (fixed_size_mode::includes_p): Check for constant-sized modes.
22324 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
22325 return a poly_uint16 rather than an unsigned short.
22326 (emit_mode_size): Change the type of mode_size from unsigned short
22327 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
22328 (emit_mode_adjustments): Cope with polynomial vector sizes.
22329 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22331 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22333 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
22334 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
22335 * caller-save.c (setup_save_areas): Likewise.
22336 (replace_reg_with_saved_mem): Likewise.
22337 * calls.c (emit_library_call_value_1): Likewise.
22338 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
22339 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
22340 (gen_lowpart_for_combine): Likewise.
22341 * convert.c (convert_to_integer_1): Likewise.
22342 * cse.c (equiv_constant, cse_insn): Likewise.
22343 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
22344 (cselib_subst_to_values): Likewise.
22345 * dce.c (word_dce_process_block): Likewise.
22346 * df-problems.c (df_word_lr_mark_ref): Likewise.
22347 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
22348 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
22349 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
22350 (rtl_for_decl_location): Likewise.
22351 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
22352 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
22353 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
22354 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
22355 (expand_expr_real_1): Likewise.
22356 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
22357 (pad_below): Likewise.
22358 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
22359 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
22360 * ira.c (get_subreg_tracking_sizes): Likewise.
22361 * ira-build.c (ira_create_allocno_objects): Likewise.
22362 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
22363 (ira_sort_regnos_for_alter_reg): Likewise.
22364 * ira-costs.c (record_operand_costs): Likewise.
22365 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
22366 (resolve_simple_move): Likewise.
22367 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
22368 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
22369 (lra_constraints): Likewise.
22370 (CONST_POOL_OK_P): Reject variable-sized modes.
22371 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
22372 (add_pseudo_to_slot, lra_spill): Likewise.
22373 * omp-low.c (omp_clause_aligned_alignment): Likewise.
22374 * optabs-query.c (get_best_extraction_insn): Likewise.
22375 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
22376 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
22377 (expand_mult_highpart, valid_multiword_target_p): Likewise.
22378 * recog.c (offsettable_address_addr_space_p): Likewise.
22379 * regcprop.c (maybe_mode_change): Likewise.
22380 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
22381 * regrename.c (build_def_use): Likewise.
22382 * regstat.c (dump_reg_info): Likewise.
22383 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
22384 (find_reloads, find_reloads_subreg_address): Likewise.
22385 * reload1.c (eliminate_regs_1): Likewise.
22386 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
22387 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
22388 (simplify_binary_operation_1, simplify_subreg): Likewise.
22389 * targhooks.c (default_function_arg_padding): Likewise.
22390 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
22391 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
22392 (verify_gimple_assign_ternary): Likewise.
22393 * tree-inline.c (estimate_move_cost): Likewise.
22394 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22395 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
22396 (get_address_cost_ainc): Likewise.
22397 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
22398 (vect_supportable_dr_alignment): Likewise.
22399 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
22400 (vectorizable_reduction): Likewise.
22401 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
22402 (vectorizable_operation, vectorizable_load): Likewise.
22403 * tree.c (build_same_sized_truth_vector_type): Likewise.
22404 * valtrack.c (cleanup_auto_inc_dec): Likewise.
22405 * var-tracking.c (emit_note_insn_var_location): Likewise.
22406 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
22407 (ADDR_VEC_ALIGN): Likewise.
22409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22410 Alan Hayward <alan.hayward@arm.com>
22411 David Sherwood <david.sherwood@arm.com>
22413 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
22415 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
22416 or if measurement_type is polynomial.
22417 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
22418 * combine.c (make_extraction): Likewise.
22419 * dse.c (find_shift_sequence): Likewise.
22420 * dwarf2out.c (mem_loc_descriptor): Likewise.
22421 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
22422 (extract_bit_field, extract_low_bits): Likewise.
22423 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
22424 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
22425 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
22426 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
22427 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
22428 * reload.c (find_reloads): Likewise.
22429 * reload1.c (alter_reg): Likewise.
22430 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
22431 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
22432 * tree-if-conv.c (predicate_mem_writes): Likewise.
22433 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
22434 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
22435 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
22436 * valtrack.c (dead_debug_insert_temp): Likewise.
22437 * varasm.c (mergeable_constant_section): Likewise.
22438 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
22440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22441 Alan Hayward <alan.hayward@arm.com>
22442 David Sherwood <david.sherwood@arm.com>
22444 * expr.c (expand_assignment): Cope with polynomial mode sizes
22445 when assigning to a CONCAT.
22447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22448 Alan Hayward <alan.hayward@arm.com>
22449 David Sherwood <david.sherwood@arm.com>
22451 * machmode.h (mode_precision): Change from unsigned short to
22453 (mode_to_precision): Return a poly_uint16 rather than an unsigned
22455 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
22456 or if measurement_type is not polynomial.
22457 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
22458 in which the mode is already known to be a scalar_int_mode.
22459 * genmodes.c (emit_mode_precision): Change the type of mode_precision
22460 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
22462 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22463 for GET_MODE_PRECISION.
22464 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22465 for GET_MODE_PRECISION.
22466 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
22468 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
22469 (expand_field_assignment, make_extraction): Likewise.
22470 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
22471 (get_last_value): Likewise.
22472 * convert.c (convert_to_integer_1): Likewise.
22473 * cse.c (cse_insn): Likewise.
22474 * expr.c (expand_expr_real_1): Likewise.
22475 * lra-constraints.c (simplify_operand_subreg): Likewise.
22476 * optabs-query.c (can_atomic_load_p): Likewise.
22477 * optabs.c (expand_atomic_load): Likewise.
22478 (expand_atomic_store): Likewise.
22479 * ree.c (combine_reaching_defs): Likewise.
22480 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
22481 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
22482 * tree.h (type_has_mode_precision_p): Likewise.
22483 * ubsan.c (instrument_si_overflow): Likewise.
22485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22486 Alan Hayward <alan.hayward@arm.com>
22487 David Sherwood <david.sherwood@arm.com>
22489 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
22490 polynomial numbers of units.
22491 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
22492 (valid_vector_subparts_p): New function.
22493 (build_vector_type): Remove temporary shim and take the number
22494 of units as a poly_uint64 rather than an int.
22495 (build_opaque_vector_type): Take the number of units as a
22496 poly_uint64 rather than an int.
22497 * tree.c (build_vector_from_ctor): Handle polynomial
22498 TYPE_VECTOR_SUBPARTS.
22499 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
22500 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
22501 (build_vector_from_val): If the number of units is variable,
22502 use build_vec_duplicate_cst for constant operands and
22503 VEC_DUPLICATE_EXPR otherwise.
22504 (make_vector_type): Remove temporary is_constant ().
22505 (build_vector_type, build_opaque_vector_type): Take the number of
22506 units as a poly_uint64 rather than an int.
22507 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
22509 * cfgexpand.c (expand_debug_expr): Likewise.
22510 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
22511 (store_constructor, expand_expr_real_1): Likewise.
22512 (const_scalar_mask_from_tree): Likewise.
22513 * fold-const-call.c (fold_const_reduction): Likewise.
22514 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
22515 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
22516 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
22517 (fold_relational_const): Likewise.
22518 (native_interpret_vector): Likewise. Change the size from an
22519 int to an unsigned int.
22520 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
22521 TYPE_VECTOR_SUBPARTS.
22522 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
22523 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
22524 duplicating a non-constant operand into a variable-length vector.
22525 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
22526 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
22527 * ipa-icf.c (sem_variable::equals): Likewise.
22528 * match.pd: Likewise.
22529 * omp-simd-clone.c (simd_clone_subparts): Likewise.
22530 * print-tree.c (print_node): Likewise.
22531 * stor-layout.c (layout_type): Likewise.
22532 * targhooks.c (default_builtin_vectorization_cost): Likewise.
22533 * tree-cfg.c (verify_gimple_comparison): Likewise.
22534 (verify_gimple_assign_binary): Likewise.
22535 (verify_gimple_assign_ternary): Likewise.
22536 (verify_gimple_assign_single): Likewise.
22537 * tree-pretty-print.c (dump_generic_node): Likewise.
22538 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22539 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
22540 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
22541 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
22542 (vect_shift_permute_load_chain): Likewise.
22543 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
22544 (expand_vector_condition, optimize_vector_constructor): Likewise.
22545 (lower_vec_perm, get_compute_type): Likewise.
22546 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
22547 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
22548 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
22549 (vect_recog_mask_conversion_pattern): Likewise.
22550 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
22551 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
22552 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22553 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
22554 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
22555 (vectorizable_shift, vectorizable_operation, vectorizable_store)
22556 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
22557 (supportable_widening_operation): Likewise.
22558 (supportable_narrowing_operation): Likewise.
22559 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
22561 * varasm.c (output_constant): Likewise.
22563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22564 Alan Hayward <alan.hayward@arm.com>
22565 David Sherwood <david.sherwood@arm.com>
22567 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
22568 so that both the length == 3 and length != 3 cases set up their
22569 own permute vectors. Add comments explaining why we know the
22570 number of elements is constant.
22571 (vect_permute_load_chain): Likewise.
22573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22574 Alan Hayward <alan.hayward@arm.com>
22575 David Sherwood <david.sherwood@arm.com>
22577 * machmode.h (mode_nunits): Change from unsigned char to
22579 (ONLY_FIXED_SIZE_MODES): New macro.
22580 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
22581 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
22582 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
22584 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
22585 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
22586 or if measurement_type is not polynomial.
22587 * genmodes.c (ZERO_COEFFS): New macro.
22588 (emit_mode_nunits_inline): Make mode_nunits_inline return a
22590 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
22591 Use ZERO_COEFFS when emitting initializers.
22592 * data-streamer.h (bp_pack_poly_value): New function.
22593 (bp_unpack_poly_value): Likewise.
22594 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22595 for GET_MODE_NUNITS.
22596 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22597 for GET_MODE_NUNITS.
22598 * tree.c (make_vector_type): Remove temporary shim and make
22599 the real function take the number of units as a poly_uint64
22600 rather than an int.
22601 (build_vector_type_for_mode): Handle polynomial nunits.
22602 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
22603 * emit-rtl.c (const_vec_series_p_1): Likewise.
22604 (gen_rtx_CONST_VECTOR): Likewise.
22605 * fold-const.c (test_vec_duplicate_folding): Likewise.
22606 * genrecog.c (validate_pattern): Likewise.
22607 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
22608 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
22609 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
22610 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
22611 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
22612 * rtlanal.c (subreg_get_info): Likewise.
22613 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22614 (vect_grouped_load_supported): Likewise.
22615 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
22616 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
22617 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
22618 (simplify_const_unary_operation, simplify_binary_operation_1)
22619 (simplify_const_binary_operation, simplify_ternary_operation)
22620 (test_vector_ops_duplicate, test_vector_ops): Likewise.
22621 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
22622 instead of CONST_VECTOR_NUNITS.
22623 * varasm.c (output_constant_pool_2): Likewise.
22624 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
22625 explicit-encoded elements in the XVEC for variable-length vectors.
22627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22629 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
22631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22632 Alan Hayward <alan.hayward@arm.com>
22633 David Sherwood <david.sherwood@arm.com>
22635 * coretypes.h (fixed_size_mode): Declare.
22636 (fixed_size_mode_pod): New typedef.
22637 * builtins.h (target_builtins::x_apply_args_mode)
22638 (target_builtins::x_apply_result_mode): Change type to
22639 fixed_size_mode_pod.
22640 * builtins.c (apply_args_size, apply_result_size, result_vector)
22641 (expand_builtin_apply_args_1, expand_builtin_apply)
22642 (expand_builtin_return): Update accordingly.
22644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22646 * cse.c (hash_rtx_cb): Hash only the encoded elements.
22647 * cselib.c (cselib_hash_rtx): Likewise.
22648 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
22649 CONST_VECTOR encoding.
22651 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22652 Jeff Law <law@redhat.com>
22655 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
22656 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
22657 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
22658 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
22661 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
22662 explicitly probe *sp in a noreturn function if there were any callee
22663 register saves or frame pointer is needed.
22665 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22668 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
22669 BLKmode for ternary, binary or unary expressions.
22672 * var-tracking.c (delete_vta_debug_insn): New inline function.
22673 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
22674 insns from get_insns () to NULL instead of each bb separately.
22675 Use delete_vta_debug_insn. No longer static.
22676 (vt_debug_insns_local, variable_tracking_main_1): Adjust
22677 delete_vta_debug_insns callers.
22678 * rtl.h (delete_vta_debug_insns): Declare.
22679 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
22680 instead of variable_tracking_main.
22682 2018-01-03 Martin Sebor <msebor@redhat.com>
22684 PR tree-optimization/83603
22685 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
22686 arguments past the endof the argument list in functions declared
22687 without a prototype.
22688 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
22689 Avoid checking when arguments are null.
22691 2018-01-03 Martin Sebor <msebor@redhat.com>
22694 * doc/extend.texi (attribute const): Fix a typo.
22695 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
22696 issuing -Wsuggest-attribute for void functions.
22698 2018-01-03 Martin Sebor <msebor@redhat.com>
22700 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
22701 offset_int::from instead of wide_int::to_shwi.
22702 (maybe_diag_overlap): Remove assertion.
22703 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
22704 * gimple-ssa-sprintf.c (format_directive): Same.
22705 (parse_directive): Same.
22706 (sprintf_dom_walker::compute_format_length): Same.
22707 (try_substitute_return_value): Same.
22709 2018-01-03 Jeff Law <law@redhat.com>
22711 PR middle-end/83654
22712 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
22713 non-constant residual for zero at runtime and avoid probing in
22714 that case. Reorganize code for trailing problem to mirror handling
22717 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22719 PR tree-optimization/83501
22720 * tree-ssa-strlen.c (get_string_cst): New.
22721 (handle_char_store): Call get_string_cst.
22723 2018-01-03 Martin Liska <mliska@suse.cz>
22725 PR tree-optimization/83593
22726 * tree-ssa-strlen.c: Include tree-cfg.h.
22727 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
22728 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
22729 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
22731 (strlen_dom_walker::before_dom_children): Call
22732 gimple_purge_dead_eh_edges. Dump tranformation with details
22734 (strlen_dom_walker::before_dom_children): Update call by adding
22735 new argument cleanup_eh.
22736 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
22738 2018-01-03 Martin Liska <mliska@suse.cz>
22741 * cif-code.def (VARIADIC_THUNK): New enum value.
22742 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
22745 2018-01-03 Jan Beulich <jbeulich@suse.com>
22747 * sse.md (mov<mode>_internal): Tighten condition for when to use
22748 vmovdqu<ssescalarsize> for TI and OI modes.
22750 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22752 Update copyright years.
22754 2018-01-03 Martin Liska <mliska@suse.cz>
22757 * ipa-visibility.c (function_and_variable_visibility): Skip
22758 functions with noipa attribure.
22760 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22762 * gcc.c (process_command): Update copyright notice dates.
22763 * gcov-dump.c (print_version): Ditto.
22764 * gcov.c (print_version): Ditto.
22765 * gcov-tool.c (print_version): Ditto.
22766 * gengtype.c (create_file): Ditto.
22767 * doc/cpp.texi: Bump @copying's copyright year.
22768 * doc/cppinternals.texi: Ditto.
22769 * doc/gcc.texi: Ditto.
22770 * doc/gccint.texi: Ditto.
22771 * doc/gcov.texi: Ditto.
22772 * doc/install.texi: Ditto.
22773 * doc/invoke.texi: Ditto.
22775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22777 * vector-builder.h (vector_builder::m_full_nelts): Change from
22778 unsigned int to poly_uint64.
22779 (vector_builder::full_nelts): Update prototype accordingly.
22780 (vector_builder::new_vector): Likewise.
22781 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
22782 (vector_builder::operator ==): Likewise.
22783 (vector_builder::finalize): Likewise.
22784 * int-vector-builder.h (int_vector_builder::int_vector_builder):
22785 Take the number of elements as a poly_uint64 rather than an
22787 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
22788 from unsigned int to poly_uint64.
22789 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
22790 (vec_perm_indices::new_vector): Likewise.
22791 (vec_perm_indices::length): Likewise.
22792 (vec_perm_indices::nelts_per_input): Likewise.
22793 (vec_perm_indices::input_nelts): Likewise.
22794 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
22795 number of elements per input as a poly_uint64 rather than an
22796 unsigned int. Use the original encoding for variable-length
22797 vectors, rather than clamping each individual element.
22798 For the second and subsequent elements in each pattern,
22799 clamp the step and base before clamping their sum.
22800 (vec_perm_indices::series_p): Handle polynomial element counts.
22801 (vec_perm_indices::all_in_range_p): Likewise.
22802 (vec_perm_indices_to_tree): Likewise.
22803 (vec_perm_indices_to_rtx): Likewise.
22804 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
22805 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
22806 (tree_vector_builder::new_binary_operation): Handle polynomial
22807 element counts. Return false if we need to know the number
22808 of elements at compile time.
22809 * fold-const.c (fold_vec_perm): Punt if the number of elements
22810 isn't known at compile time.
22812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22814 * vec-perm-indices.h (vec_perm_builder): Change element type
22815 from HOST_WIDE_INT to poly_int64.
22816 (vec_perm_indices::element_type): Update accordingly.
22817 (vec_perm_indices::clamp): Handle polynomial element_types.
22818 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22819 (vec_perm_indices::all_in_range_p): Likewise.
22820 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
22822 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
22823 polynomial vec_perm_indices element types.
22824 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
22825 * fold-const.c (fold_vec_perm): Likewise.
22826 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
22827 * tree-vect-generic.c (lower_vec_perm): Likewise.
22828 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22829 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
22830 element type to HOST_WIDE_INT.
22832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22833 Alan Hayward <alan.hayward@arm.com>
22834 David Sherwood <david.sherwood@arm.com>
22836 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
22837 rather than an int. Use plus_constant.
22838 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
22839 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
22841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22842 Alan Hayward <alan.hayward@arm.com>
22843 David Sherwood <david.sherwood@arm.com>
22845 * calls.c (emit_call_1, expand_call): Change struct_value_size from
22846 a HOST_WIDE_INT to a poly_int64.
22848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22849 Alan Hayward <alan.hayward@arm.com>
22850 David Sherwood <david.sherwood@arm.com>
22852 * calls.c (load_register_parameters): Cope with polynomial
22853 mode sizes. Require a constant size for BLKmode parameters
22854 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
22855 forces a parameter to be padded at the lsb end in order to
22856 fill a complete number of words, require the parameter size
22857 to be ordered wrt UNITS_PER_WORD.
22859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22860 Alan Hayward <alan.hayward@arm.com>
22861 David Sherwood <david.sherwood@arm.com>
22863 * reload1.c (spill_stack_slot_width): Change element type
22864 from unsigned int to poly_uint64_pod.
22865 (alter_reg): Treat mode sizes as polynomial.
22867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22868 Alan Hayward <alan.hayward@arm.com>
22869 David Sherwood <david.sherwood@arm.com>
22871 * reload.c (complex_word_subreg_p): New function.
22872 (reload_inner_reg_of_subreg, push_reload): Use it.
22874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22875 Alan Hayward <alan.hayward@arm.com>
22876 David Sherwood <david.sherwood@arm.com>
22878 * lra-constraints.c (process_alt_operands): Reject matched
22879 operands whose sizes aren't ordered.
22880 (match_reload): Refer to this check here.
22882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22883 Alan Hayward <alan.hayward@arm.com>
22884 David Sherwood <david.sherwood@arm.com>
22886 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
22887 that the mode size is in the set {1, 2, 4, 8, 16}.
22889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22890 Alan Hayward <alan.hayward@arm.com>
22891 David Sherwood <david.sherwood@arm.com>
22893 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
22894 Use plus_constant instead of gen_rtx_PLUS.
22896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22897 Alan Hayward <alan.hayward@arm.com>
22898 David Sherwood <david.sherwood@arm.com>
22900 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
22901 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
22902 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
22903 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
22904 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
22905 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
22906 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
22907 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
22908 * config/i386/i386.c (ix86_push_rounding): ...this new function.
22909 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
22911 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
22912 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
22913 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
22914 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
22915 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
22916 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
22917 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
22918 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
22919 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
22920 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
22922 * expr.c (emit_move_resolve_push): Treat the input and result
22923 of PUSH_ROUNDING as a poly_int64.
22924 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
22925 (emit_push_insn): Likewise.
22926 * lra-eliminations.c (mark_not_eliminable): Likewise.
22927 * recog.c (push_operand): Likewise.
22928 * reload1.c (elimination_effects): Likewise.
22929 * rtlanal.c (nonzero_bits1): Likewise.
22930 * calls.c (store_one_arg): Likewise. Require the padding to be
22931 known at compile time.
22933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22934 Alan Hayward <alan.hayward@arm.com>
22935 David Sherwood <david.sherwood@arm.com>
22937 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
22938 Use plus_constant instead of gen_rtx_PLUS.
22940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22941 Alan Hayward <alan.hayward@arm.com>
22942 David Sherwood <david.sherwood@arm.com>
22944 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
22945 rather than an int.
22947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22948 Alan Hayward <alan.hayward@arm.com>
22949 David Sherwood <david.sherwood@arm.com>
22951 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
22952 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
22953 via stack temporaries. Treat the mode size as polynomial too.
22955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22956 Alan Hayward <alan.hayward@arm.com>
22957 David Sherwood <david.sherwood@arm.com>
22959 * expr.c (expand_expr_real_2): When handling conversions involving
22960 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
22961 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
22962 as a poly_uint64 too.
22964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22965 Alan Hayward <alan.hayward@arm.com>
22966 David Sherwood <david.sherwood@arm.com>
22968 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
22970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22971 Alan Hayward <alan.hayward@arm.com>
22972 David Sherwood <david.sherwood@arm.com>
22974 * combine.c (can_change_dest_mode): Handle polynomial
22975 REGMODE_NATURAL_SIZE.
22976 * expmed.c (store_bit_field_1): Likewise.
22977 * expr.c (store_constructor): Likewise.
22978 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
22979 and polynomial REGMODE_NATURAL_SIZE.
22980 (gen_lowpart_common): Likewise.
22981 * reginfo.c (record_subregs_of_mode): Likewise.
22982 * rtlanal.c (read_modify_subreg_p): Likewise.
22984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22985 Alan Hayward <alan.hayward@arm.com>
22986 David Sherwood <david.sherwood@arm.com>
22988 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
22989 numbers of elements.
22991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22992 Alan Hayward <alan.hayward@arm.com>
22993 David Sherwood <david.sherwood@arm.com>
22995 * match.pd: Cope with polynomial numbers of vector elements.
22997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22998 Alan Hayward <alan.hayward@arm.com>
22999 David Sherwood <david.sherwood@arm.com>
23001 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
23002 in a POINTER_PLUS_EXPR.
23004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23005 Alan Hayward <alan.hayward@arm.com>
23006 David Sherwood <david.sherwood@arm.com>
23008 * omp-simd-clone.c (simd_clone_subparts): New function.
23009 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
23010 (ipa_simd_modify_function_body): Likewise.
23012 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23013 Alan Hayward <alan.hayward@arm.com>
23014 David Sherwood <david.sherwood@arm.com>
23016 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
23017 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
23018 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
23019 (expand_vector_condition, vector_element): Likewise.
23020 (subparts_gt): New function.
23021 (get_compute_type): Use subparts_gt.
23022 (count_type_subparts): Delete.
23023 (expand_vector_operations_1): Use subparts_gt instead of
23024 count_type_subparts.
23026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23027 Alan Hayward <alan.hayward@arm.com>
23028 David Sherwood <david.sherwood@arm.com>
23030 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
23031 (vect_compile_time_alias): ...this new function. Do the calculation
23032 on poly_ints rather than trees.
23033 (vect_prune_runtime_alias_test_list): Update call accordingly.
23035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23036 Alan Hayward <alan.hayward@arm.com>
23037 David Sherwood <david.sherwood@arm.com>
23039 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
23041 (vect_schedule_slp_instance): Likewise.
23043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23044 Alan Hayward <alan.hayward@arm.com>
23045 David Sherwood <david.sherwood@arm.com>
23047 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
23048 constant and extern definitions for variable-length vectors.
23049 (vect_get_constant_vectors): Note that the number of units
23050 is known to be constant.
23052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23053 Alan Hayward <alan.hayward@arm.com>
23054 David Sherwood <david.sherwood@arm.com>
23056 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
23057 of units as polynomial. Choose between WIDE and NARROW based
23060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23061 Alan Hayward <alan.hayward@arm.com>
23062 David Sherwood <david.sherwood@arm.com>
23064 * tree-vect-stmts.c (simd_clone_subparts): New function.
23065 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
23067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23068 Alan Hayward <alan.hayward@arm.com>
23069 David Sherwood <david.sherwood@arm.com>
23071 * tree-vect-stmts.c (vectorizable_call): Treat the number of
23072 vectors as polynomial. Use build_index_vector for
23073 IFN_GOMP_SIMD_LANE.
23075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23076 Alan Hayward <alan.hayward@arm.com>
23077 David Sherwood <david.sherwood@arm.com>
23079 * tree-vect-stmts.c (get_load_store_type): Treat the number of
23080 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
23081 for variable-length vectors.
23082 (vectorizable_mask_load_store): Treat the number of units as
23083 polynomial, asserting that it is constant if the condition has
23084 already been enforced.
23085 (vectorizable_store, vectorizable_load): Likewise.
23087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23088 Alan Hayward <alan.hayward@arm.com>
23089 David Sherwood <david.sherwood@arm.com>
23091 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
23092 of units as polynomial. Punt if we can't tell at compile time
23093 which vector contains the final result.
23095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23096 Alan Hayward <alan.hayward@arm.com>
23097 David Sherwood <david.sherwood@arm.com>
23099 * tree-vect-loop.c (vectorizable_induction): Treat the number
23100 of units as polynomial. Punt on SLP inductions. Use an integer
23101 VEC_SERIES_EXPR for variable-length integer reductions. Use a
23102 cast of such a series for variable-length floating-point
23105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23106 Alan Hayward <alan.hayward@arm.com>
23107 David Sherwood <david.sherwood@arm.com>
23109 * tree.h (build_index_vector): Declare.
23110 * tree.c (build_index_vector): New function.
23111 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
23112 of units as polynomial, forcibly converting it to a constant if
23113 vectorizable_reduction has already enforced the condition.
23114 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
23115 to create a {1,2,3,...} vector.
23116 (vectorizable_reduction): Treat the number of units as polynomial.
23117 Choose vectype_in based on the largest scalar element size rather
23118 than the smallest number of units. Enforce the restrictions
23121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23122 Alan Hayward <alan.hayward@arm.com>
23123 David Sherwood <david.sherwood@arm.com>
23125 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
23126 number of units as polynomial.
23128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23129 Alan Hayward <alan.hayward@arm.com>
23130 David Sherwood <david.sherwood@arm.com>
23132 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
23133 * target.def (autovectorize_vector_sizes): Return the vector sizes
23134 by pointer, using vector_sizes rather than a bitmask.
23135 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
23136 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
23137 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
23139 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
23140 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
23141 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
23142 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
23143 * omp-general.c (omp_max_vf): Likewise.
23144 * omp-low.c (omp_clause_aligned_alignment): Likewise.
23145 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
23146 * tree-vect-loop.c (vect_analyze_loop): Likewise.
23147 * tree-vect-slp.c (vect_slp_bb): Likewise.
23148 * doc/tm.texi: Regenerate.
23149 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
23151 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
23152 the vector size as a poly_uint64 rather than an unsigned int.
23153 (current_vector_size): Change from an unsigned int to a poly_uint64.
23154 (get_vectype_for_scalar_type): Update accordingly.
23155 * tree.h (build_truth_vector_type): Take the size and number of
23156 units as a poly_uint64 rather than an unsigned int.
23157 (build_vector_type): Add a temporary overload that takes
23158 the number of units as a poly_uint64 rather than an unsigned int.
23159 * tree.c (make_vector_type): Likewise.
23160 (build_truth_vector_type): Take the number of units as a poly_uint64
23161 rather than an unsigned int.
23163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23164 Alan Hayward <alan.hayward@arm.com>
23165 David Sherwood <david.sherwood@arm.com>
23167 * target.def (get_mask_mode): Take the number of units and length
23168 as poly_uint64s rather than unsigned ints.
23169 * targhooks.h (default_get_mask_mode): Update accordingly.
23170 * targhooks.c (default_get_mask_mode): Likewise.
23171 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
23172 * doc/tm.texi: Regenerate.
23174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23175 Alan Hayward <alan.hayward@arm.com>
23176 David Sherwood <david.sherwood@arm.com>
23178 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
23179 * omp-general.c (omp_max_vf): Likewise.
23180 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
23181 (expand_omp_simd): Handle polynomial safelen.
23182 * omp-low.c (omplow_simd_context): Add a default constructor.
23183 (omplow_simd_context::max_vf): Change from int to poly_uint64.
23184 (lower_rec_simd_input_clauses): Update accordingly.
23185 (lower_rec_input_clauses): Likewise.
23187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23188 Alan Hayward <alan.hayward@arm.com>
23189 David Sherwood <david.sherwood@arm.com>
23191 * tree-vectorizer.h (vect_nunits_for_cost): New function.
23192 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
23193 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
23194 (vect_analyze_slp_cost): Likewise.
23195 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
23196 (vect_model_load_cost): Likewise.
23198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23199 Alan Hayward <alan.hayward@arm.com>
23200 David Sherwood <david.sherwood@arm.com>
23202 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
23203 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
23204 from an unsigned int * to a poly_uint64_pod *.
23205 (calculate_unrolling_factor): New function.
23206 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
23208 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23209 Alan Hayward <alan.hayward@arm.com>
23210 David Sherwood <david.sherwood@arm.com>
23212 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
23213 from an unsigned int to a poly_uint64.
23214 (_loop_vec_info::slp_unrolling_factor): Likewise.
23215 (_loop_vec_info::vectorization_factor): Change from an int
23217 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
23218 (vect_get_num_vectors): New function.
23219 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
23220 (vect_get_num_copies): Use vect_get_num_vectors.
23221 (vect_analyze_data_ref_dependences): Change max_vf from an int *
23222 to an unsigned int *.
23223 (vect_analyze_data_refs): Change min_vf from an int * to a
23225 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
23226 than an unsigned HOST_WIDE_INT.
23227 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
23228 (vect_analyze_data_ref_dependence): Change max_vf from an int *
23229 to an unsigned int *.
23230 (vect_analyze_data_ref_dependences): Likewise.
23231 (vect_compute_data_ref_alignment): Handle polynomial vf.
23232 (vect_enhance_data_refs_alignment): Likewise.
23233 (vect_prune_runtime_alias_test_list): Likewise.
23234 (vect_shift_permute_load_chain): Likewise.
23235 (vect_supportable_dr_alignment): Likewise.
23236 (dependence_distance_ge_vf): Take the vectorization factor as a
23237 poly_uint64 rather than an unsigned HOST_WIDE_INT.
23238 (vect_analyze_data_refs): Change min_vf from an int * to a
23240 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
23241 vfm1 as a poly_uint64 rather than an int. Make the same change
23242 for the returned bound_scalar.
23243 (vect_gen_vector_loop_niters): Handle polynomial vf.
23244 (vect_do_peeling): Likewise. Update call to
23245 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
23246 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
23248 * tree-vect-loop.c (vect_determine_vectorization_factor)
23249 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
23250 (vect_get_known_peeling_cost): Likewise.
23251 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
23252 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
23253 (vect_transform_loop): Likewise. Use the lowest possible VF when
23254 updating the upper bounds of the loop.
23255 (vect_min_worthwhile_factor): Make static. Return an unsigned int
23256 rather than an int.
23257 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
23258 polynomial unroll factors.
23259 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
23260 (vect_make_slp_decision): Likewise.
23261 (vect_supported_load_permutation_p): Likewise, and polynomial
23263 (vect_analyze_slp_cost): Handle polynomial vf.
23264 (vect_slp_analyze_node_operations): Likewise.
23265 (vect_slp_analyze_bb_1): Likewise.
23266 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
23267 than an unsigned HOST_WIDE_INT.
23268 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
23269 (vectorizable_load): Handle polynomial vf.
23270 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
23272 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
23274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23275 Alan Hayward <alan.hayward@arm.com>
23276 David Sherwood <david.sherwood@arm.com>
23278 * match.pd: Handle bit operations involving three constants
23279 and try to fold one pair.
23281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23283 * tree-vect-loop-manip.c: Include gimple-fold.h.
23284 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
23285 niters_maybe_zero parameters. Handle other cases besides a step of 1.
23286 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
23287 Add a path that uses a step of VF instead of 1, but disable it
23289 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
23290 and niters_no_overflow parameters. Update calls to
23291 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
23292 Create a new SSA name if the latter choses to use a ste other
23293 than zero, and return it via niters_vector_mult_vf_var.
23294 * tree-vect-loop.c (vect_transform_loop): Update calls to
23295 vect_do_peeling, vect_gen_vector_loop_niters and
23296 slpeel_make_loop_iterate_ntimes.
23297 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
23298 (vect_gen_vector_loop_niters): Update declarations after above changes.
23300 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
23302 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
23303 128-bit round to integer instructions.
23304 (ceil<mode>2): Likewise.
23305 (btrunc<mode>2): Likewise.
23306 (round<mode>2): Likewise.
23308 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23310 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
23311 unaligned VSX load/store on P8/P9.
23312 (expand_block_clear): Allow the use of unaligned VSX
23313 load/store on P8/P9.
23315 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23317 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
23319 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
23320 swap associated with both a load and a store.
23322 2018-01-02 Andrew Waterman <andrew@sifive.com>
23324 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
23325 * config/riscv/riscv.md (clear_cache): Use it.
23327 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
23329 * web.c: Remove out-of-date comment.
23331 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23333 * expr.c (fixup_args_size_notes): Check that any existing
23334 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
23335 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
23336 (emit_single_push_insn): ...here.
23338 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23340 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
23341 (const_vector_encoded_nelts): New function.
23342 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
23343 (const_vector_int_elt, const_vector_elt): Declare.
23344 * emit-rtl.c (const_vector_int_elt_1): New function.
23345 (const_vector_elt): Likewise.
23346 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
23347 of CONST_VECTOR_ELT.
23349 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23351 * expr.c: Include rtx-vector-builder.h.
23352 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
23353 directly on the tree encoding.
23354 (const_vector_from_tree): Likewise.
23355 * optabs.c: Include rtx-vector-builder.h.
23356 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
23357 sequence of "u" values.
23358 * vec-perm-indices.c: Include rtx-vector-builder.h.
23359 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
23360 directly on the vec_perm_indices encoding.
23362 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23364 * doc/rtl.texi (const_vector): Describe new encoding scheme.
23365 * Makefile.in (OBJS): Add rtx-vector-builder.o.
23366 * rtx-vector-builder.h: New file.
23367 * rtx-vector-builder.c: Likewise.
23368 * rtl.h (rtx_def::u2): Add a const_vector field.
23369 (CONST_VECTOR_NPATTERNS): New macro.
23370 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
23371 (CONST_VECTOR_DUPLICATE_P): Likewise.
23372 (CONST_VECTOR_STEPPED_P): Likewise.
23373 (CONST_VECTOR_ENCODED_ELT): Likewise.
23374 (const_vec_duplicate_p): Check for a duplicated vector encoding.
23375 (unwrap_const_vec_duplicate): Likewise.
23376 (const_vec_series_p): Check for a non-duplicated vector encoding.
23377 Say that the function only returns true for integer vectors.
23378 * emit-rtl.c: Include rtx-vector-builder.h.
23379 (gen_const_vec_duplicate_1): Delete.
23380 (gen_const_vector): Call gen_const_vec_duplicate instead of
23381 gen_const_vec_duplicate_1.
23382 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
23383 (gen_const_vec_duplicate): Use rtx_vector_builder.
23384 (gen_const_vec_series): Likewise.
23385 (gen_rtx_CONST_VECTOR): Likewise.
23386 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
23387 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
23388 Build a new vector rather than modifying a CONST_VECTOR in-place.
23389 (handle_special_swappables): Update call accordingly.
23390 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
23391 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
23392 Build a new vector rather than modifying a CONST_VECTOR in-place.
23393 (handle_special_swappables): Update call accordingly.
23395 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23397 * simplify-rtx.c (simplify_const_binary_operation): Use
23398 CONST_VECTOR_ELT instead of XVECEXP.
23400 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23402 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
23403 the selector elements to be different from the data elements
23404 if the selector is a VECTOR_CST.
23405 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
23406 ssizetype for the selector.
23408 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23410 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
23411 before testing each element individually.
23412 * tree-vect-generic.c (lower_vec_perm): Likewise.
23414 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23416 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
23417 * selftest-run-tests.c (selftest::run_tests): Call it.
23418 * vector-builder.h (vector_builder::operator ==): New function.
23419 (vector_builder::operator !=): Likewise.
23420 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
23421 (vec_perm_indices::all_from_input_p): New function.
23422 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
23423 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
23424 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
23425 instead of reading the VECTOR_CST directly. Detect whether both
23426 vector inputs are the same before constructing the vec_perm_indices,
23427 and update the number of inputs argument accordingly. Use the
23428 utility functions added above. Only construct sel2 if we need to.
23430 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23432 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
23433 the broadcast of the low byte.
23434 (expand_mult_highpart): Use an explicit encoding for the permutes.
23435 * optabs-query.c (can_mult_highpart_p): Likewise.
23436 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
23437 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23438 (vectorizable_bswap): Likewise.
23439 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
23440 explicit encoding for the power-of-2 permutes.
23441 (vect_permute_store_chain): Likewise.
23442 (vect_grouped_load_supported): Likewise.
23443 (vect_permute_load_chain): Likewise.
23445 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23447 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
23448 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
23449 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
23450 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
23451 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
23452 (vect_gen_perm_mask_any): Likewise.
23454 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23456 * int-vector-builder.h: New file.
23457 * vec-perm-indices.h: Include int-vector-builder.h.
23458 (vec_perm_indices): Redefine as an int_vector_builder.
23459 (auto_vec_perm_indices): Delete.
23460 (vec_perm_builder): Redefine as a stand-alone class.
23461 (vec_perm_indices::vec_perm_indices): New function.
23462 (vec_perm_indices::clamp): Likewise.
23463 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
23464 (vec_perm_indices::new_vector): New function.
23465 (vec_perm_indices::new_expanded_vector): Update for new
23466 vec_perm_indices class.
23467 (vec_perm_indices::rotate_inputs): New function.
23468 (vec_perm_indices::all_in_range_p): Operate directly on the
23469 encoded form, without computing elided elements.
23470 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
23471 encoding. Update for new vec_perm_indices class.
23472 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
23473 the given vec_perm_builder.
23474 (expand_vec_perm_var): Update vec_perm_builder constructor.
23475 (expand_mult_highpart): Use vec_perm_builder instead of
23476 auto_vec_perm_indices.
23477 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
23478 vec_perm_indices instead of auto_vec_perm_indices. Use a single
23479 or double series encoding as appropriate.
23480 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
23481 vec_perm_indices instead of auto_vec_perm_indices.
23482 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23483 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23484 (vect_permute_store_chain): Likewise.
23485 (vect_grouped_load_supported): Likewise.
23486 (vect_permute_load_chain): Likewise.
23487 (vect_shift_permute_load_chain): Likewise.
23488 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23489 (vect_transform_slp_perm_load): Likewise.
23490 (vect_schedule_slp_instance): Likewise.
23491 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23492 (vectorizable_mask_load_store): Likewise.
23493 (vectorizable_bswap): Likewise.
23494 (vectorizable_store): Likewise.
23495 (vectorizable_load): Likewise.
23496 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
23497 vec_perm_indices instead of auto_vec_perm_indices. Use
23498 tree_to_vec_perm_builder to read the vector from a tree.
23499 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
23500 vec_perm_builder instead of a vec_perm_indices.
23501 (have_whole_vector_shift): Use vec_perm_builder and
23502 vec_perm_indices instead of auto_vec_perm_indices. Leave the
23503 truncation to calc_vec_perm_mask_for_shift.
23504 (vect_create_epilog_for_reduction): Likewise.
23505 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
23506 from auto_vec_perm_indices to vec_perm_indices.
23507 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
23508 instead of changing individual elements.
23509 (aarch64_vectorize_vec_perm_const): Use new_vector to install
23510 the vector in d.perm.
23511 * config/arm/arm.c (expand_vec_perm_d::perm): Change
23512 from auto_vec_perm_indices to vec_perm_indices.
23513 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
23514 instead of changing individual elements.
23515 (arm_vectorize_vec_perm_const): Use new_vector to install
23516 the vector in d.perm.
23517 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
23518 Update vec_perm_builder constructor.
23519 (rs6000_expand_interleave): Likewise.
23520 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
23521 (rs6000_expand_interleave): Likewise.
23523 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23525 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
23526 to qimode could truncate the indices.
23527 * optabs.c (expand_vec_perm_var): Likewise.
23529 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23531 * Makefile.in (OBJS): Add vec-perm-indices.o.
23532 * vec-perm-indices.h: New file.
23533 * vec-perm-indices.c: Likewise.
23534 * target.h (vec_perm_indices): Replace with a forward class
23536 (auto_vec_perm_indices): Move to vec-perm-indices.h.
23537 * optabs.h: Include vec-perm-indices.h.
23538 (expand_vec_perm): Delete.
23539 (selector_fits_mode_p, expand_vec_perm_var): Declare.
23540 (expand_vec_perm_const): Declare.
23541 * target.def (vec_perm_const_ok): Replace with...
23542 (vec_perm_const): ...this new hook.
23543 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
23544 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
23545 * doc/tm.texi: Regenerate.
23546 * optabs.def (vec_perm_const): Delete.
23547 * doc/md.texi (vec_perm_const): Likewise.
23548 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
23549 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
23550 expand_vec_perm for constant permutation vectors. Assert that
23551 the mode of variable permutation vectors is the integer equivalent
23552 of the mode that is being permuted.
23553 * optabs-query.h (selector_fits_mode_p): Declare.
23554 * optabs-query.c: Include vec-perm-indices.h.
23555 (selector_fits_mode_p): New function.
23556 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
23557 is defined, instead of checking whether the vec_perm_const_optab
23558 exists. Use targetm.vectorize.vec_perm_const instead of
23559 targetm.vectorize.vec_perm_const_ok. Check whether the indices
23560 fit in the vector mode before using a variable permute.
23561 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
23562 vec_perm_indices instead of an rtx.
23563 (expand_vec_perm): Replace with...
23564 (expand_vec_perm_const): ...this new function. Take the selector
23565 as a vec_perm_indices rather than an rtx. Also take the mode of
23566 the selector. Update call to shift_amt_for_vec_perm_mask.
23567 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
23568 Use vec_perm_indices::new_expanded_vector to expand the original
23569 selector into bytes. Check whether the indices fit in the vector
23570 mode before using a variable permute.
23571 (expand_vec_perm_var): Make global.
23572 (expand_mult_highpart): Use expand_vec_perm_const.
23573 * fold-const.c: Includes vec-perm-indices.h.
23574 * tree-ssa-forwprop.c: Likewise.
23575 * tree-vect-data-refs.c: Likewise.
23576 * tree-vect-generic.c: Likewise.
23577 * tree-vect-loop.c: Likewise.
23578 * tree-vect-slp.c: Likewise.
23579 * tree-vect-stmts.c: Likewise.
23580 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
23582 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
23583 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
23584 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
23585 (aarch64_vectorize_vec_perm_const): ...this new function.
23586 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23587 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23588 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
23589 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
23590 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23591 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23592 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
23594 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
23595 check for NEON modes.
23596 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
23597 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
23598 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
23599 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
23601 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
23602 the old VEC_PERM_CONST conditions.
23603 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
23604 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
23605 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
23606 (ia64_vectorize_vec_perm_const_ok): Merge into...
23607 (ia64_vectorize_vec_perm_const): ...this new function.
23608 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
23609 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
23610 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
23611 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
23612 * config/mips/mips.c (mips_expand_vec_perm_const)
23613 (mips_vectorize_vec_perm_const_ok): Merge into...
23614 (mips_vectorize_vec_perm_const): ...this new function.
23615 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
23616 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
23617 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
23618 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
23619 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
23620 (rs6000_expand_vec_perm_const): Delete.
23621 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
23623 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23624 (altivec_expand_vec_perm_const_le): Take each operand individually.
23625 Operate on constant selectors rather than rtxes.
23626 (altivec_expand_vec_perm_const): Likewise. Update call to
23627 altivec_expand_vec_perm_const_le.
23628 (rs6000_expand_vec_perm_const): Delete.
23629 (rs6000_vectorize_vec_perm_const_ok): Delete.
23630 (rs6000_vectorize_vec_perm_const): New function.
23631 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23632 an element count and rtx array.
23633 (rs6000_expand_extract_even): Update call accordingly.
23634 (rs6000_expand_interleave): Likewise.
23635 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
23636 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
23637 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
23638 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
23639 (rs6000_expand_vec_perm_const): Delete.
23640 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23641 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23642 (altivec_expand_vec_perm_const_le): Take each operand individually.
23643 Operate on constant selectors rather than rtxes.
23644 (altivec_expand_vec_perm_const): Likewise. Update call to
23645 altivec_expand_vec_perm_const_le.
23646 (rs6000_expand_vec_perm_const): Delete.
23647 (rs6000_vectorize_vec_perm_const_ok): Delete.
23648 (rs6000_vectorize_vec_perm_const): New function. Remove stray
23649 reference to the SPE evmerge intructions.
23650 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23651 an element count and rtx array.
23652 (rs6000_expand_extract_even): Update call accordingly.
23653 (rs6000_expand_interleave): Likewise.
23654 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
23655 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
23657 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23659 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23661 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
23662 vector mode and that that mode matches the mode of the data
23664 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
23665 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
23666 directly using expand_vec_perm_1 when forcing selectors into
23668 (expand_vec_perm_var): New function, split out from expand_vec_perm.
23670 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23672 * optabs-query.h (can_vec_perm_p): Delete.
23673 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
23674 * optabs-query.c (can_vec_perm_p): Split into...
23675 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
23676 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
23677 particular selector is valid.
23678 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23679 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23680 (vect_grouped_load_supported): Likewise.
23681 (vect_shift_permute_load_chain): Likewise.
23682 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23683 (vect_transform_slp_perm_load): Likewise.
23684 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23685 (vectorizable_bswap): Likewise.
23686 (vect_gen_perm_mask_checked): Likewise.
23687 * fold-const.c (fold_ternary_loc): Likewise. Don't take
23688 implementations of variable permutation vectors into account
23689 when deciding which selector to use.
23690 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
23691 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
23692 with a false third argument.
23693 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
23694 to test whether the constant selector is valid and can_vec_perm_var_p
23695 to test whether a variable selector is valid.
23697 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23699 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
23700 * optabs-query.c (can_vec_perm_p): Likewise.
23701 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
23702 instead of vec_perm_indices.
23703 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
23704 (vect_gen_perm_mask_checked): Likewise,
23705 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
23706 (vect_gen_perm_mask_checked): Likewise,
23708 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23710 * optabs-query.h (qimode_for_vec_perm): Declare.
23711 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
23712 (qimode_for_vec_perm): ...this new function.
23713 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
23715 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23717 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
23718 does not have a conditional at the top.
23720 2018-01-02 Richard Biener <rguenther@suse.de>
23722 * ipa-inline.c (big_speedup_p): Fix expression.
23724 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23727 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
23730 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23734 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
23735 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
23736 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
23737 cond_taken_branch_cost 3->4.
23739 2018-01-01 Jakub Jelinek <jakub@redhat.com>
23741 PR tree-optimization/83581
23742 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
23743 TODO_cleanup_cfg if any changes have been made.
23745 PR middle-end/83608
23746 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
23747 convert_modes if target mode has the right side, but different mode
23750 PR middle-end/83609
23751 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
23752 last argument when extracting from CONCAT. If either from_real or
23753 from_imag is NULL, use expansion through memory. If result is not
23754 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
23755 the parts directly to inner mode, if even that fails, use expansion
23758 PR middle-end/83623
23759 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
23760 check for bswap in mode rather than HImode and use that in expand_unop
23763 Copyright (C) 2018 Free Software Foundation, Inc.
23765 Copying and distribution of this file, with or without modification,
23766 are permitted in any medium without royalty provided the copyright
23767 notice and this notice are preserved.