1 2018-08-21 Jan Hubicka <jh@suse.cz>
3 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
4 * tree.h (is_redundant_typedef): Remove.
5 * dwarf2out.c (is_redundant_typedef): Turn into static function.
7 2018-08-21 Jan Hubicka <jh@suse.cz>
9 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
12 2018-08-21 Tamar Christina <tamar.christina@arm.com>
14 * expmed.c (extract_low_bits): Reject invalid subregs early.
16 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
19 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
22 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
24 * config/vxworks.h: Guard vxworks_asm_out_constructor and
25 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
26 * config/vxworks.c: Likewise.
28 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
30 * config/vxworks.c: Set targetm.have_ctors_dtors
31 if HAVE_INITFINI_ARRAY_SUPPORT.
32 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
33 if HAVE_INITFINI_ARRAY_SUPPORT.
35 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
37 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
38 default search path for VxWorks < 7.
40 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
42 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
43 (get_format_string): Refer to c_getstr.
45 2018-08-21 Tom de Vries <tdevries@suse.de>
47 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
48 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
49 (debuginfo_early_stop): Declare.
50 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
51 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
52 (debuginfo_early_stop): New function.
53 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
54 and debuginfo_early_stop.
55 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
56 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
57 (general_init): Call debuginfo_early_init.
58 (finalize): Call debuginfo_fini.
59 (do_compile): Call debuginfo_init.
60 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
62 (@item -fdump-debug, @item -fdump-earlydebug): Add.
64 2018-08-21 Tom de Vries <tdevries@suse.de>
66 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
67 flag_dump_noaddr and flag_dump_unnumbered.
69 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
71 * wide-int-range.cc (wide_int_range_abs): New.
72 (wide_int_range_order_set): Rename from wide_int_range_min_max.
73 * wide-int-range.h (wide_int_range_abs): New.
74 (wide_int_range_min_max): New.
75 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
76 case to call wide_int_range_abs.
77 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
78 (extract_range_from_abs_expr): Delete.
80 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
83 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
84 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
87 2018-08-20 Nathan Sidwell <nathan@acm.org>
88 Jeff Law <law@redhat.com>
90 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
91 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
93 2018-08-20 David Malcolm <dmalcolm@redhat.com>
96 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
97 (decl_attributes): Likewise.
98 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
100 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
101 * diagnostic-core.h (class auto_diagnostic_group): New class.
102 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
103 (diagnostic_report_diagnostic): Handle the first diagnostics within
105 (emit_diagnostic): Add auto_diagnostic_group instance.
107 (inform_n): Likewise.
109 (warning_at): Likewise.
110 (warning_n): Likewise.
112 (permerror): Likewise.
115 (error_at): Likewise.
117 (fatal_error): Likewise.
118 (internal_error): Likewise.
119 (internal_error_no_backtrace): Likewise.
120 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
121 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
122 * diagnostic.h (struct diagnostic_context): Add fields
123 "diagnostic_group_nesting_depth",
124 "diagnostic_group_emission_count", "begin_group_cb",
126 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
127 Add auto_diagnostic_group instance(s).
128 (find_explicit_erroneous_behavior): Likewise.
129 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
130 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
131 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
132 (gimplify_va_arg_expr): Likewise.
133 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
134 (HSA_SORRY_AT): Likewise.
135 * ipa-devirt.c (compare_virtual_tables): Likewise.
136 (warn_odr): Likewise.
137 * multiple_target.c (expand_target_clones): Likewise.
138 * opts-common.c (cmdline_handle_error): Likewise.
139 * reginfo.c (globalize_reg): Likewise.
140 * substring-locations.c (format_warning_n_va): Likewise.
141 * tree-inline.c (expand_call_inline): Likewise.
142 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
143 * tree-ssa-loop-niter.c
144 (do_warn_aggressive_loop_optimizations): Likewise.
145 * tree-ssa-uninit.c (warn_uninit): Likewise.
146 * tree.c (warn_deprecated_use): Likewise.
148 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
151 * config/i386/i386.md (eh_return): Always update EH return
152 address in word_mode.
154 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
156 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
157 TARGET_SPLIT_COMPLEX_ARG is defined.
159 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
161 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
163 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
166 * expr.c (expand_assignment): Assert that bitpos is positive.
167 (store_field): Likewise
168 (expand_expr_real_1): Make sure that bitpos is positive.
169 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
172 2018-08-20 Nathan Sidwell <nathan@acm.org>
174 * Makefile.in (CPP_ID_DATA_H): Delete.
175 (CPP_INTERNAL_H): Don't add it.
176 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
177 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
179 2018-08-20 Richard Biener <rguenther@suse.de>
181 PR tree-optimization/78655
182 * tree-vrp.c (extract_range_from_binary_expr_1): Make
183 pointer + offset nonnull if either operand is nonnull work.
185 2018-08-20 Tom de Vries <tdevries@suse.de>
187 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
188 unless the referenced die describes the added property using
189 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
190 Otherwise, add a DW_AT_location to the referenced die.
192 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
195 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
196 register_operand when calling ix86_set_reg_reg_cost.
197 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
198 Set *total to 0 for operands that satisfy x86_64_immediate_operand
199 predicate and to 1 otherwise.
201 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
203 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
204 emit a diagnostic that it is not supported and reset the option.
205 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
206 supported and consume the option. (ASM_FINAL_SPEC): New.
208 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
210 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
213 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
215 C-SKY port: Documentation
217 * doc/extend.texi (C-SKY Function Attributes): New section.
218 * doc/invoke.texi (Option Summary): Add C-SKY options.
219 (C-SKY Options): New section.
220 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
222 2018-08-17 Jojo <jijie_rong@c-sky.com>
223 Huibin Wang <huibin_wang@c-sky.com>
224 Sandra Loosemore <sandra@codesourcery.com>
225 Chung-Lin Tang <cltang@codesourcery.com>
227 C-SKY port: Backend implementation
229 * config/csky/*: New.
230 * common/config/csky/*: New.
232 2018-08-17 Jojo <jijie_rong@c-sky.com>
233 Huibin Wang <huibin_wang@c-sky.com>
234 Sandra Loosemore <sandra@codesourcery.com>
235 Chung-Lin Tang <cltang@codesourcery.com>
236 Andrew Jenner <andrew@codesourcery.com>
238 C-SKY port: Configury
240 * config.gcc (csky-*-*): New.
241 * configure.ac: Add csky to targets for dwarf2 debug_line support.
242 * configure: Regenerated.
244 2018-08-17 David Malcolm <dmalcolm@redhat.com>
246 * dump-context.h: Include "dumpfile.h".
247 (dump_context::dump_printf_va): Convert final param from va_list
248 to va_list *. Convert from ATTRIBUTE_PRINTF to
249 ATTRIBUTE_GCC_DUMP_PRINTF.
250 (dump_context::dump_printf_loc_va): Likewise.
251 * dumpfile.c: Include "stringpool.h".
252 (make_item_for_dump_printf_va): Delete.
253 (make_item_for_dump_printf): Delete.
254 (class dump_pretty_printer): New class.
255 (dump_pretty_printer::dump_pretty_printer): New ctor.
256 (dump_pretty_printer::emit_items): New member function.
257 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
259 (dump_pretty_printer::emit_item): New member function.
260 (dump_pretty_printer::stash_item): New member function.
261 (dump_pretty_printer::format_decoder_cb): New member function.
262 (dump_pretty_printer::decode_format): New member function.
263 (dump_context::dump_printf_va): Reimplement in terms of
265 (dump_context::dump_printf_loc_va): Convert final param from va_list
267 (dump_context::begin_scope): Reimplement call to
268 make_item_for_dump_printf.
269 (dump_printf): Update for change to dump_printf_va.
270 (dump_printf_loc): Likewise.
271 (selftest::test_capture_of_dump_calls): Convert "stmt" from
272 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
274 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
275 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
276 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
277 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
278 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
279 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
280 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
281 within a dump_printf_loc call to "%wu".
282 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
283 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
284 missing space after "=".
285 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
286 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
287 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
288 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
289 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
290 duplicate "vectorized" from message.
292 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
294 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
295 polyNxK_t element's TYPE_STRING_FLAG.
297 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
299 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
300 (they were unnamed before). Fix comments.
302 2018-08-17 Nathan Sidwell <nathan@acm.org>
304 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
306 2018-08-17 Richard Biener <rguenther@suse.de>
308 PR tree-optimization/86841
309 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
311 2018-08-17 Martin Liska <mliska@suse.cz>
313 * common.opt: Remove Warn, Init and Report for options with
314 Ignore/Deprecated flag. Warning is done automatically for
316 * config/i386/i386.opt: Likewise.
317 * config/ia64/ia64.opt: Likewise.
318 * config/rs6000/rs6000.opt: Likewise.
319 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
320 Remove usage of flag_check_pointer_bounds.
321 * lto-wrapper.c (merge_and_complain): Do not handle
322 OPT_fcheck_pointer_bounds.
323 (append_compiler_options): Likewise.
324 * opt-functions.awk: Do not handle Deprecated.
325 * optc-gen.awk: Check that Var, Report and Init are not
326 used for an option with Ignore/Deprecated flag.
327 * opts-common.c (decode_cmdline_option): Do not report
329 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
331 * opts.h (struct cl_option): Remove cl_deprecated flag.
332 (CL_ERR_DEPRECATED): Remove error enum value.
334 2018-08-17 Richard Biener <rguenther@suse.de>
337 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
338 across a va-arg-pack using call adjust its return value accordingly.
340 2018-08-16 Martin Sebor <msebor@redhat.com>
342 PR tree-optimization/86853
343 * gimple-ssa-sprintf.c (struct format_result): Rename member.
344 (struct fmtresult): Add member and initialize it in ctors.
345 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
346 (format_string): Handle %S the same as %ls. Set MAYFAIL.
347 (format_directive): Set POSUNDER4K when MAYFAIL is set.
348 (parse_directive): Handle %C same as %c.
349 (sprintf_dom_walker::compute_format_length): Adjust.
350 (is_call_safe): Adjust.
352 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
354 * builtins.c (c_strlen): Add new parameter eltsize. Use it
355 for determining how to count the elements.
356 * builtins.h (c_strlen): Adjust prototype.
357 * expr.c (string_constant): Add new parameter mem_size.
358 Set *mem_size appropriately.
359 * expr.h (string_constant): Adjust protoype.
360 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
361 * gimple-fold.h (get_range_strlen): Adjust prototype.
362 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
363 (format_string): Call get_string_length with eltsize.
365 2018-08-16 David Malcolm <dmalcolm@redhat.com>
367 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
368 to emit the span, rather than setting it as the prefix.
370 2018-08-16 David Malcolm <dmalcolm@redhat.com>
372 * diagnostic-show-locus.c (layout::start_annotation_line): Add
373 "margin_char" parameter, defaulting to space. Use it in place
374 of pp_space for the initial part of the margin.
375 (layout::print_leading_fixits): Use '+' when filling the margin
376 of line-insertion fix-it hints.
378 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
380 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
383 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
385 * config/rs6000/altivec.md: Don't set length attribute to the default
387 * config/rs6000/darwin.md: Ditto.
388 * config/rs6000/dfp.md: Ditto.
389 * config/rs6000/htm.md: Ditto.
390 * config/rs6000/rs6000.md: Ditto.
391 * config/rs6000/sync.md: Ditto.
392 * config/rs6000/vsx.md: Ditto.
394 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
396 * config/rs6000/altivec.md: Don't set length attribute to the default
397 value, for branch instructions.
398 * config/rs6000/darwin.md: Ditto.
399 * config/rs6000/rs6000.md: Ditto.
401 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
403 * config/rs6000/rs6000.md (length): Always define as const_int 4.
404 (unnamed conditional branch define_insn): Set length to 4 or 8
406 (<bd>_<mode>): Similar, for alternative 0.
407 (<bd>tf_<mode>): Ditto.
409 2018-08-16 Tamar Christina <tamar.christina@arm.com>
411 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
413 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
415 * doc/rtl.texi: Replace old RTX class names with new names.
418 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
420 * expmed.h (canonicalize_comparison): New declaration.
421 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
422 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
423 * optabs.c (prepare_cmp_insn): Likewise.
424 * rtl.h (unsigned_condition_p): New function which checks if a
425 comparison operator is unsigned.
427 2018-08-16 Nathan Sidwell <nathan@acm.org>
429 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
430 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
432 2018-08-16 Tamar Christina <tamar.christina@arm.com>
435 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
436 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
437 (mov<mov>): ..this and enable unconditionally.
439 2018-08-16 Tamar Christina <tamar.christina@arm.com>
441 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
443 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
445 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
446 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
447 "Common" with "Target".
449 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
451 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
452 * doc/invoke.texi (mmitigate-rop): Remove.
453 * config/i386/i386.c: Do not include "regrename.h".
454 (ix86_rop_should_change_byte_p, reg_encoded_number)
455 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
457 (ix86_reorg): Remove call to ix86_mitigate_rop.
458 * config/i386/i386.md (attr "modrm_class"): Remove.
459 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
460 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
461 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
463 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
465 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
466 allow folding of mergeh() and mergel() for the float and double types.
467 (fold_mergehl_helper): Rework to handle building a permute tree
470 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
472 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
475 2018-08-15 David Malcolm <dmalcolm@redhat.com>
477 * common.opt (fdiagnostics-show-labels): New option.
478 * diagnostic-show-locus.c (class layout_range): Add field
480 (class layout): Add field "m_show_labels_p".
481 (layout_range::layout_range): Add param "label" and use it to
483 (make_range): Pass in NULL for new "label" param of layout_range's
485 (layout::layout): Initialize m_show_labels_p.
486 (layout::maybe_add_location_range): Pass in loc_range->m_label
487 when constructing layout_range instances.
488 (struct line_label): New struct.
489 (layout::print_any_labels): New member function.
490 (layout::print_line): Call it if label-printing is enabled.
491 (selftest::test_one_liner_labels): New test.
492 (selftest::test_diagnostic_show_locus_one_liner): Call it.
493 * diagnostic.c (diagnostic_initialize): Initialize
494 context->show_labels_p.
495 * diagnostic.h (struct diagnostic_context): Add field
497 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
498 -fno-diagnostics-show-labels.
499 * dwarf2out.c (gen_producer_string): Add
500 OPT_fdiagnostics_show_labels to the ignored options.
501 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
503 (gcc_rich_location::maybe_add_expr): Likewise.
504 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
505 label" param, defaulting to NULL.
506 (gcc_rich_location::add_expr): Add "label" param.
507 (gcc_rich_location::maybe_add_expr): Likewise.
508 (class text_range_label): New class.
509 (class range_label_for_type_mismatch): New class.
510 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
511 of format_warning_va.
512 (fmtwarn_n): Likewise for new params of format_warning_n_va.
513 * lto-wrapper.c (merge_and_complain): Add
514 OPT_fdiagnostics_show_labels to the "pick one setting" options.
515 (append_compiler_options): Likewise to the dropped options.
516 (append_diag_options): Likewise to the passed-on options.
517 * opts.c (common_handle_option): Handle the new option.
518 * selftest-diagnostic.c
519 (test_diagnostic_context::test_diagnostic_context): Enable
521 * substring-locations.c: Include "gcc-rich-location.h".
522 (format_warning_n_va): Add "fmt_label" and "param_label" params
523 and use them as appropriate.
524 (format_warning_va): Add "fmt_label" and "param_label" params,
525 passing them on to format_warning_n_va.
526 (format_warning_at_substring): Likewise.
527 (format_warning_at_substring_n): Likewise.
528 * substring-locations.h (format_warning_va): Add "fmt_label" and
529 "param_label" params.
530 (format_warning_n_va): Likewise.
531 (format_warning_at_substring): Likewise.
532 (format_warning_at_substring_n): Likewise.
533 * toplev.c (general_init): Initialize global_dc->show_labels_p.
535 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
538 * builtins.c (expand_builtin_memcmp): Do not expand the call
539 when overflow is detected.
541 2018-08-15 Martin Sebor <msebor@redhat.com>
543 PR tree-optimization/71625
544 * config/aarch64/aarch64-builtins.c
545 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
547 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
549 * config/s390/s390.c (s390_reorg): Remove loop.
551 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
554 (darwin_function_switched_text_sections): Delete.
555 * gcc/config/darwin.h
556 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
558 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
561 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
562 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
563 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
565 2018-08-15 Martin Liska <mliska@suse.cz>
567 PR tree-optimization/86925
568 * predict.c (expr_expected_value_1): When taking
569 later predictor, assign also probability.
570 Use fold_build2_initializer_loc in order to fold
571 the expression in -frounding-math.
573 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
575 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
577 (expand_vec_perm_1): Try the new method.
579 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
582 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
583 Check whether lra_live_max_point is 0 before dividing.
585 2018-08-14 Martin Sebor <msebor@redhat.com>
587 PR tree-optimization/86650
588 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
589 (vrp_prop::check_mem_ref): Same.
591 2018-08-13 Liu Hao <lh_mouse@126.com>
593 * pretty-print.c (eat_esc_sequence): Swap the foreground and
594 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
595 and clear it thereafter, as it only works for DBCS.
597 2018-08-13 Liu Hao <lh_mouse@126.com>
599 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
600 handle returned by _get_osf_handle().
602 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
604 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
605 for folding vec_perm.
607 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
609 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
610 Add support for gimple-folding of vec_pack() and vec_unpack()
613 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
615 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
616 vec_xst variants to the list.
617 (rs6000_gimple_fold_builtin): Add support for folding unaligned
618 vector loads and stores.
620 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
622 * config.gcc (rs6000-ibm-aix4.x): Delete.
623 (rs6000-ibm-aix5.1): Delete.
624 (rs6000-ibm-aix5.2): Delete.
625 (rs6000-ibm-aix5.3): Delete.
626 * config/rs6000/aix43.h: Delete.
627 * config/rs6000/aix51.h: Delete.
628 * config/rs6000/aix52.h: Delete.
629 * config/rs6000/t-aix43: Delete.
631 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
633 * config/s390/s390.c (s390_decompose_constant_pool_ref):
635 (s390_decompose_address): Factor out constant pool ref
638 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
640 * config/nds32/nds32-predicates.c
641 (nds32_can_use_bclr_p): Change return type as bool.
642 (nds32_can_use_bset_p): Ditto.
643 (nds32_can_use_btgl_p): Ditto.
644 (nds32_can_use_bitci_p): Ditto.
645 * config/nds32/nds32-protos.h
646 (nds32_can_use_bclr_p): Change declaration.
647 (nds32_can_use_bset_p): Ditto.
648 (nds32_can_use_btgl_p): Ditto.
649 (nds32_can_use_bitci_p): Ditto.
651 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
653 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
654 Support -msched-prolog-epilog option.
655 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
657 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
659 * common/config/nds32/nds32-common.c
660 (nds32_option_optimization_table): Enalbe -malways-align.
662 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
664 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
666 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
667 OPT_misr_secure_ case.
668 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
669 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
670 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
672 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
673 (nds32_isr_info): New field security_level.
674 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
675 * config/nds32/nds32.md (return_internal): Consider critical attribute.
676 * config/nds32/nds32.opt (misr-secure): New option.
677 * config/nds32/nds32_init.inc: New file.
678 * config/nds32/nds32_isr.h: New file.
680 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
682 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
683 Update comment for atomic instructions.
684 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
685 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
687 (atomic_loaddi): Revise fence expansion to only emit fence prior to
688 load for __ATOMIC_SEQ_CST model.
689 (atomic_loaddi_1): Remove float register target.
690 (atomic_storedi): Handle CONST_INT values.
691 (atomic_storedi_1): Remove float register source. Add special case
693 (memory_barrier): New expander and insn.
695 2018-08-11 Jakub Jelinek <jakub@redhat.com>
697 PR tree-optimization/86835
698 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
699 new_stmt after def_gsi, make sure to insert new_square_stmt after
700 that stmt, not 2 stmts before it.
702 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
705 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
708 2018-08-10 Martin Liska <mliska@suse.cz>
711 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
713 * builtins.c (expand_builtin_expect_with_probability):
715 (expand_builtin_expect_with_probability): New function.
716 (build_builtin_expect_predicate): Add new argumnet probability
717 for BUILT_IN_EXPECT_WITH_PROBABILITY.
718 (fold_builtin_expect):
721 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
722 * builtins.h (fold_builtin_expect): Set new argument.
723 * doc/extend.texi: Document __builtin_expect_with_probability.
724 * doc/invoke.texi: Likewise.
725 * gimple-fold.c (gimple_fold_call): Pass new argument.
726 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
727 also BUILT_IN_EXPECT_WITH_PROBABILITY.
728 * predict.c (get_predictor_value): New function.
729 (expr_expected_value): Add new argument probability. Assume
730 that predictor and probability are always non-null.
731 (expr_expected_value_1): Likewise. For __builtin_expect and
732 __builtin_expect_with_probability set probability. Handle
733 combination in binary expressions.
734 (tree_predict_by_opcode): Simplify code by simply calling
736 (pass_strip_predict_hints::execute): Add handling of
737 BUILT_IN_EXPECT_WITH_PROBABILITY.
738 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
740 * tree.h (DECL_BUILT_IN_P): New function.
742 2018-08-10 Martin Liska <mliska@suse.cz>
744 PR tree-optimization/85799
745 * passes.def: Add argument for pass_strip_predict_hints.
746 * predict.c (class pass_strip_predict_hints): Add new argument
748 (strip_predictor_early): New function.
749 (pass_strip_predict_hints::execute): Call the function to
751 (strip_predict_hints): New function.
752 * predict.def: Fix comment.
754 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
756 * Makefile.in: Clarify which tm.texi to copy over to assert the
757 right to grant a GFDL license for all.
759 2018-08-09 Jeff Law <law@redhat.com>
761 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
764 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
766 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
769 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
771 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
772 reductions for variable-length vectors.
774 2018-08-09 David Malcolm <dmalcolm@redhat.com>
777 * common.opt (fdiagnostics-show-line-numbers): New option.
778 * diagnostic-show-locus.c (class layout): Add fields
779 "m_show_line_numbers_p" and "m_linenum_width";
780 (num_digits): New function.
781 (test_num_digits): New function.
782 (layout::layout): Initialize new fields. Update m_x_offset
783 logic to handle any left margin.
784 (layout::print_source_line): Print line number when requested.
785 (layout::start_annotation_line): New member function.
786 (layout::print_annotation_line): Call it.
787 (layout::print_leading_fixits): Likewise.
788 (layout::print_trailing_fixits): Likewise. Update calls to
789 move_to_column for new parameter.
790 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
791 it to potentially call start_annotation_line.
792 (layout::show_ruler): Call start_annotation_line.
793 (selftest::test_line_numbers_multiline_range): New selftest.
794 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
795 and selftest::test_line_numbers_multiline_range.
796 * diagnostic.c (diagnostic_initialize): Initialize
798 * diagnostic.h (struct diagnostic_context): Add field
799 "show_line_numbers_p".
800 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
801 -fno-diagnostics-show-line-numbers.
802 * dwarf2out.c (gen_producer_string): Add
803 OPT_fdiagnostics_show_line_numbers to the ignored options.
804 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
805 one setting" options.
806 (append_compiler_options): Likewise to the dropped options.
807 (append_diag_options): Likewise to the passed-on options.
808 * opts.c (common_handle_option): Handle the new option.
809 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
811 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
813 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
814 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
815 third argument of type "const signed char" to descriptions of
816 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
817 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
818 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
819 __builtin_bcdsub_ov functions.
821 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
823 PR tree-optimization/86858
824 * tree-vect-loop.c (vect_is_simple_reduction): Restore
825 flow_bb_inside_loop_p calls.
827 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
829 PR tree-optimization/86871
830 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
831 instead of gimple_assign_lhs.
833 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
836 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
837 register constraint to operand 0.
838 (add<mode>3_carryinC): Likewise.
839 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
841 2018-08-09 Martin Liska <mliska@suse.cz>
844 * common.opt: Remove extra line.
846 2018-08-09 Martin Liska <mliska@suse.cz>
848 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
849 at the end of a line, make first letter capital and end up
850 a sentence with a dot.
851 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
852 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
853 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
854 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
855 (PARAM_MAX_ISL_OPERATIONS): Likewise.
856 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
857 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
858 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
859 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
860 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
861 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
862 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
863 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
864 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
865 (PARAM_TREE_REASSOC_WIDTH): Likewise.
866 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
867 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
868 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
870 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
873 * config/s390/s390.c (s390_option_override_internal): Reduce the
874 stack-clash-protection-probe-interval param if it would be too big
877 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
880 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
881 * config/m68k/m68k.c (handle_move_double): Don't call
882 m68k_final_prescan_insn.
883 (m68k_adjust_decorated_operand): Renamed from
884 m68k_final_prescan_insn, remove first and third operand and
886 (print_operand): Call it.
887 (print_operand_address): Call it.
889 2018-08-08 Nathan Sidwell <nathan@acm.org>
891 * diagnostic.c (diagnostic_report_current_module): Use
892 linemap_included_from & linemap_included_from_linemap.
894 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
896 * config/aarch64/aarch64-cores.def: Add phecda core.
897 * config/aarch64/aarch64-tune.md: Regenerate.
898 * doc/invoke.texi: Add phecda core.
900 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
903 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
905 * config/s390/s390.md ("movti"): Add more alternatives for
906 constant to GPR copies.
908 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
910 * config/s390/s390.c: Fix whitespace damage throughout the file.
911 * config/s390/s390.h: Likewise.
912 * config/s390/tpf.h: Likewise.
914 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
916 * config/s390/s390.c (s390_loadrelative_operand_p):
917 Remove TARGET_CPU_ZARCH usages.
918 (s390_rtx_costs): Likewise.
919 (s390_legitimate_constant_p): Likewise.
920 (s390_cannot_force_const_mem): Likewise.
921 (legitimate_reload_constant_p): Likewise.
922 (s390_preferred_reload_class): Likewise.
923 (legitimize_pic_address): Likewise.
924 (legitimize_tls_address): Likewise.
925 (s390_split_branches): Removed.
926 (s390_add_execute): Removed.
927 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
928 (s390_mainpool_start): Likewise.
929 (s390_mainpool_finish): Likewise.
930 (s390_mainpool_cancel): Removed.
931 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
932 (s390_chunkify_cancel): Likewise.
933 (s390_return_addr_rtx): Likewise.
934 (s390_register_info): Remove split_branches_pending_p uages.
935 (s390_optimize_register_info): Likewise.
936 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
937 split_branches_pending_p usages.
938 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
939 (s390_load_got): Likewise.
940 (s390_expand_split_stack_prologue): Likewise.
941 (output_asm_nops): Likewise.
942 (s390_function_profiler): Likewise.
943 (s390_emit_call): Likewise.
944 (s390_conditional_register_usage): Likewise.
945 (s390_optimize_prologue): Likewise.
946 (s390_reorg): Remove TARGET_CPU_ZARCH and
947 split_branches_pending_p usages.
948 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
950 (s390_output_indirect_thunk_function): Likewise.
951 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
952 (TARGET_CPU_ZARCH_P): Removed.
953 (struct machine_function): Remove split_branches_pending_p.
954 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
956 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
958 * common/config/s390/s390-common.c (processor_flags_table):
960 * config.gcc: Remove with_arch/with_tune support.
961 * config/s390/2064.md: Remove cpu attribute comparisons.
962 * config/s390/driver-native.c (s390_host_detect_local_cpu):
964 * config/s390/linux.h (ASM_SPEC):
965 Remove -march support.
966 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
967 Use a table to get an arch level.
968 * config/s390/s390-opts.h (enum processor_type):
971 (processor_table): Remove entries, add arch_level values.
972 (s390_issue_rate): Remove cases.
973 (s390_option_override): Adjust
974 s390_option_override_internal() call.
975 (s390_option_override_internal): Remove deprecation warning.
976 (s390_valid_target_attribute_tree): Adjust
977 s390_option_override_internal() call.
978 * config/s390/s390.h (struct s390_processor):
979 Share with s390-c.c, add arch_level field.
980 * config/s390/s390.md:
981 Remove occurrences in cpu attribute.
982 * config/s390/s390.opt: Remove -march/-mtune support.
983 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
984 * doc/invoke.texi: Remove deprecation warning.
986 2018-08-08 Luis Machado <luis.machado@linaro.org>
988 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
990 (qdf24xx_tunings): Set vector cost structure to
993 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
994 <register_sextend>: Set to 3.
996 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
999 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
1000 * config/aarch64/aarch64-simd.md
1001 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
1002 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
1003 (aarch64_frecpx<mode>): ...this new pattern.
1004 * config/aarch64/aarch64-simd-builtins.def: Remove comment
1005 about aarch64_frecp<FRECP:frecp_suffix><mode>.
1007 2018-08-07 Martin Liska <mliska@suse.cz>
1010 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
1011 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
1012 * predict.def (PRED_MALLOC_NONNULL): New predictor.
1013 * doc/extend.texi: Document that malloc attribute adds
1016 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
1019 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1020 Define to speculation_safe_value_not_needed.
1022 2018-08-06 Jeff Law <law@redhat.com>
1024 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
1025 the vr_values instance to cprop_into_stmt.
1026 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
1027 (cprop_operand): Also query EVRP to determine if OP is a constant.
1029 2018-08-06 Nathan Sidwell <nathan@acm.org>
1031 * diagnostic.c (diagnostic_report_current_module): Reroll
1032 included-at loop. Translate text.
1034 2018-08-06 David Malcolm <dmalcolm@redhat.com>
1036 * function-tests.c (selftest::test_expansion_to_rtl): Call
1037 free_after_compilation.
1039 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1041 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
1043 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
1045 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
1046 loops with memory block operations from getting unrolled.
1048 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
1051 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1052 Define to speculation_safe_value_not_needed.
1054 2018-08-06 Jeff Law <law@redhat.com>
1056 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
1059 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1062 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
1063 with all enabled __intN types.
1065 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
1067 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1069 * alias.c (record_set): Check for clobber high.
1070 * cfgexpand.c (expand_gimple_stmt): Likewise.
1071 * combine-stack-adj.c (single_set_for_csa): Likewise.
1072 * combine.c (find_single_use_1): Likewise.
1073 (set_nonzero_bits_and_sign_copies): Likewise.
1074 (get_combine_src_dest): Likewise.
1075 (is_parallel_of_n_reg_sets): Likewise.
1076 (try_combine): Likewise.
1077 (record_dead_and_set_regs_1): Likewise.
1078 (reg_dead_at_p_1): Likewise.
1079 (reg_dead_at_p): Likewise.
1080 * dce.c (deletable_insn_p): Likewise.
1081 (mark_nonreg_stores_1): Likewise.
1082 (mark_nonreg_stores_2): Likewise.
1083 * df-scan.c (df_find_hard_reg_defs): Likewise.
1084 (df_uses_record): Likewise.
1085 (df_get_call_refs): Likewise.
1086 * dwarf2out.c (mem_loc_descriptor): Likewise.
1087 * haifa-sched.c (haifa_classify_rtx): Likewise.
1088 * ira-build.c (create_insn_allocnos): Likewise.
1089 * ira-costs.c (scan_one_insn): Likewise.
1090 * ira.c (equiv_init_movable_p): Likewise.
1091 (rtx_moveable_p): Likewise.
1092 (interesting_dest_for_shprep): Likewise.
1093 * jump.c (mark_jump_label_1): Likewise.
1094 * postreload-gcse.c (record_opr_changes): Likewise.
1095 * postreload.c (reload_cse_simplify): Likewise.
1096 (struct reg_use): Add source expr.
1097 (reload_combine): Check for clobber high.
1098 (reload_combine_note_use): Likewise.
1099 (reload_cse_move2add): Likewise.
1100 (move2add_note_store): Likewise.
1101 * print-rtl.c (print_pattern): Likewise.
1102 * recog.c (decode_asm_operands): Likewise.
1103 (store_data_bypass_p): Likewise.
1104 (if_test_bypass_p): Likewise.
1105 * regcprop.c (kill_clobbered_value): Likewise.
1106 (kill_set_value): Likewise.
1107 * reginfo.c (reg_scan_mark_refs): Likewise.
1108 * reload1.c (maybe_fix_stack_asms): Likewise.
1109 (eliminate_regs_1): Likewise.
1110 (elimination_effects): Likewise.
1111 (mark_not_eliminable): Likewise.
1112 (scan_paradoxical_subregs): Likewise.
1113 (forget_old_reloads_1): Likewise.
1114 * reorg.c (find_end_label): Likewise.
1115 (try_merge_delay_insns): Likewise.
1116 (redundant_insn): Likewise.
1117 (own_thread_p): Likewise.
1118 (fill_simple_delay_slots): Likewise.
1119 (fill_slots_from_thread): Likewise.
1120 (dbr_schedule): Likewise.
1121 * resource.c (update_live_status): Likewise.
1122 (mark_referenced_resources): Likewise.
1123 (mark_set_resources): Likewise.
1124 * rtl.c (copy_rtx): Likewise.
1125 * rtlanal.c (reg_referenced_p): Likewise.
1126 (single_set_2): Likewise.
1127 (noop_move_p): Likewise.
1128 (note_stores): Likewise.
1129 * sched-deps.c (sched_analyze_reg): Likewise.
1130 (sched_analyze_insn): Likewise.
1132 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1134 * cse.c (invalidate_reg): New function extracted from...
1135 (invalidate): ...here.
1136 (canonicalize_insn): Check for clobber high.
1137 (invalidate_from_clobbers): invalidate clobber highs.
1138 (invalidate_from_sets_and_clobbers): Likewise.
1139 (count_reg_usage): Check for clobber high.
1140 (insn_live_p): Likewise.
1141 * cselib.c (cselib_expand_value_rtx_1):Likewise.
1142 (cselib_invalidate_regno): Check for clobber in setter.
1143 (cselib_invalidate_rtx): Pass through setter.
1144 (cselib_invalidate_rtx_note_stores):
1145 (cselib_process_insn): Check for clobber high.
1146 * cselib.h (cselib_invalidate_rtx): Add operand.
1148 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1150 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
1151 (mark_not_eliminable): Likewise.
1152 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
1153 * lra-lives.c (process_bb_lives): Check for clobber high.
1154 * lra.c (new_insn_reg): Remember clobber highs.
1155 (collect_non_operand_hard_regs): Check for clobber high.
1156 (lra_set_insn_recog_data): Likewise.
1157 (add_regs_to_insn_regno_info): Likewise.
1158 (lra_update_insn_regno_info): Likewise.
1160 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1162 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
1163 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
1165 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1167 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
1168 (copy_insn_1): Likewise.
1169 (gen_hard_reg_clobber_high): New gen function.
1170 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
1171 * genemit.c (gen_exp): Likewise.
1172 (gen_emit_seq): Pass through info.
1173 (gen_insn): Check for CLOBBER_HIGH.
1174 (gen_expand): Pass through info.
1175 (gen_split): Likewise.
1176 (output_add_clobbers): Likewise.
1177 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
1178 (remove_clobbers): Likewise.
1179 * rtl.h (gen_hard_reg_clobber_high): New declaration.
1181 2018-08-06 Alan Hayward <alan.hayward@arm.com>
1183 * doc/rtl.texi (clobber_high): Add.
1184 (parallel): Add in clobber high
1185 * rtl.c (rtl_check_failed_code3): Add function.
1186 * rtl.def (CLOBBER_HIGH): Add expression.
1187 * rtl.h (RTL_CHECKC3): Add macro.
1188 (rtl_check_failed_code3): Add declaration.
1189 (XC3EXP): Add macro.
1191 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
1194 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
1195 cfun->machine->max_used_stack_alignment if needed.
1197 2018-08-04 Martin Sebor <msebor@redhat.com>
1199 PR tree-optimization/86571
1200 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
1203 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
1206 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1209 2018-08-03 Jeff Law <law@redhat.com>
1212 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1213 Define to speculation_safe_value_not_needed.
1215 2018-08-03 David Malcolm <dmalcolm@redhat.com>
1217 * doc/gcov.texi (-x): Remove duplicate "to".
1218 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
1219 (-Wif-not-aligned): Remove duplicate "is".
1220 (-flto): Remove duplicate "the".
1221 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
1222 duplicate "v5.00.b".
1223 (MSP430 Options): Remove duplicate "and" from the description
1224 of "-mgprel-sec=regexp".
1225 (x86 Options): Remove duplicate copies of "vmldLog102" and
1226 vmlsLog104 from description of "-mveclibabi=type".
1228 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
1230 * internal-fn.h (first_commutative_argument): Declare.
1231 * internal-fn.c (first_commutative_argument): New function.
1232 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
1233 restrictions for pattern statements. Use first_commutative_argument
1234 to look for commutative operands in calls to internal functions.
1236 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
1238 * Makefile.in (wide-int-range.o): New.
1239 * tree-vrp.c: Move all the wide_int_* functions to...
1240 * wide-int-range.cc: ...here.
1241 * tree-vrp.h: Move all the wide_int_* prototypes to...
1242 * wide-int-range.h: ...here.
1244 2018-08-03 Tom de Vries <tdevries@suse.de>
1246 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
1248 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
1249 * except.c (output_function_exception_table): Do early exit if
1250 targetm_common.except_unwind_info (&global_options) == UI_NONE.
1252 2018-08-03 Martin Liska <mliska@suse.cz>
1254 * predict.c (dump_prediction): Change to 2 digits
1257 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
1259 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
1261 (falkor_am_1_gtov_gtov): ... a new insn reservation.
1263 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
1265 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
1266 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
1267 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
1268 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
1269 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
1271 2018-08-02 David Malcolm <dmalcolm@redhat.com>
1273 * diagnostic-show-locus.c (diagnostic_show_locus): Use
1274 pp_take_prefix when saving the existing prefix.
1275 * diagnostic.c (diagnostic_append_note): Likewise.
1276 * langhooks.c (lhd_print_error_function): Likewise.
1277 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
1278 param's type. Free the existing prefix.
1279 (pp_take_prefix): New function.
1280 (pretty_printer::pretty_printer): Drop the prefix parameter.
1281 Rename the length parameter to match the comment.
1282 (pretty_printer::~pretty_printer): Free the prefix.
1283 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
1285 (struct pretty_printer): Drop the "const" from "prefix" field's
1286 type and clarify memory management.
1287 (pp_set_prefix): Drop the "const" from the 2nd param.
1288 (pp_take_prefix): New decl.
1290 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
1292 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
1293 for word_mode_ok here instead of passing as argument.
1294 (expand_block_compare): Change select_block_compare_mode() call.
1295 (expand_strncmp_gpr_sequence): New function.
1296 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
1298 2018-08-02 Jeff Law <law@redhat.com>
1301 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1302 Define to speculation_safe_value_not_needed.
1305 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1306 Define to speculation_safe_value_not_needed.
1308 2018-08-02 Tom de Vries <tdevries@suse.de>
1311 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
1312 function. Return UI_TARGET unconditionally.
1313 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
1314 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
1316 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1318 * genemit.c (print_overload_test): Fix typo.
1320 2018-08-02 Richard Biener <rguenther@suse.de>
1322 PR tree-optimization/86816
1323 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
1324 which checks for value availability before querying it.
1325 (gvn_uses_equal): Use it.
1326 (same_succ_hash): Likewise.
1327 (gimple_equal_p): Likewise.
1329 2018-08-02 Nick Clifton <nickc@redhat.com>
1332 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1333 Define to speculation_safe_value_not_needed.
1336 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1337 Define to speculation_safe_value_not_needed.
1340 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1341 Define to speculation_safe_value_not_needed.
1344 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1345 Define to speculation_safe_value_not_needed.
1348 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1349 Define to speculation_safe_value_not_needed.
1352 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1353 Define to speculation_safe_value_not_needed.
1356 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1357 Define to speculation_safe_value_not_needed.
1360 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1361 Define to speculation_safe_value_not_needed.
1364 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1365 speculation_safe_value_not_needed.
1368 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
1369 to speculation_safe_value_not_needed.
1371 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1373 * doc/md.texi: Expand the documentation of instruction names
1374 to mention port-local uses. Document '@' in pattern names.
1375 * read-md.h (overloaded_instance, overloaded_name): New structs.
1377 (md_reader::handle_overloaded_name): New member function.
1378 (md_reader::get_overloads): Likewise.
1379 (md_reader::m_first_overload): New member variable.
1380 (md_reader::m_next_overload_ptr): Likewise.
1381 (md_reader::m_overloads_htab): Likewise.
1382 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
1383 m_next_overload_ptr and m_overloads_htab.
1384 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
1385 (get_mode_token, get_code_token, get_int_token): New functions.
1386 (map_attr_string): Add an optional argument that passes back
1387 the associated iterator.
1388 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
1389 (md_reader::handle_overloaded_name, add_overload_instance): New
1391 (apply_iterators): Handle '@' names. Report an error if '@'
1392 is used without iterators.
1393 (initialize_iterators): Initialize the new iterator_group fields.
1394 * genopinit.c (handle_overloaded_code_for)
1395 (handle_overloaded_gen): New functions.
1396 (main): Use them to print declarations of maybe_code_for_* and
1397 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
1398 * genemit.c (print_overload_arguments, print_overload_test)
1399 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
1400 (main): Use it to print definitions of maybe_code_for_* and
1401 maybe_gen_* functions.
1402 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
1403 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
1404 instead of explicit mode checks.
1405 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
1406 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
1407 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
1408 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
1409 (aarch64_expand_compare_and_swap): Likewise
1410 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
1411 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
1412 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
1413 (aarch64_constant_pool_reload_icode): Delete.
1414 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
1415 instead of aarch64_constant_pool_reload_icode. Use
1416 code_for_aarch64_reload_mov instead of explicit mode checks.
1417 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
1418 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
1419 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
1420 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
1421 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
1422 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
1423 (aarch64_atomic_load_op_code): Delete.
1424 (aarch64_emit_atomic_load_op): Likewise.
1425 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
1426 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
1427 instead of aarch64_emit_atomic_load_op.
1428 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
1429 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
1430 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
1431 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
1432 character before the pattern name.
1433 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
1434 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
1435 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
1436 (aarch64_frecps<mode>): Likewise.
1437 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
1438 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
1439 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
1440 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
1441 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
1443 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
1445 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
1446 Allow HFmode constants if TARGET_FP_F16INST.
1448 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
1451 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
1452 No longer check last store for clobber of address register.
1454 2018-08-02 Martin Liska <mliska@suse.cz>
1456 PR gcov-profile/86817
1457 * gcov.c (process_all_functions): New function.
1459 (process_file): Move functions processing to
1460 process_all_functions.
1462 2018-08-02 David Malcolm <dmalcolm@redhat.com>
1464 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
1465 "const" to the "gimple *" and "rtx_insn *" parameters.
1466 * dumpfile.h (dump_user_location_t::dump_user_location_t):
1468 (dump_location_t::dump_location_t): Likewise.
1470 2018-08-01 Martin Sebor <msebor@redhat.com>
1472 PR tree-optimization/86650
1473 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
1474 rather than a "gcall *". Directly pass the data of interest
1475 to percent_K_format, rather than building a temporary CALL_EXPR
1477 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
1478 (gimple_fold_builtin_strncat): Adjust.
1479 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
1480 gcall* argument with gimple*.
1481 * gimple-ssa-warn-restrict.c (check_call): Same.
1482 (wrestrict_dom_walker::before_dom_children): Same.
1483 (builtin_access::builtin_access): Same.
1484 (check_bounds_or_overlap): Same
1485 (maybe_diag_overlap): Same.
1486 (maybe_diag_offset_bounds): Same.
1487 * tree-diagnostic.c (default_tree_printer): Move usage of
1488 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
1490 * tree-pretty-print.c (percent_K_format): Add argument.
1491 * tree-pretty-print.h: Add argument.
1492 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
1493 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
1494 (maybe_diag_stxncpy_trunc): Same.
1495 (handle_builtin_stxncpy): Same.
1496 (handle_builtin_strcat): Same.
1498 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1500 * match.pd: Optimise pointer range checks.
1502 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1504 PR tree-optimization/86758
1505 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
1506 to remove pattern statements.
1508 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1510 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
1511 result of dfs_enumerate_from when constructing stmt_vec_infos,
1512 instead of additionally calling get_loop_body.
1514 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1516 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
1518 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
1519 When creating an iv, assert that the step is not known to be zero.
1520 (vect_setup_realignment): Update call accordingly.
1521 * tree-vect-stmts.c (vectorizable_store): Likewise.
1522 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
1524 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1526 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
1527 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
1528 (vectorizable_reduction): Likewise.
1529 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
1530 (vect_detect_hybrid_slp_stmts): Likewise.
1531 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
1533 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
1535 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
1536 (wide_int_set_zero_nonzero_bits): ...this.
1537 (zero_nonzero_bits_from_vr): Rename to...
1538 (vrp_set_zero_nonzero_bits): ...this.
1539 (extract_range_from_multiplicative_op_1): Abstract wide int
1541 (wide_int_range_multiplicative_op): ...here.
1542 (extract_range_from_binary_expr_1): Extract wide int binary
1543 operations into their own functions.
1544 (wide_int_range_lshift): New.
1545 (wide_int_range_can_optimize_bit_op): New.
1546 (wide_int_range_shift_undefined_p): New.
1547 (wide_int_range_bit_xor): New.
1548 (wide_int_range_bit_ior): New.
1549 (wide_int_range_bit_and): New.
1550 (wide_int_range_trunc_mod): New.
1551 (extract_range_into_wide_ints): New.
1552 (vrp_shift_undefined_p): New.
1553 (extract_range_from_multiplicative_op): New.
1554 (vrp_can_optimize_bit_op): New.
1555 * tree-vrp.h (value_range::dump): New.
1556 (wide_int_range_multiplicative_op): New.
1557 (wide_int_range_lshift):New.
1558 (wide_int_range_shift_undefined_p): New.
1559 (wide_int_range_bit_xor): New.
1560 (wide_int_range_bit_ior): New.
1561 (wide_int_range_bit_and): New.
1562 (wide_int_range_trunc_mod): New.
1563 (zero_nonzero_bits_from_bounds): Rename to...
1564 (wide_int_set_zero_nonzero_bits): ...this.
1565 (zero_nonzero_bits_from_vr): Rename to...
1566 (vrp_set_zero_nonzero_bits): ...this.
1567 (range_easy_mask_min_max): Rename to...
1568 (wide_int_range_can_optimize_bit_op): this.
1569 * vr-values.c (simplify_bit_ops_using_ranges): Rename
1570 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
1572 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1574 * tree-vectorizer.h (vect_orig_stmt): New function.
1575 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
1576 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
1577 (vect_create_epilog_for_reduction): Likewise.
1578 (vectorizable_live_operation): Likewise.
1579 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
1580 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
1581 * tree-vect-stmts.c (vectorizable_call): Likewise.
1582 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
1584 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1586 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
1588 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
1589 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
1590 (vect_transform_loop): Likewise.
1591 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1593 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1595 * tree-vectorizer.h (vect_schedule_slp): Return void.
1596 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
1597 (vect_schedule_slp): Likewise.
1599 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1601 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
1603 (vect_transform_loop): Update calls accordingly. Schedule SLP
1604 instances before the main loop, if any exist.
1606 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
1608 PR tree-optimization/86749
1609 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
1610 If the lhs is used in a COND_EXPR, check that it is being used
1611 as the "then" or "else" value.
1613 2018-08-01 Tom de Vries <tdevries@suse.de>
1616 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
1617 speculation_safe_value_not_needed.
1619 2018-08-01 Richard Biener <rguenther@suse.de>
1621 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
1624 2018-08-01 Martin Liska <mliska@suse.cz>
1626 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
1627 format how successful transformation is dumped.
1628 (gimple_mod_pow2_value_transform): Likewise.
1629 (gimple_mod_subtract_transform): Likewise.
1630 (gimple_stringops_transform): Likewise.
1632 2018-08-01 Martin Liska <mliska@suse.cz>
1635 * value-prof.c (interesting_stringop_to_profile_p):
1636 Simplify the code and add BUILT_IN_MEMMOVE.
1637 (gimple_stringops_transform): Likewise.
1639 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
1641 * config/aarch64/aarch64-simd.md
1642 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
1643 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
1644 use GPI iterator instead of SI mode.
1646 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1648 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
1649 rs6000_speculation_barrier.
1650 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
1651 new barrier pattern name.
1653 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
1655 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
1656 (speculation_barrier): New insn.
1658 2018-08-01 Richard Biener <rguenther@suse.de>
1661 * graphite.h: Include isl/id.h and isl/space.h to allow build
1664 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
1667 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1668 mode for COFF targets.
1669 * defaults.h (TARGET_COFF): Define.
1670 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
1671 TARGET_COFF): Define.
1672 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
1673 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
1674 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
1676 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1678 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
1679 * gimple-streamer-out.c (output_bb): Save it.
1680 * lto-streamer-in.c (input_struct_function_base): Restore
1681 instance discriminator if available. Create map on demand.
1682 * lto-streamer-out.c (output_struct_function_base): Save it if
1684 * final.c (decl_to_instance_map): Document LTO strategy.
1686 2018-07-31 Alexandre Oliva <oliva@adacore.com>
1687 Olivier Hainque <hainque@adacore.com>
1689 * debug.h (decl_to_instance_map_t): New type.
1690 (decl_to_instance_map): Declare.
1691 (maybe_create_decl_to_instance_map): New inline function.
1692 * final.c (bb_discriminator, last_bb_discriminator): New statics,
1693 to track basic block discriminators.
1694 (final_start_function_1): Initialize them.
1695 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
1697 (decl_to_instance_map): New variable.
1698 (map_decl_to_instance, maybe_set_discriminator): New functions.
1699 (notice_source_line): Set discriminator.
1701 2018-07-31 Ian Lance Taylor <iant@golang.org>
1703 * targhooks.c (default_have_speculation_safe_value): Add
1706 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1708 * dump-context.h: Include "pretty-print.h".
1709 (dump_context::refresh_dumps_are_enabled): New decl.
1710 (dump_context::emit_item): New decl.
1711 (class dump_context): Add fields "m_test_pp" and
1713 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1714 (temp_dump_context::get_dumped_text): New decl.
1715 (class temp_dump_context): Add field "m_pp".
1716 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
1717 (dump_context::refresh_dumps_are_enabled): ...and add a test for
1719 (set_dump_file): Update for above change.
1720 (set_alt_dump_file): Likewise.
1721 (dump_loc): New overload, taking a pretty_printer *.
1722 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
1723 to any test pretty-printer.
1724 (make_item_for_dump_gimple_stmt): New function, adapted from
1725 optinfo::add_gimple_stmt.
1726 (dump_context::dump_gimple_stmt): Call it, and use the result,
1727 eliminating the direct usage of dump_file and alt_dump_file in
1728 favor of indirectly using them via emit_item.
1729 (make_item_for_dump_gimple_expr): New function, adapted from
1730 optinfo::add_gimple_expr.
1731 (dump_context::dump_gimple_expr): Call it, and use the result,
1732 eliminating the direct usage of dump_file and alt_dump_file in
1733 favor of indirectly using them via emit_item.
1734 (make_item_for_dump_generic_expr): New function, adapted from
1736 (dump_context::dump_generic_expr): Call it, and use the result,
1737 eliminating the direct usage of dump_file and alt_dump_file in
1738 favor of indirectly using them via emit_item.
1739 (make_item_for_dump_printf_va): New function, adapted from
1740 optinfo::add_printf_va.
1741 (make_item_for_dump_printf): New function.
1742 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
1743 and use the result, eliminating the direct usage of dump_file and
1744 alt_dump_file in favor of indirectly using them via emit_item.
1745 (make_item_for_dump_dec): New function.
1746 (dump_context::dump_dec): Call it, and use the result,
1747 eliminating the direct usage of dump_file and alt_dump_file in
1748 favor of indirectly using them via emit_item.
1749 (make_item_for_dump_symtab_node): New function, adapted from
1750 optinfo::add_symtab_node.
1751 (dump_context::dump_symtab_node): Call it, and use the result,
1752 eliminating the direct usage of dump_file and alt_dump_file in
1753 favor of indirectly using them via emit_item.
1754 (dump_context::begin_scope): Reimplement, avoiding direct usage
1755 of dump_file and alt_dump_file in favor of indirectly using them
1757 (dump_context::emit_item): New member function.
1758 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
1759 Set up test pretty-printer on the underlying context. Call
1760 refresh_dumps_are_enabled.
1761 (temp_dump_context::~temp_dump_context): Call
1762 refresh_dumps_are_enabled.
1763 (temp_dump_context::get_dumped_text): New member function.
1764 (selftest::verify_dumped_text): New function.
1765 (ASSERT_DUMPED_TEXT_EQ): New macro.
1766 (selftest::test_capture_of_dump_calls): Run all tests twice, with
1767 and then without optinfo enabled. Add uses of
1768 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
1769 * dumpfile.h: Update comment for the dump_* API.
1770 * optinfo-emit-json.cc
1771 (selftest::test_building_json_from_dump_calls): Update for new
1772 param for temp_dump_context ctor.
1773 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
1774 and "m_owned" field.
1775 (optinfo_item::~optinfo_item): Likewise.
1776 (optinfo::add_item): New member function.
1777 (optinfo::emit): Update comment.
1778 (optinfo::add_string): Delete.
1779 (optinfo::add_printf): Delete.
1780 (optinfo::add_printf_va): Delete.
1781 (optinfo::add_gimple_stmt): Delete.
1782 (optinfo::add_gimple_expr): Delete.
1783 (optinfo::add_tree): Delete.
1784 (optinfo::add_symtab_node): Delete.
1785 (optinfo::add_dec): Delete.
1786 * optinfo.h (class dump_context): New forward decl.
1787 (optinfo::add_item): New decl.
1788 (optinfo::add_string): Delete.
1789 (optinfo::add_printf): Delete.
1790 (optinfo::add_printf_va): Delete.
1791 (optinfo::add_gimple_stmt): Delete.
1792 (optinfo::add_gimple_expr): Delete.
1793 (optinfo::add_tree): Delete.
1794 (optinfo::add_symtab_node): Delete.
1795 (optinfo::add_dec): Delete.
1796 (optinfo::add_poly_int): Delete.
1797 (optinfo_item::optinfo_item): Remove "owned" param.
1798 (class optinfo_item): Remove field "m_owned".
1800 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1803 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
1804 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
1805 requested variable alignment.
1806 (expand_one_ssa_partition): Likewise.
1807 (expand_one_var): Likewise.
1809 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1811 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1812 to speculation_safe_value_not_needed.
1814 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1816 * targhooks.h (speculation_safe_value_not_needed): New prototype.
1817 * targhooks.c (speculation_safe_value_not_needed): New function.
1818 * target.def (have_speculation_safe_value): Update documentation.
1819 * doc/tm.texi: Regenerated.
1821 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1823 * config/aarch64/iterators.md (ALLI_TI): New iterator.
1824 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
1826 (despeculate_copy<ALLI:mode>_insn): New insn.
1827 (despeculate_copyti_insn): New insn.
1828 (despeculate_simple<ALLI:mode>): New insn
1829 (despeculate_simpleti): New insn.
1830 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
1832 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
1833 aarch64_speculation_safe_value.
1834 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
1836 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1838 * config/aarch64/aarch64-speculation.cc: New file.
1839 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
1840 before pass_reorder_blocks.
1841 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
1843 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
1844 X14 and X15 when tracking speculation.
1845 * config/aarch64/aarch64.md (register name constants): Add
1846 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
1847 (unspec): Add UNSPEC_SPECULATION_TRACKER.
1848 (speculation_barrier): New insn attribute.
1849 (cmp<mode>): Allow SP in comparisons.
1850 (speculation_tracker): New insn.
1851 (speculation_barrier): Add speculation_barrier attribute.
1852 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
1853 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
1854 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
1856 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1858 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
1859 aarch64_track_speculation is true.
1860 (tb<optab><mode>1): Likewise.
1861 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
1862 generate CB[N]Z when tracking speculation.
1863 (aarch64_split_compare_and_swap): Likewise.
1864 (aarch64_split_atomic_op): Likewise.
1866 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1868 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
1870 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1872 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
1873 (speculation_barrier): New insn.
1875 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1877 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
1878 * config/arm/arm.md (speculation_barrier): New expand.
1879 (speculation_barrier_insn): New pattern.
1881 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
1883 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
1884 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
1885 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
1886 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
1888 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
1889 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
1890 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
1891 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
1892 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
1893 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
1894 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
1895 * builtins.c (expand_speculation_safe_value): New function.
1896 (expand_builtin): Call it.
1897 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
1898 * doc/extend.texi: Document __builtin_speculation_safe_value.
1899 * doc/md.texi: Document "speculation_barrier" pattern.
1900 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
1901 TARGET_HAVE_SPECULATION_SAFE_VALUE.
1902 * doc/tm.texi: Regenerated.
1903 * target.def (have_speculation_safe_value, speculation_safe_value): New
1905 * targhooks.c (default_have_speculation_safe_value): New function.
1906 (default_speculation_safe_value): New function.
1907 * targhooks.h (default_have_speculation_safe_value): Add prototype.
1908 (default_speculation_safe_value): Add prototype.
1910 2018-07-31 David Malcolm <dmalcolm@redhat.com>
1912 * dump-context.h (dump_context::dump_loc): New decl.
1913 * dumpfile.c (dump_context::dump_loc): New member function.
1914 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
1915 and dump_gimple_stmt.
1916 (dump_context::dump_gimple_expr_loc): Likewise, using
1918 (dump_context::dump_generic_expr_loc): Likewise, using
1920 (dump_context::dump_printf_loc_va): Likewise, using
1922 (dump_context::begin_scope): Explicitly using the global function
1923 "dump_loc", rather than the member function.
1925 2018-07-31 Martin Sebor <msebor@redhat.com>
1927 PR tree-optimization/86741
1928 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
1930 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
1932 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
1933 depend on whether prefetch instructions will be emitted or not.
1934 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
1935 will be emitted or not.
1936 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
1937 (TARGET_SETMEM_PFD): New macros.
1939 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1941 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
1942 (NULL_STMT_VEC_INFO): Delete.
1943 (stmt_vec_info::operator*): Likewise.
1944 (stmt_vec_info::operator gimple *): Likewise.
1945 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
1946 of NULL_STMT_VEC_INFO.
1947 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
1948 (vect_reassociating_reduction_p): Likewise.
1949 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
1950 (vectorizable_store): Likewise.
1951 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
1952 (vec_info::free_stmt_vec_infos): Likewise.
1954 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1956 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
1957 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
1959 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1961 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
1962 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1963 (vec_info::free_stmt_vec_info): New private member functions.
1964 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
1965 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
1966 * tree-parloops.c (gather_scalar_reductions): Remove calls to
1967 set_stmt_vec_info_vec and free_stmt_vec_infos.
1968 * tree-vect-loop.c (_loop_vec_info): Remove call to
1969 set_stmt_vec_info_vec.
1970 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
1971 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
1972 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
1973 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
1974 (vec_info::free_stmt_vec_info): ...these new functions. Remove
1975 assignments in {vec_info::,}new_stmt_vec_info that are redundant
1976 with the clearing in the xcalloc.
1977 (stmt_vec_info_vec): Delete.
1978 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
1979 (vectorize_loops): Likewise.
1980 (vec_info::~vec_info): Remove argument from call to
1981 free_stmt_vec_infos.
1982 (vec_info::add_stmt): Remove vinfo argument from call to
1985 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1987 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
1988 rather than a gimple stmt.
1989 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
1990 information for pattern statements when passed the original
1991 statement; instead wait to be passed the pattern statement itself.
1992 Don't call set_vinfo_for_stmt here.
1993 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
1994 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
1995 stmt_vec_infos here.
1996 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
1997 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
1998 stmt_vec_infos entry.
2000 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2002 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
2003 * tree-vectorizer.c (vec_info::replace_stmt): New function.
2004 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
2005 * tree-vect-stmts.c (vectorizable_call): Likewise.
2006 (vectorizable_simd_clone_call): Likewise.
2008 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2010 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
2011 * tree-vectorizer.c (vec_info::remove_stmt): New function.
2012 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
2013 * tree-vect-loop.c (vect_transform_loop): Likewise.
2014 * tree-vect-slp.c (vect_schedule_slp): Likewise.
2015 * tree-vect-stmts.c (vect_remove_stores): Likewise.
2017 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2019 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
2020 (vect_dr_stmt): Delete.
2021 * tree-vectorizer.c (vec_info::lookup_dr): New function.
2022 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
2024 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
2025 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
2026 (vect_verify_datarefs_alignment, vect_peeling_supportable)
2027 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
2028 (vect_analyze_data_refs): Likewise.
2029 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
2031 (vect_find_same_alignment_drs): Likewise.
2032 (vect_slp_analyze_node_dependences): Update calls accordingly.
2033 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
2034 instead of DR_VECT_AUX.
2035 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
2036 of a vector data references. Use vec_info::lookup_dr instead of
2038 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
2039 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
2040 instead of DR_VECT_AUX.
2042 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2044 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
2046 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
2048 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2049 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2050 (vect_gen_prolog_loop_niters): Likewise.
2052 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2054 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
2055 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
2056 (vect_known_alignment_in_bytes, vect_dr_behavior)
2057 (vect_get_scalar_dr_size): Take references as dr_vec_infos
2058 instead of data_references. Update calls to other routines for
2059 which the same change has been made.
2060 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
2061 dr_vec_infos instead of stmt_vec_infos.
2062 (vect_analyze_data_ref_dependence): Update call accordingly.
2063 (vect_slp_analyze_data_ref_dependence)
2064 (vect_record_base_alignments): Use DR_VECT_AUX.
2065 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
2066 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
2067 (vector_alignment_reachable_p, vect_get_data_access_cost)
2068 (vect_peeling_supportable, vect_analyze_group_access_1)
2069 (vect_analyze_group_access, vect_analyze_data_ref_access)
2070 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
2071 (vect_compile_time_alias, vect_small_gap_p)
2072 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
2073 (vect_supportable_dr_alignment): Take references as dr_vec_infos
2074 instead of data_references. Update calls to other routines for
2075 which the same change has been made.
2076 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
2077 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
2078 (vect_slp_analyze_and_verify_node_alignment)
2079 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
2080 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
2081 (vect_setup_realignment): Use dr_vec_infos. Update calls after
2083 (_vect_peel_info::dr): Replace with...
2084 (_vect_peel_info::dr_info): ...this new field.
2085 (vect_peeling_hash_get_most_frequent)
2086 (vect_peeling_hash_choose_best_peeling): Update accordingly.
2087 (vect_peeling_hash_get_lowest_cost):
2088 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
2089 routines for which the same change has been made.
2090 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
2092 * tree-vect-loop-manip.c (get_misalign_in_elems)
2093 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
2095 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2096 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
2097 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
2098 (get_group_load_store_type, get_negative_load_store_type)
2099 (vect_get_data_ptr_increment, vectorizable_store)
2100 (vectorizable_load): Likewise.
2101 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
2102 Update calls to other routines for which the same change has been made.
2104 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2106 * tree-vectorizer.h (vec_info::move_dr): New member function.
2107 (dataref_aux): Rename to...
2108 (dr_vec_info): ...this and add "dr" and "stmt" fields.
2109 (_stmt_vec_info::dr_aux): Update accordingly.
2110 (_stmt_vec_info::data_ref_info): Delete.
2111 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
2112 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
2113 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
2114 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
2116 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
2117 (STMT_VINFO_DR_INFO): New macro.
2118 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
2119 (set_dr_misalignment): Update after rename of dataref_aux.
2120 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
2121 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
2122 initialization of STMT_VINFO_DATA_REF.
2123 * tree-vectorizer.c (vec_info::move_dr): New function.
2124 * tree-vect-patterns.c (vect_recog_bool_pattern)
2125 (vect_recog_mask_conversion_pattern)
2126 (vect_recog_gather_scatter_pattern): Use it.
2127 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
2128 the "dr" and "stmt" fields of dr_vec_info instead of
2129 STMT_VINFO_DATA_REF.
2131 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2133 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
2134 (is_pattern_stmt_p): Use it.
2135 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
2136 on pattern statements.
2138 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2140 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
2141 original stmt as a stmt_vec_info rather than a gimple stmt.
2142 (vect_pattern_recog_1): Take the statement directly as a
2143 stmt_vec_info, rather than via a gimple_stmt_iterator.
2144 Update call to vect_mark_pattern_stmts.
2145 (vect_pattern_recog): Update calls accordingly.
2147 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2149 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
2150 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
2151 a vect_def_type for the first argument.
2152 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
2153 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
2154 operand if it isn't defined by a vectorized statement.
2155 (vect_build_gather_load_calls): Remove the mask_dt argument and
2156 update calls to vect_get_vec_def_for_stmt_copy.
2157 (vectorizable_bswap): Likewise the dt argument.
2158 (vectorizable_call): Update calls to vectorizable_bswap and
2159 vect_get_vec_def_for_stmt_copy.
2160 (vectorizable_simd_clone_call, vectorizable_assignment)
2161 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
2162 (vectorizable_comparison): Update calls to
2163 vect_get_vec_def_for_stmt_copy.
2164 (vectorizable_store): Likewise. Remove now-unnecessary calls to
2166 (vect_get_loop_based_defs): Remove dt argument and update call
2167 to vect_get_vec_def_for_stmt_copy.
2168 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
2169 and vect_get_vec_def_for_stmt_copy.
2170 (vectorizable_load): Update calls to vect_build_gather_load_calls
2171 and vect_get_vec_def_for_stmt_copy.
2172 * tree-vect-loop.c (vect_create_epilog_for_reduction)
2173 (vectorizable_reduction, vectorizable_live_operation): Update calls
2174 to vect_get_vec_def_for_stmt_copy.
2176 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2178 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
2179 and gimple stmt arguments with a stmt_vec_info.
2180 (vect_record_base_alignments): Update calls accordingly.
2181 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
2182 and gimple stmt arguments with a stmt_vec_info.
2183 (vect_build_slp_tree_1): Remove vinfo argument and update call
2184 to vect_record_max_nunits.
2185 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
2186 and vect_record_max_nunits.
2188 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2190 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
2191 file and take a stmt_vec_info instead of a gimple stmt.
2192 (supportable_widening_operation, vect_finish_replace_stmt)
2193 (vect_finish_stmt_generation, vect_get_store_rhs)
2194 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
2195 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
2196 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
2197 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
2198 (vect_create_data_ref_ptr, bump_vector_ptr)
2199 (vect_permute_store_chain, vect_setup_realignment)
2200 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
2201 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
2202 (vectorizable_reduction, vectorizable_induction)
2203 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
2204 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
2205 than gimple stmts as arguments.
2206 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
2207 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
2208 (can_group_stmts_p, vect_check_gather_scatter)
2209 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
2210 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
2211 (vect_permute_load_chain, vect_shift_permute_load_chain)
2212 (vect_transform_grouped_load)
2213 (vect_record_grouped_load_vectors): Likewise.
2214 * tree-vect-loop.c (vect_fixup_reduc_chain)
2215 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
2216 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
2217 (vectorizable_reduction, vectorizable_induction)
2218 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
2219 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
2220 (vect_get_load_store_mask): Likewise.
2221 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
2222 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
2223 * tree-vect-stmts.c (vect_mark_relevant)
2224 (is_simple_and_all_uses_invariant)
2225 (exist_non_indexing_operands_for_use_p, process_use)
2226 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
2227 (vect_get_vec_def_for_operand, vect_get_vec_defs)
2228 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
2229 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
2230 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
2231 (get_negative_load_store_type, get_load_store_type)
2232 (vect_check_load_store_mask, vect_check_store_rhs)
2233 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
2234 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2235 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2236 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2237 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
2238 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
2239 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
2240 (supportable_widening_operation): Likewise.
2242 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2244 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
2245 a stmt_vec_info instead of a gcall.
2246 (vect_check_gather_scatter): Update call accordingly.
2247 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
2249 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
2250 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
2251 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
2252 instead of a gimple stmt.
2253 (vect_transform_loop): Update calls accordingly.
2254 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
2255 stmt_vec_infos instead of gimple stmts.
2256 (vect_analyze_slp_instance): Update use accordingly.
2257 * tree-vect-stmts.c (read_vector_array, write_vector_array)
2258 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
2259 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
2260 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
2261 (vect_gen_widened_results_half, vect_get_loop_based_defs)
2262 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
2263 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
2264 down to subroutines.
2266 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2268 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
2269 of the worklist from a vector of gimple stmts to a vector of
2271 * tree-vect-stmts.c (vect_mark_relevant, process_use)
2272 (vect_mark_stmts_to_be_vectorized): Likewise
2274 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2276 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
2277 statement before passing it to vect_analyze_stmt.
2278 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
2279 the chain of phi vector definitions. Track the exit phi via its
2281 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
2282 STMT_VINFO_REDUC_DEF.
2283 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
2284 stmt_vec_infos to handle the statement chains.
2285 (vect_get_slp_defs): Record the first statement in the node
2286 using a stmt_vec_info.
2287 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
2288 statements here and pass their stmt_vec_info down to subroutines.
2289 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
2290 down to vect_finish_stmt_generation.
2291 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
2292 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
2293 stmt_vec_infos to subroutines.
2294 (vect_remove_stores): Use stmt_vec_infos to handle the statement
2297 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2299 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
2300 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
2301 (vect_permute_store_chain, vect_setup_realignment)
2302 (vect_permute_load_chain, vect_shift_permute_load_chain)
2303 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
2304 stmts internally, and when passing values to other vectorizer routines.
2305 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2306 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
2307 (vect_analyze_loop_operations, get_initial_def_for_reduction)
2308 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
2309 (vectorizable_reduction, vectorizable_induction)
2310 (vectorizable_live_operation, vect_transform_loop_stmt)
2311 (vect_transform_loop): Likewise.
2312 * tree-vect-patterns.c (vect_reassociating_reduction_p)
2313 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
2314 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
2315 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2316 (vect_slp_analyze_node_operations_1): Likewise.
2317 * tree-vect-stmts.c (vect_mark_relevant, process_use)
2318 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
2319 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
2320 (vect_finish_stmt_generation_1, get_group_load_store_type)
2321 (get_load_store_type, vect_build_gather_load_calls)
2322 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2323 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2324 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2325 (vectorizable_store, vectorizable_load, vectorizable_condition)
2326 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
2327 (supportable_widening_operation): Likewise.
2328 (vect_get_vector_types_for_stmt): Likewise.
2329 * tree-vectorizer.h (vect_dr_behavior): Likewise.
2331 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2333 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2334 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
2335 (vect_permute_store_chain, vect_permute_load_chain)
2336 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
2337 repeated stmt_vec_info lookups.
2338 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
2339 (vect_update_ivs_after_vectorizer): Likewise.
2340 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
2341 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2342 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
2343 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2344 (vect_bb_slp_scalar_cost): Likewise.
2345 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
2347 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2349 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
2350 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
2351 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
2352 to get gassigns and gcalls, rather than operating on generc gimple
2354 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
2355 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
2356 (vectorizable_load, vect_analyze_stmt): Likewise.
2357 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
2359 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2361 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
2362 return stmt_vec_infos rather than gimple stmts. Do not accept
2364 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
2366 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
2367 Update use of get_later_stmt.
2368 (vect_get_constant_vectors): Update call accordingly.
2369 (vect_schedule_slp_instance): Likewise
2370 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
2371 (vect_slp_analyze_instance_dependence): Likewise.
2372 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
2374 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2376 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
2377 (stmt_info_for_cost::stmt_info): ...this new field.
2378 (add_stmt_costs): Update accordingly.
2379 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2380 (vect_get_known_peeling_cost): Likewise.
2381 (vect_estimate_min_profitable_iters): Likewise.
2382 * tree-vect-stmts.c (record_stmt_cost): Likewise.
2384 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2386 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
2387 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2388 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
2390 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
2392 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2394 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
2395 a gimple stmt to a stmt_vec_info.
2396 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
2398 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2400 * tree-vectorizer.h (vec_info::grouped_stores): Change from
2401 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2402 (_loop_vec_info::reduction_chains): Likewise.
2403 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
2405 * tree-vect-slp.c (vect_analyze_slp): Likewise.
2407 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2409 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
2410 a gimple stmt to a stmt_vec_info.
2411 (_stmt_vec_info::next_element): Likewise.
2412 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
2413 (vect_slp_analyze_and_verify_node_alignment)
2414 (vect_analyze_group_access_1, vect_analyze_group_access)
2415 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
2416 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
2417 (vect_supportable_dr_alignment): Update accordingly.
2418 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
2419 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
2420 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
2421 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2422 * tree-vect-slp.c (vect_build_slp_tree_1)
2423 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
2424 (vect_split_slp_store_group, vect_analyze_slp_instance)
2425 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
2426 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
2427 (get_group_load_store_type, get_load_store_type)
2428 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
2429 (vect_transform_stmt, vect_remove_stores): Likewise.
2431 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2433 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
2435 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2436 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
2437 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
2438 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
2439 (vector_alignment_reachable_p, vect_get_data_access_cost)
2440 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
2441 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
2442 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
2443 (vect_analyze_group_access_1, vect_analyze_group_access)
2444 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
2445 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
2446 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
2447 result of vect_dr_stmt and use the stmt_vec_info instead of
2448 the associated gimple stmt.
2449 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
2450 (vect_gen_prolog_loop_niters): Likewise.
2451 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
2453 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2455 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
2456 to a vec<stmt_vec_info>.
2457 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
2458 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
2460 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
2461 to a vec<stmt_vec_info>.
2462 (bst_traits::value_type, bst_traits::value_type): Likewise.
2463 (bst_traits::hash): Update accordingly.
2464 (vect_get_and_check_slp_defs): Change the stmts parameter from
2465 a vec<gimple *> to a vec<stmt_vec_info>.
2466 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
2467 (vect_build_slp_tree): Likewise.
2468 (vect_build_slp_tree_2): Likewise. Update uses of
2469 SLP_TREE_SCALAR_STMTS.
2470 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
2471 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
2472 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
2473 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
2474 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
2475 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
2476 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
2477 (vect_get_constant_vectors, vect_get_slp_defs)
2478 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
2479 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
2480 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
2481 instead of gimple stmts.
2482 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
2483 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
2484 (vect_slp_analyze_instance_dependence): Update uses of
2485 SLP_TREE_SCALAR_STMTS.
2486 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2487 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
2488 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
2489 (get_initial_defs_for_reduction): Likewise.
2490 (vect_create_epilog_for_reduction): Likewise.
2491 (vectorize_fold_left_reduction): Likewise.
2492 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
2493 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
2494 (can_vectorize_live_stmts): Likewise.
2496 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2498 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
2499 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
2500 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
2502 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
2504 (gather_scalar_reductions): Update after above interface changes.
2505 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
2506 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
2508 (vect_force_simple_reduction): Likewise.
2509 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
2510 LOOP_VINFO_REDUCTIONS.
2511 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
2513 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2515 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
2516 a gimple stmt to a stmt_vec_info.
2517 * tree-vect-loop.c (vect_active_double_reduction_p)
2518 (vect_force_simple_reduction, vectorizable_reduction): Update
2521 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2523 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
2524 vec<gimple *> to a vec<stmt_vec_info>.
2525 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
2526 the reduction_phis argument from a vec<gimple *> to a
2528 (vectorizable_reduction): Likewise the phis local variable that
2529 is passed to vect_create_epilog_for_reduction. Update for new type
2530 of SLP_TREE_VEC_STMTS.
2531 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
2532 (vectorizable_live_operation): Likewise.
2533 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
2534 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
2536 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2538 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
2539 a gimple stmt to a stmt_vec_info.
2540 (vectorizable_condition, vectorizable_live_operation)
2541 (vectorizable_reduction, vectorizable_induction): Pass back the
2542 vectorized statement as a stmt_vec_info.
2543 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
2544 use of STMT_VINFO_VEC_STMT.
2545 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
2546 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
2547 as stmt_vec_infos rather than gimple stmts.
2548 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
2550 (vectorizable_live_operation): Likewise.
2551 (vectorizable_reduction, vectorizable_induction): Likewise,
2552 updating use of STMT_VINFO_VEC_STMT.
2553 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
2554 of STMT_VINFO_VEC_STMT.
2555 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
2556 (vectorizable_simd_clone_call, vectorizable_conversion)
2557 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2558 (vectorizable_store, vectorizable_load, vectorizable_condition)
2559 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
2560 from a gimple stmt to a stmt_vec_info.
2561 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
2562 pointer to a stmt_vec_info to the vectorizable_* routines.
2564 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2566 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
2567 a gimple stmt to a stmt_vec_info.
2568 (is_pattern_stmt_p): Update accordingly.
2569 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
2570 (vect_record_grouped_load_vectors): Likewise.
2571 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2572 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
2573 (vect_model_reduction_cost): Likewise.
2574 (vect_create_epilog_for_reduction): Likewise.
2575 (vectorizable_reduction, vectorizable_induction): Likewise.
2576 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2577 Return the stmt_vec_info for the pattern statement.
2578 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
2579 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
2580 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2581 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
2582 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2583 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
2584 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
2585 (free_stmt_vec_info, vect_is_simple_use): Likewise.
2587 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2589 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
2590 (vect_finish_stmt_generation): Likewise.
2591 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2592 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
2593 (vect_build_gather_load_calls): Use the return value of the above
2594 functions instead of a separate call to vinfo_for_stmt. Use narrow
2595 scopes for the input gimple stmt and wider scopes for the associated
2596 stmt_vec_info. Use vec_info::lookup_def when setting these
2597 stmt_vec_infos from an SSA_NAME definition.
2598 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
2599 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
2600 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
2601 (vectorizable_store, vectorizable_load, vectorizable_condition)
2602 (vectorizable_comparison): Likewise.
2603 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
2604 (vectorizable_reduction): Likewise.
2606 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2608 * tree-vectorizer.h (vect_is_simple_use): Add an optional
2609 stmt_vec_info * parameter before the optional gimple **.
2610 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2611 (process_use, vect_get_vec_def_for_operand_1): Update callers.
2612 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
2613 * tree-vect-loop.c (vectorizable_reduction): Likewise.
2614 (vectorizable_live_operation): Likewise.
2615 * tree-vect-patterns.c (type_conversion_p): Likewise.
2616 (vect_look_through_possible_promotion): Likewise.
2617 (vect_recog_rotate_pattern): Likewise.
2618 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2620 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2622 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
2623 a typedef to a wrapper class.
2624 (NULL_STMT_VEC_INFO): New macro.
2625 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
2626 (stmt_vec_info::operator*): New function.
2627 (stmt_vec_info::operator gimple *): Likewise.
2628 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
2629 (add_stmt_costs): Likewise.
2630 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
2631 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2632 (vect_get_known_peeling_cost): Likewise.
2633 (vect_estimate_min_profitable_iters): Likewise.
2634 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2635 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
2636 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
2637 (vectorizable_store, free_stmt_vec_infos): Likewise.
2638 (new_stmt_vec_info): Change return type of xcalloc to
2641 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2643 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
2644 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
2645 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
2646 a single_imm_use-based sequence.
2647 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
2649 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2651 * tree-vectorizer.h (vec_info::lookup_def): Declare.
2652 * tree-vectorizer.c (vec_info::lookup_def): New function.
2653 * tree-vect-patterns.c (vect_get_internal_def): Use it.
2654 (vect_widened_op_tree): Likewise.
2655 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
2656 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
2657 (vectorizable_reduction): Likewise.
2658 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
2660 (vect_is_slp_reduction): Update calls accordingly. Use
2661 vec_info::lookup_def.
2662 (vect_is_simple_reduction): Likewise
2663 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
2665 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2667 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
2668 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
2669 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
2671 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
2672 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
2673 (vect_update_vf_for_slp, vect_analyze_loop_operations)
2674 (vect_is_slp_reduction, vectorizable_induction)
2675 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
2676 * tree-vect-patterns.c (vect_init_pattern_stmt):
2677 (vect_determine_min_output_precision_1, vect_determine_precisions)
2678 (vect_pattern_recog): Likewise.
2679 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
2680 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
2681 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
2682 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
2683 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
2684 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
2685 info field from a loop to a loop_vec_info.
2687 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2689 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
2690 (vec_info::add_stmt): Declare.
2691 * tree-vectorizer.c (vec_info::add_stmt): New function.
2692 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
2693 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
2694 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
2695 (vectorizable_induction): Likewise.
2696 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
2697 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
2698 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
2699 (vectorizable_load): Likewise.
2700 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
2701 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
2702 (vect_recog_gather_scatter_pattern): Likewise.
2703 (append_pattern_def_seq): Likewise. Remove a check that is
2704 performed by add_stmt itself.
2706 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2708 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
2709 which make_ssa_name was called with new_stmt before new_stmt
2712 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2714 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
2716 (vect_is_slp_reduction): ...here...
2717 (vect_is_simple_reduction): ...and here. Remove repetition of tests
2718 that are already known to be false.
2720 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2722 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
2723 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
2724 STMT_VINFO_NUM_SLP_USES when it's true.
2725 (vect_free_slp_instance): Add a final_p parameter and pass it to
2727 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
2728 (vect_analyze_slp_instance): Likewise.
2729 (vect_slp_analyze_operations): Likewise.
2730 (vect_slp_analyze_bb_1): Likewise.
2731 * tree-vectorizer.c (vec_info): Likewise.
2732 * tree-vect-loop.c (vect_transform_loop): Likewise.
2734 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2736 * tree-vect-loop.c (vectorizable_reduction): Assert that the
2737 function is not called for second and subsequent members of
2740 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
2742 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
2743 cases for nested loops from here to ...
2744 (vect_create_epilog_for_reduction): ...here. Only call
2745 vect_is_simple_use for inner-loop reductions.
2747 2018-07-31 Martin Liska <mliska@suse.cz>
2749 PR gcov-profile/85338
2750 PR gcov-profile/85350
2751 PR gcov-profile/85372
2752 * profile.c (struct location_triplet): New.
2753 (struct location_triplet_hash): Likewise.
2754 (output_location): Do not output a BB that
2755 is already recorded for a line.
2756 (branch_prob): Use streamed_locations.
2758 2018-07-31 Martin Liska <mliska@suse.cz>
2760 PR gcov-profile/85370
2761 * coverage.c (coverage_begin_function): Do not mark target
2762 clones as artificial functions.
2764 2018-07-31 Martin Liska <mliska@suse.cz>
2766 PR gcov-profile/83813
2767 PR gcov-profile/84758
2768 PR gcov-profile/85217
2769 PR gcov-profile/85332
2770 * profile.c (branch_prob): Do not record GOTO expressions
2771 for GIMPLE statements which locations are already streamed.
2773 2018-07-31 Olivier Hainque <hainque@adacore.com>
2775 * gcc.c (handle_spec_function): Accept a soft_matched_part
2776 argument, as do_spec_1. Pass it down to ...
2777 (eval_spec_function): Accept a soft_matched_part argument,
2778 and pass it down to ...
2779 (do_spec_2): Accept a soft_matched_part argument, and pass
2780 it down to do_spec_1.
2781 (do_spec_1): Pass soft_matched_part to handle_spec_function.
2782 (handle_braces): Update call to handle_spec_function.
2783 (driver::set_up_specs): Update calls to do_spec_2.
2784 (compare_debug_dump_opt_spec_function): Likewise.
2785 (compare_debug_self_opt_spec_function): Likewise.
2787 2018-07-31 Olivier Hainque <hainque@adacore.com>
2789 * common.opt (nolibc): New option.
2790 * doc/invoke.texi (Link Options): Document it.
2791 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
2792 * config/alpha/linux.h: Likewise.
2793 * config/arc/elf.h: Likewise.
2794 * config/arm/uclinux-elf.h: Likewise.
2795 * config/arm/unknown-elf.h: Likewise.
2796 * config/avr/avrlibc.h: Likewise.
2797 * config/bfin/bfin.h: Likewise.
2798 * config/bfin/linux.h: Likewise.
2799 * config/bfin/uclinux.h: Likewise.
2800 * config/darwin.h: Likewise.
2801 * config/darwin10.h: Likewise.
2802 * config/darwin12.h: Likewise.
2803 * config/gnu-user.h: Likewise.
2804 * config/lm32/uclinux-elf.h: Likewise.
2805 * config/pa/pa-hpux11.h: Likewise.
2806 * config/pa/pa64-hpux.h: Likewise.
2807 * config/sparc/sparc.h: Likewise.
2809 2018-07-31 Olivier Hainque <hainque@adacore.com>
2811 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
2812 undefined variables.
2814 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2817 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
2820 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2822 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2823 terminated string literal.
2825 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
2827 PR rtl-optimization/85160
2828 * combine.c (is_just_move): New function.
2829 (try_combine): Allow combining two instructions into two if neither of
2830 the original instructions was a move.
2832 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
2835 * doc/extend.texi (Global Register Variables): Discourage use of type
2837 (Local Register Variables): Likewise.
2839 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
2841 PR tree-optimization/86506
2842 * hwint.h (ceil_log2): Resync with hwint.c implementation.
2844 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
2847 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
2848 hard_regno, make sure no insn between `from` and `to` clobbers it.
2850 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
2851 Tom de Vries <tdevries@suse.de>
2853 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
2854 (PTX_DEFAULT_RUNTIME_DIM): ... this.
2855 (nvptx_goacc_validate_dims): Set default worker and gang dims to
2856 PTX_DEFAULT_RUNTIME_DIM.
2857 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
2859 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
2861 * config/pa/pa.c (pa_output_addr_vec): Align address table.
2862 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
2863 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
2865 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
2867 * config/rs6000/constraints.md (wG constraint): Delete, no longer
2869 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
2870 predicate to reflect toc fusion has been deleted.
2871 (toc_fusion_mem_raw): Delete, no longer used.
2872 (toc_fusion_mem_wrapped): Likewise.
2873 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
2875 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
2876 Delete, no longer used.
2877 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
2878 meant to be used for toc fusion.
2879 (rs6000_debug_print_mode): Delete toc fusion debugging.
2880 (rs6000_debug_reg_global): Likewise.
2881 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
2882 fusion and secondary reload support that were never used.
2883 (rs6000_option_override_internal): Delete TOC fusion, that was only
2884 partially defined, and it did not work unless you also used the
2886 (rs6000_legitimate_address_p): Delete TOC fusion support.
2887 (rs6000_opt_masks): Likewise.
2888 (fusion_wrap_memory_address): Delete function, no longer used.
2889 (fusion_split_address); Delete TOC fusion support.
2890 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
2891 longer used with toc fusion being deleted.
2892 (TARGET_TOC_FUSION_FP): Likewise.
2893 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
2895 (toc fusion spliter): Delete TOC fusion support.
2896 (toc_fusionload_<mode>): Likewise.
2897 (toc_fusionload_di): Likewise.
2898 (fusion_gpr_load_<mode>): Delete generator function, this insn no
2899 longer needs to be named. Rename predicate to delete TOC fusion.
2900 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2901 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2902 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
2903 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
2904 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
2906 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
2908 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
2909 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
2910 __int128 in built-in function prototypes.
2911 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
2912 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
2914 2018-07-27 Martin Sebor <msebor@redhat.com>
2916 PR tree-optimization/86696
2917 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
2918 types, including enums.
2919 (handle_char_store): Be prepared for the above function to fail.
2921 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
2923 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
2924 when optimization level is lower than 2 or optimize for size.
2926 2018-07-26 Martin Sebor <msebor@redhat.com>
2928 PR tree-optimization/86043
2929 PR tree-optimization/86042
2930 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
2931 (get_string_cst_length): Rename...
2932 (get_min_string_length): ...to this. Add argument.
2933 (handle_char_store): Extend to handle multi-character stores by
2935 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
2936 * tree.h (initializer_zerop): Add argument.
2938 2018-07-26 Jakub Jelinek <jakub@redhat.com>
2941 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
2942 declare target to variables if they have always,{to,from,tofrom} map
2945 2018-07-26 Martin Liska <mliska@suse.cz>
2948 * lto-wrapper.c: Add linker_output as prefix
2949 for ltrans_output_file.
2951 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
2953 PR rtl-optimization/85805
2954 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
2955 value for hard registers if that was written in the same mode.
2957 2018-07-26 Martin Liska <mliska@suse.cz>
2959 PR gcov-profile/86536
2960 * gcov.c (format_gcov): Use printf format %.*f directly
2961 and do not handle special values.
2963 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2965 * common/config/arc/arc-common.c (arc_option_optimization_table):
2966 Update default optimizations for size.
2968 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2970 * config/arc/arc.md (movsf_insn): Add short instruction selection.
2971 * config/arc/constraints.md (CfZ): New constraint.
2972 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
2973 (subsf3_fpu): Likewise.
2974 (cmpsf_fpu): Likewise.
2975 (cmpsf_fpu_uneq): Likewise.
2977 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2979 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
2981 (arc_is_uncached_mem_p): uncached applies to both the variable and
2984 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
2986 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
2989 2018-07-25 David Malcolm <dmalcolm@redhat.com>
2991 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
2992 field "m_scopes" from vec to auto_vec.
2994 2018-07-25 Martin Liska <mliska@suse.cz>
2996 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
2999 2018-07-25 Richard Biener <rguenther@suse.de>
3002 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
3003 special wrt context_die late.
3004 (gen_subprogram_die): Re-use DIEs in local scope.
3006 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
3008 PR tree-optimization/86644
3009 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
3011 2018-07-25 Martin Liska <mliska@suse.cz>
3014 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
3015 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
3017 2018-07-25 Martin Liska <mliska@suse.cz>
3020 * params.def: Explain ASan abbreviation and provide
3021 a documentation link.
3023 2018-07-24 Martin Sebor <msebor@redhat.com>
3025 PR tree-optimization/86622
3026 PR tree-optimization/86532
3027 * builtins.h (string_length): Declare.
3028 * builtins.c (c_strlen): Correct handling of non-constant offsets.
3029 (check_access): Be prepared for non-constant length ranges.
3030 (string_length): Make extern.
3031 * expr.c (string_constant): Only handle the minor non-constant
3032 array index. Use string_constant to compute the length of
3033 a generic string constant.
3035 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
3037 PR tree-optimization/86618
3038 * tree-vect-stmts.c (vectorizable_call): Don't take the address
3039 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
3041 2018-07-24 David Malcolm <dmalcolm@redhat.com>
3043 PR tree-optimization/86636
3044 * json.cc (json::object::set): Fix comment. Add assertions.
3045 (json::array::append): Move here from json.h. Add comment and an
3047 (json::string::string): Likewise.
3048 * json.h (json::array::append): Move to json.cc.
3049 (json::string::string): Likewise.
3050 * optinfo-emit-json.cc
3051 (optrecord_json_writer::impl_location_to_json): Assert that we
3052 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
3053 wrapper around it. Expand the location once, rather than three
3055 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
3056 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
3058 (optrecord_json_writer::optinfo_to_json): Likewise, in four
3059 places. Fix some overlong lines.
3061 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
3063 * config/aarch64/aarch64-simd.md
3064 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
3065 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
3066 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
3067 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
3068 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
3069 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
3070 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
3071 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
3072 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
3074 2018-07-24 Jakub Jelinek <jakub@redhat.com>
3077 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
3078 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
3079 and abs_d == d, do the power of two handling if profitable.
3081 2018-07-24 Richard Biener <rguenther@suse.de>
3083 * match.pd: Add BIT_FIELD_REF canonicalizations.
3085 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3088 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
3090 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3092 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
3093 terminated STRING_CST object.
3095 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
3097 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
3099 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
3101 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
3102 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
3103 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
3104 the elements into a register.
3105 (rs6000_split_v4si_init_di_reg): Delete.
3106 (rs6000_split_v4si_init): Delete.
3107 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
3108 (vsx_init_v4si): Rewrite as a define_expand.
3110 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
3112 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
3113 zero_extend argument from memory): New.
3115 2018-07-22 Martin Sebor <msebor@redhat.com>
3118 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
3119 diagnosing calls with unknown arguments unless -Walloca-larger-than
3120 is restricted to less than PTRDIFF_MAX bytes.
3122 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
3124 * doc/gcov.texi (Invoking Gcov): Editorial changes.
3126 2018-07-20 David Malcolm <dmalcolm@redhat.com>
3128 * pretty-print.c (text_info::set_location): Remove redundant
3129 "line_table" parameter from call to rich_location::set_range.
3131 2018-07-20 Martin Sebor <msebor@redhat.com>
3134 * builtins.c (expand_builtin_alloca): Adjust.
3135 * calls.c (alloc_max_size): Simplify.
3136 * cgraphunit.c (cgraph_node::expand): Adjust.
3137 * common.opt (larger_than_size, warn_frame_larger_than): Remove
3139 (frame_larger_than_size): Same.
3140 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
3141 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
3143 * doc/invoke.texi (GCC Command Options): Document option arguments.
3144 Explain byte-size arguments and suffixes.
3145 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
3146 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
3147 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
3148 * doc/options.texi (UInteger): Expand.
3149 (Host_Wide_Int, ByteSize): Document new properties.
3150 * final.c (final_start_function_1): Include sizes in an error message.
3151 * function.c (frame_offset_overflow): Same.
3152 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
3153 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
3154 Diagnose unbounded alloca calls only for limits of less than
3156 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
3157 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
3159 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
3160 only for limits of less than PTRDIFF_MAX.
3161 * langhooks-def.h (lhd_handle_option): Change function argument
3163 * langhooks.c (lhd_handle_option): Same.
3164 * langhooks.h (handle_option): Same.
3165 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
3167 (var_type, var_type_struct): Same.
3168 (var_set): Handle ByteSize flag.
3169 * optc-gen.awk: Add comments to output to ease debugging. Make
3170 use of HOST_WIDE_INT where appropriate.
3171 * opts-gen-save.awk: Use %lx to format unsigned long.
3172 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
3173 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
3174 arguments. Parse bytes-size suffixes.
3175 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
3176 (enum_value_to_arg): Same.
3177 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
3178 (handle_option): Adjust.
3179 (generate_option): Change function argument to HOST_WIDE_INT.
3180 (cmdline_handle_error): Adjust.
3181 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
3182 (set_option): Change function argument to HOST_WIDE_INT.
3183 (option_enabled): Handle cl_host_wide_int.
3184 (get_option_state): Handle CLVC_SIZE.
3185 (control_warning_option): Same.
3186 * opts.c (common_handle_option): Change function argument to
3187 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
3188 OPT_Wvla_larger_than_.
3189 * opts.h (enum cl_var_type): Add an enumerator.
3190 * stor-layout.c (layout_decl): Print a more meaningful warning.
3191 * toplev.c (output_stack_usage): Adjust.
3193 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
3195 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
3196 call to inline_expand_builtin_string_cmp.
3197 (expand_builtin_strcmp): Likewise.
3198 (expand_builtin_strncmp): Likewise.
3199 (inline_string_cmp): Delete the last parameter, change char_type_node
3200 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
3202 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
3203 the inlining expansion on target where the type of the call has same or
3204 narrower precision than unsigned char.
3206 2018-07-20 David Malcolm <dmalcolm@redhat.com>
3208 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
3209 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
3210 * common.opt (fsave-optimization-record): New option.
3211 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
3212 * doc/invoke.texi (-fsave-optimization-record): New option.
3213 * dumpfile.c: Include "optinfo-emit-json.h".
3214 (struct kv_pair): Move to coretypes.h.
3215 (optgroup_options): Make non-static.
3216 (dump_context::end_scope): Call
3217 optimization_records_maybe_pop_dump_scope.
3218 * dumpfile.h (optgroup_options): New decl.
3219 * json.cc: New file.
3221 * optinfo-emit-json.cc: New file.
3222 * optinfo-emit-json.h: New file.
3223 * optinfo.cc: Include "optinfo-emit-json.h".
3224 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
3225 (optinfo_enabled_p): Check optimization_records_enabled_p.
3226 (optinfo_wants_inlining_info_p): Likewise.
3227 * optinfo.h: Update comment.
3228 * profile-count.c (profile_quality_as_string): New function.
3229 * profile-count.h (profile_quality_as_string): New decl.
3230 (profile_count::quality): New accessor.
3231 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
3232 and optinfo_emit_json_cc_tests.
3233 * selftest.h (selftest::json_cc_tests): New decl.
3234 (selftest::optinfo_emit_json_cc_tests): New decl.
3235 * toplev.c: Include "optinfo-emit-json.h".
3236 (compile_file): Call optimization_records_finish.
3237 (do_compile): Call optimization_records_start.
3238 * tree-ssa-live.c: Include optinfo.h.
3239 (remove_unused_scope_block_p): Retain inlining information if
3240 optinfo_wants_inlining_info_p returns true.
3242 2018-07-20 Richard Biener <rguenther@suse.de>
3245 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
3246 to cover -flto-partition=none.
3248 2018-07-20 Martin Liska <mliska@suse.cz>
3250 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
3251 (get_decl_source_range): Remove unused function.
3253 2018-07-20 Richard Biener <rguenther@suse.de>
3255 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
3256 (struct vn_phi_s): Likewise.
3257 (struct vn_reference_s): Likewise.
3258 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
3259 for searching the slot of an entry known to be in the hash itself.
3260 (vn_phi_hasher::equal): Likewise.
3261 (vn_reference_hasher::equal): Likewise.
3262 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
3264 (optimistic_info, current_info): Remove, keeping only valid_info.
3265 (vn_reference_lookup_1): Remove fallback lookup.
3266 (vn_reference_lookup_2): Likewise.
3267 (vn_nary_op_lookup_1): Likewise.
3268 (vn_phi_lookup): Likewise.
3269 (vn_nary_build_or_lookup_1): Make sure to not chain the built
3271 (vn_reference_insert): Adjust, chain the inserted hash element
3272 at last_inserted_ref.
3273 (vn_reference_insert_pieces): Likewise.
3274 (visit_reference_op_call): Likewise.
3275 (vn_nary_op_insert_into): Chain the inserted hash element at
3277 (vn_nary_op_insert_pieces): Adjust.
3278 (vn_nary_op_insert): Likewise.
3279 (vn_nary_op_insert_stmt): Likewise.
3280 (vn_phi_insert): Adjust, chain the inserted hash element at
3282 (process_scc): Remove clearing and copying the optimistic
3283 table. Instead remove elements inserted during an optimistic
3284 iteration from the single table we maintain.
3285 (init_scc_vn): Adjust.
3286 (free_scc_vn): Likewise.
3287 (sccvn_dom_walker::record_cond): Likewise.
3288 (sccvn_dom_walker::after_dom_children): Likewise.
3290 2018-07-19 Martin Sebor <msebor@redhat.com>
3292 PR tree-optimization/84047
3293 PR tree-optimization/83776
3294 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
3295 (check_array_bounds): Call it.
3297 2018-07-19 Martin Sebor <msebor@redhat.com>
3299 * align.h (align_flags): Use member initialization.
3301 2018-07-19 David Malcolm <dmalcolm@redhat.com>
3303 * Makefile.in (OBJS): Add optinfo.o.
3304 * coretypes.h (class symtab_node): New forward decl.
3305 (struct cgraph_node): New forward decl.
3306 (class varpool_node): New forward decl.
3307 * dump-context.h: New file.
3308 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
3310 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
3311 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
3312 (set_alt_dump_file): Likewise.
3313 (dump_context::~dump_context): New dtor.
3314 (dump_gimple_stmt): Move implementation to...
3315 (dump_context::dump_gimple_stmt): ...this new member function.
3316 Add the stmt to any pending optinfo, creating one if need be.
3317 (dump_gimple_stmt_loc): Move implementation to...
3318 (dump_context::dump_gimple_stmt_loc): ...this new member function.
3319 Start a new optinfo and add the stmt to it.
3320 (dump_gimple_expr): Move implementation to...
3321 (dump_context::dump_gimple_expr): ...this new member function.
3322 Add the stmt to any pending optinfo, creating one if need be.
3323 (dump_gimple_expr_loc): Move implementation to...
3324 (dump_context::dump_gimple_expr_loc): ...this new member function.
3325 Start a new optinfo and add the stmt to it.
3326 (dump_generic_expr): Move implementation to...
3327 (dump_context::dump_generic_expr): ...this new member function.
3328 Add the tree to any pending optinfo, creating one if need be.
3329 (dump_generic_expr_loc): Move implementation to...
3330 (dump_context::dump_generic_expr_loc): ...this new member
3331 function. Add the tree to any pending optinfo, creating one if
3333 (dump_printf): Move implementation to...
3334 (dump_context::dump_printf_va): ...this new member function. Add
3335 the text to any pending optinfo, creating one if need be.
3336 (dump_printf_loc): Move implementation to...
3337 (dump_context::dump_printf_loc_va): ...this new member function.
3338 Start a new optinfo and add the stmt to it.
3339 (dump_dec): Move implementation to...
3340 (dump_context::dump_dec): ...this new member function. Add the
3341 value to any pending optinfo, creating one if need be.
3342 (dump_context::dump_symtab_node): New member function.
3343 (dump_context::get_scope_depth): New member function.
3344 (dump_context::begin_scope): New member function.
3345 (dump_context::end_scope): New member function.
3346 (dump_context::ensure_pending_optinfo): New member function.
3347 (dump_context::begin_next_optinfo): New member function.
3348 (dump_context::end_any_optinfo): New member function.
3349 (dump_context::s_current): New global.
3350 (dump_context::s_default): New global.
3351 (dump_scope_depth): Delete global.
3352 (dumpfile_ensure_any_optinfo_are_flushed): New function.
3353 (dump_symtab_node): New function.
3354 (get_dump_scope_depth): Reimplement in terms of dump_context.
3355 (dump_begin_scope): Likewise.
3356 (dump_end_scope): Likewise.
3357 (selftest::temp_dump_context::temp_dump_context): New ctor.
3358 (selftest::temp_dump_context::~temp_dump_context): New dtor.
3359 (selftest::verify_item): New function.
3360 (ASSERT_IS_TEXT): New macro.
3361 (ASSERT_IS_TREE): New macro.
3362 (ASSERT_IS_GIMPLE): New macro.
3363 (selftest::test_capture_of_dump_calls): New test.
3364 (selftest::dumpfile_c_tests): Call it.
3365 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
3366 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
3367 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
3368 descriptive comment.
3369 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
3370 (dump_node, dump_bb): Move these unrelated decls.
3371 (class dump_manager): Add leading comment.
3372 * optinfo.cc: New file.
3373 * optinfo.h: New file.
3375 2018-07-19 Michael Collison <michael.collison@arm.com>
3376 Richard Henderson <rth@redhat.com>
3378 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
3379 (subti): Handle op1 zero.
3380 (subvti4, usub4ti4): New.
3381 (*sub<GPI>3_compare1_imm): New.
3382 (sub<GPI>3_carryinCV): New.
3383 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
3384 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
3386 2018-07-19 Michael Collison <michael.collison@arm.com>
3387 Richard Henderson <rth@redhat.com>
3389 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
3390 (addti3): Create simpler code if low part is already known to be 0.
3391 (addvti4, uaddvti4): New.
3392 (*add<GPI>3_compareC_cconly_imm): New.
3393 (*add<GPI>3_compareC_cconly): New.
3394 (*add<GPI>3_compareC_imm): New.
3395 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
3396 handle constants within this pattern..
3397 (*add<GPI>3_compareV_cconly_imm): New.
3398 (*add<GPI>3_compareV_cconly): New.
3399 (*add<GPI>3_compareV_imm): New.
3400 (add<GPI>3_compareV): New.
3401 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
3402 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
3403 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
3404 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
3405 with 'comparison' operator.
3406 (*add<GPI>3_compareV_cconly_imm): Ditto.
3407 (*add<GPI>3_compareV_cconly): Ditto.
3408 (*add<GPI>3_compareV_imm): Ditto.
3409 (add<GPI>3_compareV): Ditto.
3410 (add<mode>3_carryinC): Ditto.
3411 (*add<mode>3_carryinC_zero): Ditto.
3412 (*add<mode>3_carryinC): Ditto.
3413 (add<mode>3_carryinV): Ditto.
3414 (*add<mode>3_carryinV_zero): Ditto.
3415 (*add<mode>3_carryinV): Ditto.
3417 2018-07-19 Michael Collison <michael.collison@arm.com>
3418 Richard Henderson <rth@redhat.com>
3420 * config/aarch64/aarch64-modes.def (CC_V): New.
3421 * config/aarch64/aarch64-protos.h
3422 (aarch64_addti_scratch_regs): Declare
3423 (aarch64_subvti_scratch_regs): Declare.
3424 (aarch64_expand_subvti): Declare.
3425 (aarch64_gen_unlikely_cbranch): Declare
3426 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
3427 for signed overflow using CC_Vmode.
3428 (aarch64_get_condition_code_1): Handle CC_Vmode.
3429 (aarch64_gen_unlikely_cbranch): New function.
3430 (aarch64_addti_scratch_regs): New function.
3431 (aarch64_subvti_scratch_regs): New function.
3432 (aarch64_expand_subvti): New function.
3434 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3436 * config/aarch64/aarch64-option-extensions.def: New entry for profile
3438 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
3439 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
3442 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3445 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
3446 address check not strict.
3448 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
3450 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
3452 (store_pair_lanes<mode>): Likewise.
3453 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
3454 enum value 'ADDR_QUERY_LDP_STP_N'.
3455 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
3456 (aarch64_print_address_internal): Add declaration.
3457 (aarch64_print_ldpstp_address): Remove.
3458 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
3459 (aarch64_print_operand): Change printing of 'y'.
3460 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
3461 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
3462 'true' rather than '1'.
3463 * config/aarch64/constraints.md (Uml): Likewise.
3464 (Uml): Rename to Umn.
3467 2018-07-19 Richard Biener <rguenther@suse.de>
3469 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
3471 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
3472 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
3473 (vn_reference_hasher): Likewise.
3474 (struct vn_tables_s): Remove obstack and alloc-pool members.
3475 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
3476 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
3477 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
3478 (vn_reference_insert_pieces): Likewise.
3479 (alloc_vn_nary_op_noinit): Adjust.
3480 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
3481 (vn_phi_eq): Adjust.
3482 (shared_lookup_phiargs): Remove.
3483 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
3484 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
3485 (visit_reference_op_call): Likewise.
3486 (copy_nary, copy_phi, copy_reference): Remove.
3487 (process_scc): Rewind the obstack when iterating. Do not
3488 copy the elements to valid_info but just move them from one
3489 hashtable to the other.
3490 (allocate_vn_table): Adjust.
3491 (free_vn_table): Likewise.
3492 (init_scc_vn): Likewise.
3493 (free_scc_vn): Likewise.
3495 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
3498 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
3499 indirect_return as function type attribute.
3500 (ix86_attribute_table): Change indirect_return to function
3502 * doc/extend.texi: Update indirect_return attribute.
3504 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
3506 * wide-int.h (widest2_int): New.
3507 * gimple-fold.c (arith_overflowed_p): Use it.
3508 * tree.h (widest2_int_cst): New.
3509 * tree-vrp.c (wide_int_binop_overflow): Rename from
3510 vrp_int_const_binop.
3511 Rewrite to work on trees.
3512 (extract_range_from_multiplicative_op_1): Abstract code to...
3513 (wide_int_range_min_max): ...here.
3514 (wide_int_range_cross_product): ...and here.
3515 (extract_range_from_binary_expr_1): Abstract overflow code to...
3516 (wide_int_range_mult_wrapping): ...here.
3517 * tree-vrp.h (wide_int_range_cross_product): New.
3518 (wide_int_range_mult_wrapping): New.
3520 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
3521 Julia Koval <julia.koval@intel.com>
3523 * config/i386/x86-tune-costs.h (skylake_memcpy,
3524 skylake_memset): Replace rep_prefix with unrolling for size 512.
3526 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
3529 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
3530 comparision with EQ_EXPR in last stmt.
3532 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
3534 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
3535 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
3536 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
3537 previously known as "PowerPC AltiVec Built-in Functions". Move
3538 some material to new subsubsections "PowerPC AltiVec Built-in
3539 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
3541 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
3542 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
3543 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
3544 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
3546 2018-07-18 Richard Biener <rguenther@suse.de>
3548 PR tree-optimization/86557
3549 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
3552 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
3554 * config/s390/s390.c (s390_function_profiler): Generate CFI.
3556 2018-07-17 Jeff Law <law@redhat.com>
3558 * config/arm/arm.c (get_label_padding): Update for recent
3559 changes to label_to_alignment.
3561 PR tree-optimization/86010
3562 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
3564 * config/mips/mips.c (vr4130_align_insns): Update for recent
3565 changes to label_to_alignment.
3567 * config/frv/frv.c (frv_label_align): Update for recent changes
3568 to label_to_alignment.
3570 * config/nios2/nios2.c (nios2_label_align): Update for recent
3571 changes which dropped ALIGN_LABELS_LOG.
3573 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
3575 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
3576 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
3578 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
3580 * config/arc/arc.c (arc_label_align): Use align_labels instead of
3581 deprecated align_labels_log.
3583 2018-07-17 Richard Biener <rguenther@suse.de>
3586 * dwarf2out.c (init_sections_and_labels): Always generate
3587 a debug_line_str_section for early LTO debug.
3588 (dwarf2out_finish): Reset debug_line_str_hash output early.
3589 Bump counter for extra dwarf5 .debug_loc labels to not conflict
3591 (dwarf2out_early_finish): Output debug_line_str.
3593 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3595 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
3596 index register on z196 or later.
3598 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
3600 * config/s390/s390.c (s390_default_align): Set default function
3602 (s390_override_options_after_change): Call s390_default align.
3603 (s390_option_override_internal): Call s390_default_align.
3604 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
3606 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3609 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
3610 to _looptemp_ clauses, other than the first two.
3612 2018-07-17 Martin Liska <mliska@suse.cz>
3614 * opts.c: Do not enable OPT_falign_* for -Os.
3616 2018-07-17 Martin Liska <mliska@suse.cz>
3618 * align.h (MAX_CODE_ALIGN): New.
3619 (MAX_CODE_ALIGN_VALUE): New.
3620 * common/config/i386/i386-common.c (ix86_handle_option):
3621 (MAX_CODE_ALIGN): Moved to align.h.
3622 * final.c (MAX_CODE_ALIGN): Likewise.
3623 * opts.c (parse_and_check_align_values):
3624 (MAX_CODE_ALIGN): Likewise.
3625 (MAX_CODE_ALIGN_VALUE): Likewise.
3627 2018-07-17 Martin Liska <mliska@suse.cz>
3629 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
3630 in order to fulfil coding style.
3631 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
3632 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3633 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3634 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
3635 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
3636 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
3637 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
3638 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
3639 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3641 2018-07-17 Martin Liska <mliska@suse.cz>
3643 * align.h: New file.
3644 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
3646 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
3647 align_flags of label_to_alignment.
3648 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
3650 * config/m68k/m68k.c: Do not use removed align_labels_value and
3652 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
3653 (LOOP_ALIGN): Likewise.
3654 (LABEL_ALIGN): Likewise.
3655 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3656 Remove not used macro.
3657 (rs6000_loop_align): Change return type to align_flags.
3658 (rs6000_loop_align_max_skip): Remove.
3659 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
3660 Change return type to align_flags.
3661 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
3662 Remove not used macro.
3663 (rs6000_loop_align): Change return type to align_flags.
3664 (rs6000_loop_align_max_skip): Remove.
3665 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
3666 * config/rx/rx-protos.h (rx_align_for_label): Make it
3668 * config/rx/rx.c (rx_align_for_label): Change return type
3670 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
3672 into align_flags class.
3673 (LABEL_ALIGN): Likewise.
3674 (LOOP_ALIGN): Likewise.
3675 * config/s390/s390.c (s390_label_align): Use align_flags
3677 (s390_asm_output_function_label): Likewise.
3678 * config/sh/sh.c (sh_override_options_after_change):
3679 Use align_flags class directly without macros.
3680 (find_barrier): Likewise.
3681 (barrier_align): Likewise.
3682 (sh_loop_align): Likewise.
3683 * config/spu/spu.c (spu_option_override):
3684 Use align_flags_tuple::get_value instead of removed macros.
3685 (spu_sched_init): Likewise.
3686 * config/spu/spu.h (GTY): Likewise.
3687 * config/visium/visium.c (visium_option_override):
3688 Set "8" as default secondary alignment.
3689 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
3690 in order to guarantee secondary alignment of 8.
3691 * coretypes.h: Include align.h header file.
3692 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
3693 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
3694 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
3695 * doc/tm.texi.in: Likewise.
3696 * final.c (struct label_alignment): Remove not used structure.
3697 (LABEL_ALIGN): Change type to align_flags.
3698 (LOOP_ALIGN): Likewise.
3699 (JUMP_ALIGN): Likewise.
3700 (default_loop_align_max_skip): Remove.
3701 (default_label_align_max_skip): Likewise.
3702 (default_jump_align_max_skip): Likewise.
3703 (default_label_align_after_barrier_max_skip):
3704 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
3705 (LABEL_TO_MAX_SKIP): Remove.
3706 (label_to_alignment): Return align_flags type instead of integer.
3707 (label_to_max_skip): Remove.
3708 (align_fuzz): Use align_flags type.
3709 (compute_alignments): Use align_flags type and use align_flags::max
3710 to combine multiple alignments.
3711 (grow_label_align): Grow vec instead of C array.
3712 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
3713 (shorten_branches): Use align_flags type and use align_flags::max
3714 to combine multiple alignments.
3715 (final_scan_insn_1): Remove usage of secondary alignment that comes
3716 from label alignment, but instead use proper secondary alignment
3717 which is computed in grow_label_align.
3718 * flags.h (struct align_flags_tuple): Move to align.h.
3719 (struct align_flags): Likewise.
3720 (state_align_loops): Rename to align_loops.
3721 (state_align_jumps): Rename to align_jumps.
3722 (state_align_labels): Rename to align_labels.
3723 (state_align_functions): Rename to align_functions.
3724 (align_loops_log): Remove.
3725 (align_jumps_log): Remove.
3726 (align_labels_log): Remove.
3727 (align_functions_log): Remove.
3728 (align_loops_max_skip): Remove.
3729 (align_jumps_max_skip): Remove.
3730 (align_labels_max_skip): Remove.
3731 (align_functions_max_skip): Remove.
3732 (align_loops_value): Remove.
3733 (align_jumps_value): Remove.
3734 (align_labels_value): Remove.
3735 (align_functions_value): Remove.
3736 * output.h (label_to_alignment): Change return type to align_flags.
3737 (label_to_max_skip): Remove.
3738 * target.def: Remove loop_align_max_skip, label_align_max_skip,
3739 jump_align_max_skip macros.
3740 * targhooks.h (default_loop_align_max_skip): Remove.
3741 (default_label_align_max_skip): Likewise.
3742 (default_jump_align_max_skip): Likewise.
3743 (default_label_align_after_barrier_max_skip): Remove.
3744 * toplev.c (read_log_maxskip): Use ::normalize function.
3745 (parse_N_M): Remove not used argument and also call ::normalize.
3746 (parse_alignment_opts): Do not pass unused arguments.
3747 * varasm.c (assemble_start_function): Use directly align_functions
3748 instead of removed macros.
3749 * system.h: Do not poison removed macros.
3751 2018-07-17 Jakub Jelinek <jakub@redhat.com>
3754 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
3755 and cond temporaries don't have reference type if iterator has
3756 pointer type. For init use &for_pre_body instead of pre_p if
3757 for_pre_body is non-empty.
3759 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3761 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
3762 double-double modes to SFmode directly directly.
3763 (trunc<mode>sf2_fprs): Delete.
3765 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3767 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
3768 for conversions between IFmode and the decimal floating point modes.
3769 (init_float128_ieee): Use the correct names for conversions between
3770 KFmode and the decimal floating point modes.
3772 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
3774 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
3775 for the conversions between TDmode and IFmode.
3776 (init_float128_ieee): Use more correct names for the conversions
3777 between TDmode and KFmode.
3779 2018-07-16 Jakub Jelinek <jakub@redhat.com>
3781 PR tree-optimization/86526
3782 * builtins.c (expand_builtin_memcmp): Formatting fixes.
3783 (inline_expand_builtin_string_cmp): Likewise.
3784 (inline_string_cmp): Likewise. Use c_readstr instead of
3785 builtin_memcpy_read_str. Add unit_mode temporary.
3787 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
3790 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
3791 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
3793 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3795 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
3796 Alphabetize prototypes of built-in functions, separating out
3797 built-in functions that are listed in this section but should be
3798 described elsewhere.
3800 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
3803 * expmed.c (emit_store_flag): Do not emit setcc followed by a
3804 conditional move when trapping comparison was split to a
3805 non-trapping one (and vice versa).
3807 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3809 * config/s390/s390.c (s390_function_profiler): Generate nops
3810 instead of profiler call sequences.
3811 * config/s390/s390.opt: Add the new option.
3813 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3815 * config/s390/s390.c (s390_function_profiler): Generate
3816 __mcount_loc section.
3817 * config/s390/s390.opt: Add the new option.
3819 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
3821 * common.opt: Add the new warning.
3822 * config/s390/s390.c (s390_function_profiler): Emit "brasl
3823 %r0,__fentry__" when -mfentry is specified.
3824 (s390_option_override_internal): Disallow -mfentry for 31-bit
3826 * config/s390/s390.opt: Add the new option.
3828 2018-07-16 Richard Biener <rguenther@suse.de>
3831 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
3832 for function-local FUNCTION_DECL and RESULT_DECL immediately.
3834 2018-07-16 Martin Liska <mliska@suse.cz>
3837 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
3840 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3842 * config/arc/arcHS.md: Update ARCHS scheduling rules.
3844 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
3846 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
3848 * config/arc/arc-cpus.def (hs4x): New cpu.
3850 * config/arc/arc-tables.opt: Regenerate.
3851 * config/arc/arc.c (arc_sched_issue_rate): New function.
3852 (TARGET_SCHED_ISSUE_RATE): Define.
3853 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
3854 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
3856 (attr tune): Add ARCHS4x tune values.
3857 (attr tune_dspmpy): Define.
3858 (*tst): Correct instruction type.
3859 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
3860 * config/arc/arcHS4x.md: New file.
3861 * config/arc/fpu.md: Update instruction type attributes.
3862 * config/arc/t-multilib: Regenerate.
3864 2018-07-16 Tom de Vries <tdevries@suse.de>
3867 * var-tracking.c (vt_initialize): Fix pre_dec handling.
3869 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3871 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
3874 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
3876 PR tree-optimization/86514
3877 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
3878 conversion to a boolean type from a type with greater precision.
3880 2018-07-16 Tom de Vries <tdevries@suse.de>
3882 * var-tracking.c (vt_initialize): Print adjusted insn slim if
3883 dump_flags request TDF_SLIM.
3885 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
3887 * fold-const.c (int_const_binop_1): Abstract...
3888 (wide_int_binop): ...wide int code here.
3889 (poly_int_binop): ...poly int code here.
3890 Abstract the rest of int_const_binop_1 into int_const_binop.
3891 * fold-const.h (wide_int_binop): New.
3892 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
3893 Remove useless PLUS/MINUS_EXPR case.
3894 (zero_nonzero_bits_from_vr): Move wide int code...
3895 (zero_nonzero_bits_from_bounds): ...here.
3896 (extract_range_from_binary_expr_1): Move mask optimization code...
3897 (range_easy_mask_min_max): ...here.
3898 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
3899 (range_easy_mask_min_max): New.
3901 2018-07-15 Jeff Law <law@redhat.com>
3904 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
3907 2018-07-14 Jim Wilson <jimw@sifive.com>
3909 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
3911 2018-07-14 Paul Koning <ni1d@arrl.net>
3913 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
3915 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
3917 * lto-streamer-out.c (copy_function_or_variable): Dump info about
3920 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
3921 Steve Munroe <munroesj52@gmail.com>
3923 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
3924 (_mm_andnot_si128): Likewise.
3925 (_mm_or_si128): Likewise.
3926 (_mm_xor_si128): Likewise.
3928 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
3931 * builtins.c (expand_builtin_memcmp): Inline the calls first
3932 when result_eq is false.
3933 (expand_builtin_strcmp): Inline the calls first.
3934 (expand_builtin_strncmp): Likewise.
3935 (inline_string_cmp): New routine. Expand a string compare
3936 call by using a sequence of char comparison.
3937 (inline_expand_builtin_string_cmp): New routine. Inline expansion
3938 a call to str(n)cmp/memcmp.
3939 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
3941 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
3943 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
3945 * config/arm/driver-arm.c: Include arm-native.h.
3946 (host_detect_local_cpu): Use auto-generated data tables.
3947 (vendors, arm_cpu_table): Delete. Move part information to ...
3948 * config/arm/arm-cpus.in: ... here.
3949 * config/arm/parsecpu.awk (gen_native): New function.
3950 (vendor, part): New CPU fields.
3951 (END): Add support for building the native CPU detection tables.
3952 * config/arm/t-arm (arm-native.h): Add build rule.
3953 (driver-arm.o): Add dependency on arm-native.h.
3955 2018-07-13 Richard Biener <rguenther@suse.de>
3958 * match.pd (addr1 - addr2): Allow either of the operand to
3961 2018-07-13 Tom de Vries <tdevries@suse.de>
3963 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
3966 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
3968 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
3969 arrays instead of numbered variables.
3971 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
3973 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
3974 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
3976 2018-07-13 Richard Biener <rguenther@suse.de>
3979 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
3980 instead of get_context_die.
3982 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
3983 Richard Biener <rguenther@suse.de>
3986 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
3987 that the loop latch destination where phi is defined.
3989 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
3991 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
3992 (riscv_expand_epilogue): Add assertion to check interrupt mode.
3993 (riscv_set_current_function): Extract getting interrupt type to new
3995 (riscv_get_interrupt_type): New function.
3996 (riscv_merge_decl_attributes): New function, checking interrupt type is
3998 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
4000 2018-07-12 Paul Koning <ni1d@arrl.net>
4002 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
4005 2018-07-12 Paul Koning <ni1d@arrl.net>
4007 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
4008 zero reference, add doloop_end instead.
4009 * doc/md.texi (decrement_and_branch_until_zero): Remove.
4010 (Looping patterns): Remove decrement_and_branch_until_zero. Add
4011 detail for doloop_end.
4013 2018-07-12 Martin Sebor <msebor@redhat.com>
4016 * attribs.c (decl_attributes): Reject conflicting attributes before
4017 calling attribute handlers.
4019 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
4021 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
4023 (gcc::dump_manager::get_dump_file_name): likewise.
4024 (dump_begin): Likewise.
4025 * dumpfile.h (dump_begin): Update prototype.
4026 (gcc::dump_manager::get_dump_file_name,
4027 gcc::dump_manager::get_dump_file_name): Update prototype.
4029 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4031 * internal-fn.h (vectorizable_internal_fn_p): New function.
4032 * tree-vect-slp.c (compatible_calls_p): Likewise.
4033 (vect_build_slp_tree_1): Remove nops argument. Handle calls
4034 to internal functions.
4035 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
4037 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4039 * fold-const.h (inverse_conditions_p): Declare.
4040 * fold-const.c (inverse_conditions_p): New function.
4041 * match.pd: Use inverse_conditions_p. Add folds of view_converts
4042 that test the inverse condition of a conditional internal function.
4043 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
4044 * internal-fn.c (internal_fn_mask_index): Handle conditional
4046 (vectorized_internal_fn_supported_p): New function.
4047 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
4048 (any_pred_load_store): Replace with...
4049 (need_to_predicate): ...this new variable.
4050 (redundant_ssa_names): New variable.
4051 (ifcvt_can_use_mask_load_store): Move initial checks to...
4052 (ifcvt_can_predicate): ...this new function. Handle tree codes
4053 for which a conditional internal function exists.
4054 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
4055 instead of ifcvt_can_use_mask_load_store. Update after variable
4057 (predicate_load_or_store): New function, split out from
4058 predicate_mem_writes.
4059 (check_redundant_cond_expr): New function.
4060 (value_available_p): Likewise.
4061 (predicate_rhs_code): Likewise.
4062 (predicate_mem_writes): Rename to...
4063 (predicate_statements): ...this. Use predicate_load_or_store
4064 and predicate_rhs_code.
4065 (combine_blocks, tree_if_conversion): Update after above name changes.
4066 (ifcvt_local_dce): Handle redundant_ssa_names.
4067 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
4068 general conditional functions.
4069 * tree-vect-stmts.c (vectorizable_call): Likewise.
4071 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4072 Alan Hayward <alan.hayward@arm.com>
4073 David Sherwood <david.sherwood@arm.com>
4075 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
4076 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
4077 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
4078 plus and minus and convert them into IFN_COND_FMA-based sequences.
4079 (convert_mult_to_fma): Handle conditional plus and minus.
4081 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4083 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
4084 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
4085 (cond_fnms_optab): New optabs.
4086 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
4088 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
4089 * internal-fn.h (get_conditional_internal_fn): Declare.
4090 (get_unconditional_internal_fn): Likewise.
4091 * internal-fn.c (cond_ternary_direct): New macro.
4092 (expand_cond_ternary_optab_fn): Likewise.
4093 (direct_cond_ternary_optab_supported_p): Likewise.
4094 (FOR_EACH_COND_FN_PAIR): Likewise.
4095 (get_conditional_internal_fn): New function.
4096 (get_unconditional_internal_fn): Likewise.
4097 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
4098 (gimple_match_op::gimple_match_op): Add a new overload for 5
4100 (gimple_match_op::set_op): Likewise.
4101 (gimple_resimplify5): Declare.
4102 * genmatch.c (decision_tree::gen): Generate simplifications for
4104 * gimple-match-head.c (gimple_simplify): Define an overload for
4105 5 operands. Handle calls with 5 arguments in the top-level overload.
4106 (convert_conditional_op): Handle conversions from unconditional
4107 internal functions to conditional ones.
4108 (gimple_resimplify5): New function.
4109 (build_call_internal): Pass a fifth operand.
4110 (maybe_push_res_to_seq): Likewise.
4111 (try_conditional_simplification): Try converting conditional
4112 internal functions to unconditional internal functions.
4113 Handle 3-operand unconditional forms.
4114 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
4115 Define ternary equivalents of the current rules for binary conditional
4117 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
4119 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
4120 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
4121 (optab): Handle them.
4122 (SVE_COND_FP_TERNARY): New int iterator.
4123 (sve_fmla_op, sve_fmad_op): New int attributes.
4124 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
4125 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
4126 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
4128 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
4130 * target.def (preferred_else_value): New target hook.
4131 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
4132 * doc/tm.texi: Regenerate.
4133 * targhooks.h (default_preferred_else_value): Declare.
4134 * targhooks.c (default_preferred_else_value): New function.
4135 * internal-fn.h (conditional_internal_fn_code): Declare.
4136 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
4137 (get_conditional_internal_fn): Use it.
4138 (conditional_internal_fn_code): New function.
4139 * gimple-match.h (gimple_match_cond): New struct.
4140 (gimple_match_op): Add a cond member function.
4141 (gimple_match_op::gimple_match_op): Update all forms to take a
4143 * genmatch.c (expr::gen_transform): Use the same condition as res_op
4144 for the suboperation, but don't specify a particular else_value.
4145 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
4146 (visit_nary_op, visit_reference_op_load): Pass
4147 gimple_match_cond::UNCOND to the gimple_match_op constructor.
4148 * gimple-match-head.c: Include tree-eh.h
4149 (convert_conditional_op): New function.
4150 (maybe_resimplify_conditional_op): Likewise.
4151 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
4152 (gimple_resimplify2): Likewise.
4153 (gimple_resimplify3): Likewise.
4154 (gimple_resimplify4): Likewise.
4155 (maybe_push_res_to_seq): Return null for conditional operations.
4156 (try_conditional_simplification): New function.
4157 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
4159 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
4161 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
4163 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
4165 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
4167 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
4169 (hash_tree): Do not hash DECL_FCONTEXT
4170 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
4171 Do not stream DECL_FCONTEXT.
4172 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
4173 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
4175 2018-07-12 Richard Biener <rguenther@suse.de>
4178 * dwarf2out.c (gen_block_die): Only output blocks when they have
4179 at least one !DECL_IGNORED_P variable.
4181 2018-07-12 Richard Biener <rguenther@suse.de>
4184 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
4185 Remove -mieee-fp handling.
4187 2018-07-12 Richard Biener <rguenther@suse.de>
4189 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
4190 left-over from last patch.
4192 2018-07-12 Jakub Jelinek <jakub@redhat.com>
4194 PR tree-optimization/86492
4195 * gimple-ssa-store-merging.c
4196 (imm_store_chain_info::coalesce_immediate_stores): Call
4197 check_no_overlap even for the merge_overlapping case. Formatting fix.
4199 2018-07-12 Richard Biener <rguenther@suse.de>
4202 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
4203 move possibly trapping operations into the conditional.
4205 2018-07-12 Richard Biener <rguenther@suse.de>
4207 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
4208 (vn_lookup_simplify_result): Remove recursion limit applied
4210 (vn_nary_build_or_lookup_1): Adjust.
4211 (try_to_simplify): Likewise.
4212 * gimple-match-head.c (gimple_resimplify1): Instead apply one
4214 (gimple_resimplify2): Likewise.
4215 (gimple_resimplify3): Likewise.
4216 (gimple_resimplify4): Likewise.
4218 2018-07-11 Jakub Jelinek <jakub@redhat.com>
4220 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
4221 Use __mmask64 type instead of __mmask8 for __M argument.
4222 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
4223 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
4225 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
4226 __mmask16 for __M argument.
4227 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
4228 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
4229 to __mmask16 instead of __mmask8.
4230 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
4231 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
4232 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
4233 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
4234 instead of __mmask16 for __U argument.
4235 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
4236 __mmask16 instead of __mmask8 for __U argument.
4237 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
4239 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
4241 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
4243 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
4245 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
4247 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
4249 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
4251 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
4253 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
4255 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
4257 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
4258 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
4259 return type as well as __M argument type and all casts from __mmask8
4261 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
4262 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
4263 return type as well as __M argument type and all casts from __mmask8
4265 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
4266 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
4267 return type as well as __M argument type and all casts from __mmask8
4269 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
4270 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
4271 return type as well as __M argument type and all casts from __mmask8
4273 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
4274 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
4277 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
4279 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
4280 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
4283 2018-07-11 Paul Koning <ni1d@arrl.net>
4285 * doc/md.texi (define_subst): Document how multiple occurrences of
4286 the same argument in the replacement pattern are handled.
4288 2018-07-11 Paul Koning <ni1d@arrl.net>
4290 * doc/extend.texi (Common Variable Attributes): Move "mode" into
4292 (Common Type Attributes): Add "mode" attribute.
4294 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
4296 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
4297 stream DECL_ORIGINAL_TYPE.
4298 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
4299 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
4300 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
4301 Do not walk original type.
4302 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
4304 (write_ts_decl_non_common_tree_pointers): Do not stream
4306 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
4307 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
4309 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
4311 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
4312 thread twice from the same starting edge.
4314 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
4316 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
4317 * gimple.c (gimple_call_nonnull_result_p): ...here...
4318 (gimple_call_nonnull_arg): ...and here.
4319 * gimple.h (gimple_call_nonnull_result_p): New.
4320 (gimple_call_nonnull_arg): New.
4322 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
4324 * config/arm/arm-cpus.in: Move information from fpu field of each
4325 cpu definition to the isa field.
4326 * config/arm/parsecpu.awk (fpu): Delete match rule.
4327 (gen_comm_data): Don't add bits from the CPU's FPU entry.
4329 2018-07-11 Richard Biener <rguenther@suse.de>
4332 * dwarf2out.c (init_sections_and_labels): Use
4333 output_asm_line_debug_info consistently.
4334 (dwarf2out_early_finish): Likewise.
4335 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
4338 2018-07-11 Richard Biener <rguenther@suse.de>
4340 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
4341 Rework father_bb setting in a way to avoid propagating constants
4342 multiple times on a loop body.
4344 2018-07-10 Mark Wielaard <mark@klomp.org>
4347 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
4349 2018-07-10 Richard Biener <rguenther@suse.de>
4351 * hash-map.h (hash_map::iterator::operator*): Return
4352 references to key and value.
4354 2018-07-10 Jakub Jelinek <jakub@redhat.com>
4357 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
4358 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
4359 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
4360 (gimplify_omp_for): For composite loops, move outer
4361 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
4362 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
4363 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
4364 TREE_LIST for both the original class iterator and the "last" helper
4365 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
4366 loop, remember has_decl_expr from outer composite loops for the
4367 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
4369 2018-07-09 Martin Sebor <msebor@redhat.com>
4373 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
4374 accessing implicitly initialized array elements.
4375 * expr.c (string_constant): Handle string initializers of
4376 character arrays within aggregates.
4377 * gimple-fold.c (fold_array_ctor_reference): Add argument.
4378 Store element offset. As a special case, handle zero size.
4379 (fold_nonarray_ctor_reference): Same.
4380 (fold_ctor_reference): Add argument. Store subobject offset.
4381 * gimple-fold.h (fold_ctor_reference): Add argument.
4383 2018-07-09 Paul Koning <ni1d@arrl.net>
4385 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
4386 (pdp11_insn_cost): New function.
4387 (pdp11_md_asm_adjust): New function.
4388 (TARGET_INVALID_WITHIN_DOLOOP): Define.
4389 (pdp11_rtx_costs): Update to match machine better.
4390 (output_addr_const_pdp11): Correct format mismatch warnings.
4391 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
4392 * config/pdp11/pdp11.md: General change to add base_cost and/or
4393 length attributes for use by new pdp11_insn_cost function.
4394 (MIN_BRANCH): Correct definition.
4396 (doloop_end): Use standard pattern name for looping pattern.
4397 (doloop_end_nocc): New.
4398 (movsf): Add another constraint alternative.
4399 (zero_extendqihi2): Add constraint alternatives for not in place
4401 (zero_extendhisi2): Remove.
4402 (shift patterns): Add CC handling variants.
4406 (define_peephole2): New peephole to recognize mov that sets CC for
4409 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4411 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
4412 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
4415 2018-07-09 Jakub Jelinek <jakub@redhat.com>
4418 * real.c (real_nextafter): Return true if result is denormal.
4420 2018-07-09 Martin Liska <mliska@suse.cz>
4422 * common.opt: Add back wrongly removed attribute.
4424 2018-07-09 Richard Biener <rguenther@suse.de>
4427 * dwarf2out.c (gen_block_die): For an early generated DIE
4428 always output high/low PC attributes.
4430 2018-07-09 Tom de Vries <tdevries@suse.de>
4432 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
4433 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
4434 onto VAR_DECL with abstract origin.
4436 2018-07-07 Jim Wilson <jimw@sifive.com>
4438 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
4440 2018-07-07 Tom de Vries <tdevries@suse.de>
4442 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
4444 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
4446 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
4448 (combine_bound): Use wide-int overflow calculation instead of
4450 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
4452 * fold-const.c (int_const_binop_2): Same.
4453 (extract_muldiv_1): Same.
4454 (fold_div_compare): Same.
4455 (fold_abs_const): Same.
4457 * poly-int.h (add): Same.
4461 * predict.c (predict_iv_comparison): Same.
4462 * profile-count.c (slow_safe_scale_64bit): Same.
4463 * simplify-rtx.c (simplify_const_binary_operation): Same.
4464 * tree-chrec.c (tree_fold_binomial): Same.
4465 * tree-data-ref.c (split_constant_offset_1): Same.
4466 * tree-if-conv.c (idx_within_array_bound): Same.
4467 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
4468 * tree-ssa-phiopt.c (minmax_replacement): Same.
4469 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
4470 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
4471 * vr-values.c (vr_values::adjust_range_with_scev): Same.
4472 * wide-int.cc (wi::add_large): Same.
4473 (wi::mul_internal): Same.
4474 (wi::sub_large): Same.
4475 (wi::divmod_internal): Same.
4476 * wide-int.h: Change overflow type to overflow_type for neg, add,
4477 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
4478 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
4479 mul_internal, divmod_internal.
4480 (overflow_type): New enum.
4481 (accumulate_overflow): New.
4483 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
4485 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
4486 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
4488 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
4490 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
4491 argument is checked for zero before entering loop, avoid checking again.
4493 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
4495 * gimplify.h (generic_expr_could_trap_p): Set as global function.
4496 * gimplify.h (generic_expr_could_trap_p): Likwise.
4497 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
4499 2018-07-06 Jakub Jelinek <jakub@redhat.com>
4501 PR tree-optimization/86401
4502 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
4503 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
4504 (fold_bit_and_mask): ... here. New helper function for match.pd.
4505 * fold-const.h (fold_bit_and_mask): Declare.
4506 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
4508 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
4511 * target.def (translate_mode_attribute): New hook.
4512 * targhooks.h (default_translate_mode_attribute): Declare.
4513 * targhooks.c (default_translate_mode_attribute): New function.
4514 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
4515 * doc/tm.texi: Regenerate.
4516 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
4517 (rs6000_translate_mode_attribute): New function.
4519 2018-07-06 Paul Koning <ni1d@arrl.net>
4521 * doc/md.texi (define_split): Document DONE and FAIL.
4522 (define_peephole2): Ditto.
4524 2018-07-05 Jeff Law <law@redhat.com>
4526 PR tree-optimization/86010
4527 * tree-ssa-dse.c (compute_trims): More aggressively trim at
4528 both the head and tail of mem* and str* calls.
4530 2018-07-05 Jim Wilson <jimw@sifive.com>
4532 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
4534 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
4536 * config/aarch64/aarch64-simd.md: correct flags text for
4537 MIN_EXPR replacement.
4539 2018-07-05 James Clarke <jrtc27@jrtc27.com>
4541 * configure: Regenerated.
4543 2018-07-05 Carl Love <cel@us.ibm.com>
4545 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
4546 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
4547 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
4548 VSX_BUILTIN_DOUBLEL_V4SF.
4550 2018-07-05 Martin Sebor <msebor@redhat.com>
4553 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
4554 than its domain to compute its the upper bound of a char array.
4556 2018-07-05 Nathan Sidwell <nathan@acm.org>
4558 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
4559 * doc/cpp.texi: Update comment.
4560 * doc/tm.texi: Rebuilt.
4561 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
4562 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
4563 * doc/extend.texi (Backwards Compatibility): Clarify it is system
4564 headers affected by extern "C".
4565 * system.h: Poison NO_IMPLICIT_EXTERN_C.
4566 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
4567 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
4568 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
4569 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
4570 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
4571 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
4572 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
4573 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
4574 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
4575 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
4576 config/sparc/sp64-elf.h, config/spu/spu.h,
4577 config/stormy16/stormy16.h, config/v850/v850.h,
4578 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
4579 define NO_IMPLICIT_EXTERN_C.
4580 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
4582 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4585 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4586 instead of GET_MODE_SIZE when comparing Units.
4588 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4591 * rtlanal.c (set_noop_p): Constrain on mode change,
4592 include hard-reg-set.h
4594 2018-07-05 Tamar Christina <tamar.christina@arm.com>
4596 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
4598 2018-07-05 Jakub Jelinek <jakub@redhat.com>
4601 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4604 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4606 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4608 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
4611 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
4613 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
4615 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
4618 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
4620 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
4622 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
4623 neg_*_op* variables.
4625 2018-07-04 Martin Liska <mliska@suse.cz>
4627 * tree-switch-conversion.c: Define
4628 max_ratio_for_speed and max_ratio_for_size constants.
4630 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4631 Martin Liska <mliska@suse.cz>
4636 * common.opt: Rename align options with 'str_' prefix.
4637 * common/config/i386/i386-common.c (set_malign_value): New
4639 (ix86_handle_option): Use it to set -falign-* options/
4640 * config/aarch64/aarch64-protos.h (struct tune_params): Change
4641 type from int to string.
4642 * config/aarch64/aarch64.c: Update default values from int
4644 * config/alpha/alpha.c (alpha_override_options_after_change):
4646 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
4647 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4648 max skip conditionally.
4649 * config/i386/freebsd.h (SUBALIGN_LOG): New.
4650 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4651 max skip conditionally.
4652 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4653 max skip conditionally.
4654 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
4655 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4656 max skip conditionally.
4657 * config/i386/i386.c (struct ptt): Change type from int to
4659 (ix86_default_align): Set default values.
4660 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
4661 max skip conditionally.
4662 * config/i386/iamcu.h (SUBALIGN_LOG): New.
4663 (ASM_OUTPUT_MAX_SKIP_ALIGN):
4664 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
4665 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4666 max skip conditionally.
4667 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
4668 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
4669 * config/i386/x86-64.h (SUBALIGN_LOG): New.
4670 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
4671 max skip conditionally.
4672 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4673 * config/ia64/ia64.c (ia64_option_override): Set default values
4674 for alignment options.
4675 * config/m68k/m68k.c: Handle new str_align_* options.
4676 * config/mips/mips.c (mips_set_compression_mode): Change
4678 (mips_option_override): Set default values for options.
4679 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
4681 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4683 * config/rx/rx.c (rx_option_override): Likewise.
4684 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
4685 (LABEL_ALIGN): Use align_labels_log.
4686 (LOOP_ALIGN): Use align_loops_align.
4687 * config/s390/s390.c (s390_asm_output_function_label): Use new
4689 * config/sh/sh.c (sh_override_options_after_change):
4690 Change type of constants.
4691 * config/spu/spu.c (spu_sched_init): Likewise.
4692 * config/sparc/sparc.c (sparc_option_override): Set default
4694 * config/visium/visium.c (visium_option_override): Likewise.
4695 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
4696 emit p2align format with last argument if it's not needed.
4697 * doc/invoke.texi: Document extended format of -falign-*.
4698 * final.c: Use align_labels alignment.
4699 * flags.h (struct target_flag_state): Change type to use
4701 (struct align_flags_tuple): New.
4702 (struct align_flags): Likewise.
4703 (align_loops_log): Redefine macro to use new types.
4704 (align_loops_max_skip): Redefine macro to use new types.
4705 (align_jumps_log): Redefine macro to use new types.
4706 (align_jumps_max_skip): Redefine macro to use new types.
4707 (align_labels_log): Redefine macro to use new types.
4708 (align_labels_max_skip): Redefine macro to use new types.
4709 (align_functions_log): Redefine macro to use new types.
4710 (align_loops): Redefine macro to use new types.
4711 (align_jumps): Redefine macro to use new types.
4712 (align_labels): Redefine macro to use new types.
4713 (align_functions): Redefine macro to use new types.
4714 (align_functions_max_skip): Redefine macro to use new types.
4715 (align_loops_value): New macro.
4716 (align_jumps_value): New macro.
4717 (align_labels_value): New macro.
4718 (align_functions_value): New macro.
4719 * function.c (invoke_set_current_function_hook): Propagate
4720 alignment values from flags to global variables default in
4722 * ipa-icf.c (sem_function::equals_wpa): Use
4723 cl_optimization_option_eq instead of memcmp.
4724 * lto-streamer.h (cl_optimization_stream_out): Support streaming
4726 (cl_optimization_stream_in): Likewise.
4727 * optc-save-gen.awk: Support strings in cl_optimization.
4728 * opth-gen.awk: Likewise.
4729 * opts.c (finish_options): Remove error checking of invalid
4731 (MAX_CODE_ALIGN): Remove.
4732 (MAX_CODE_ALIGN_VALUE): Likewise.
4733 (parse_and_check_align_values): New function.
4734 (check_alignment_argument): Likewise.
4735 (common_handle_option): Use check_alignment_argument.
4736 * opts.h (parse_and_check_align_values): Declare.
4737 * toplev.c (init_alignments): Remove.
4738 (read_log_maxskip): New.
4739 (parse_N_M): Likewise.
4740 (parse_alignment_opts): Likewise.
4741 (backend_init_target): Remove usage of init_alignments.
4742 * toplev.h (parse_alignment_opts): Declare.
4743 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
4745 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
4746 * tree.c (cl_option_hasher::equal): New.
4747 * varasm.c: Use new global macros.
4749 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
4751 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
4752 Use a simpler align directive also if MAXSKIP = ALIGN-1.
4753 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4754 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4755 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4756 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
4757 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
4758 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
4759 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
4760 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4761 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4762 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4763 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
4765 2018-07-04 Martin Liska <mliska@suse.cz>
4766 Jonathan Wakely <jwakely@redhat.com>
4768 * coverage.c: Use correct type.
4769 * doc/invoke.texi: Language correction.
4771 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
4774 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
4775 ENDBRANCH for non-tail call which may return via indirect branch.
4776 * doc/extend.texi: Document indirect_return attribute.
4778 2018-07-03 Martin Sebor <msebor@redhat.com>
4780 PR tree-optimization/86274
4781 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
4783 (format_floating): Correct handling of infinities and NaNs.
4785 2018-07-03 Martin Sebor <msebor@redhat.com>
4787 * print-tree.c (print_real_cst): New function.
4788 (print_node_brief): Call it.
4789 (print_node): Ditto.
4791 2018-07-03 Jeff Law <law@redhat.com>
4793 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
4794 into a single pattern.
4796 * config/h8300/h8300.md (ors code_iterator): New.
4797 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
4798 a single pattern and single splitter.
4799 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
4800 (iorqi3_1, xorqi3_1): Likewise.
4801 (iorqi3, xorqi3 expanders): Similarly.
4803 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
4804 (movmd_internal) into a single pattern using the P mode iterator.
4805 (movmd splitters): Similarly.
4806 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
4807 (movsd splitters): Similarly.
4809 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
4810 ADDB, ADDW and ADDL into a single ADD attribute which selects the
4811 right table based on the size of the operand.
4812 * config/h8300/h8300.md (length_table): Corresponding changes. All
4813 references to "addb", "addw" and "addl" changed to "add".
4814 (btst patterns): Merge two variants into a single pattern.
4815 (tstqi, tsthi): Likewise.
4816 (addhi3_incdec, addsi3_incdec): Likewise.
4817 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
4818 (mulhi3, mulsi3): Likewise.
4819 (udivhi3, udivsi3): Likewise.
4820 (divhi3, divsi3): Likewise.
4821 (andorqi3, andorhi3, andorsi3): Likewise.
4823 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
4826 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
4827 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
4829 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4831 PR tree-optimization/85694
4832 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
4833 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
4835 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
4836 (<u>avg<mode>3_ceil): New patterns.
4838 2018-07-03 David Malcolm <dmalcolm@redhat.com>
4840 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
4841 scan-tree-dump directive.
4842 * gcc.dg/vect/slp-perm-2.c: Likewise.
4843 * gcc.dg/vect/slp-perm-3.c: Likewise.
4844 * gcc.dg/vect/slp-perm-5.c: Likewise.
4845 * gcc.dg/vect/slp-perm-6.c: Likewise.
4846 * gcc.dg/vect/slp-perm-7.c: Likewise.
4847 * gcc.dg/vect/slp-perm-8.c: Likewise.
4849 2018-07-03 Marek Polacek <polacek@redhat.com>
4852 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
4854 2018-07-03 Richard Biener <rguenther@suse.de>
4857 * tree-ssa-structalias.c (find_func_clobbers): Properly
4858 handle indirect calls.
4860 2018-07-03 Jeff Law <law@redhat.com>
4862 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
4863 (shifts): New code iterator.
4864 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
4865 expander. Fix HImode handling on H8/SX.
4866 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
4867 (subqi3, subhi3, subsi3 expanders): Likewise.
4868 (andqi3, andhi3, andsi3 expanders): Likewise.
4869 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
4870 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
4871 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
4872 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
4873 (zero_extendqihi2, zero_extendqisi2): Likewise.
4874 (extendqihi2, extendqisi2): Likewise.
4875 (rotlqi3, rotlhi3, rotlsi3): Likewise.
4876 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
4877 (rotlqi3_1, rotlhi3_1): Likewise.
4878 (logicalhi3_sn, logicalsi3_sn): Likewise.
4879 (logicalhi3, logicalsi3): Likewise.
4881 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4883 * tree-vect-patterns.c (vect_recog_rotate_pattern)
4884 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
4885 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
4886 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
4887 type to append_pattern_def_seq instead of creating a stmt_vec_info
4889 (build_mask_conversion): Likewise. Remove vinfo argument.
4890 (vect_add_conversion_to_patterm): Likewise, renaming to...
4891 (vect_add_conversion_to_pattern): ...this.
4892 (vect_recog_mask_conversion_pattern): Update call to
4893 build_mask_conversion. Pass the vector type to
4894 append_pattern_def_seq here too.
4895 (vect_recog_gather_scatter_pattern): Update call to
4896 vect_add_conversion_to_pattern.
4898 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4900 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
4901 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4902 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
4903 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
4904 STMT_VINFO_PATTERN_DEF_SEQ to null here.
4905 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
4906 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
4907 append_pattern_def_seq instead of new_pattern_def_seq.
4908 (vect_recog_divmod_pattern): Do both of the above.
4909 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
4912 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4914 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
4915 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
4916 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
4917 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
4918 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
4919 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
4920 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
4921 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4922 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4923 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
4924 parameter with a single stmt_vec_info.
4925 (vect_recog_func_ptr): Likewise.
4926 (vect_recog_gather_scatter_pattern): Likewise, folding in...
4927 (vect_try_gather_scatter_pattern): ...this.
4928 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
4929 the stmt_vec_info of the statement to be matched. Don't clear
4930 STMT_VINFO_RELATED_STMT.
4931 (vect_pattern_recog): Update call accordingly.
4933 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4935 PR tree-optimization/85694
4936 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
4937 (uavgM3_ceil): Document new optabs.
4938 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
4939 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
4941 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
4942 (savg_ceil_optab): New optabs.
4943 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
4944 (vect_vect_recog_func_ptrs): Add it.
4945 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
4946 constant directly from the associated lhs.
4948 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4950 * tree-vect-patterns.c (vect_split_statement): New function.
4951 (vect_convert_input): Use it to try to split an existing cast.
4953 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4955 * poly-int.h (print_hex): New function.
4956 * dumpfile.h (dump_dec, dump_hex): Declare.
4957 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
4958 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
4959 min_input_precision, operation_precision and operation_sign.
4960 * tree-vect-patterns.c (vect_get_range_info): New function.
4961 (vect_same_loop_or_bb_p, vect_single_imm_use)
4962 (vect_operation_fits_smaller_type): Delete.
4963 (vect_look_through_possible_promotion): Add an optional
4964 single_use_p parameter.
4965 (vect_recog_over_widening_pattern): Rewrite to use new
4966 stmt_vec_info infomration. Handle one operation at a time.
4967 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
4968 (vect_truncatable_operation_p, vect_set_operation_type)
4969 (vect_set_min_input_precision): New functions.
4970 (vect_determine_min_output_precision_1): Likewise.
4971 (vect_determine_min_output_precision): Likewise.
4972 (vect_determine_precisions_from_range): Likewise.
4973 (vect_determine_precisions_from_users): Likewise.
4974 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
4975 (vect_vect_recog_func_ptrs): Put over_widening first.
4977 (vect_pattern_recog): Call vect_determine_precisions.
4979 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
4981 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
4982 statements that have been replaced by further pattern statements.
4983 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
4985 2018-07-03 Richard Biener <rguenther@suse.de>
4987 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
4988 always set *dt. Dump vectype in vectype overload.
4989 * dumpfile.h (dump_gimple_expr): New function.
4990 (dump_gimple_expr_loc): Likewise.
4991 * dumpfile.c (dump_gimple_expr): New function.
4992 (dump_gimple_expr_loc): Likewise.
4994 2018-07-02 Jeff Law <law@redhat.com>
4996 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
4997 the H8/300, H8/300H and H8/S variants into a single pattern.
4998 (movhi_h8300, movqi_h8300hs): Similarly.
4999 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
5000 (QHI mode iterator): New.
5002 * config/h8300/h8300.md: Remove trailing whitespace.
5004 2018-07-02 Jim Wilson <jimw@sifive.com>
5006 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
5007 instead of emit_insn for interrupt returns.
5008 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
5009 (riscv_sret, riscv_uret): Likewise.
5011 2018-07-02 David Malcolm <dmalcolm@redhat.com>
5013 * pretty-print.c (selftest::test_pp_format): Move save and restore
5014 of quotes to class auto_fix_quotes, and add an instance.
5015 * selftest.c: Include "intl.h".
5016 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
5017 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
5018 * selftest.h (selftest::auto_fix_quotes): New class.
5020 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
5022 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
5023 (aarch64_sve_prepare_conditional_op): Remove.
5024 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
5025 Allow aarch64_simd_reg_or_zero as select operand; remove
5026 the aarch64_sve_prepare_conditional_op call.
5027 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
5028 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
5029 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
5030 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
5031 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
5032 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
5033 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
5034 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
5035 and a splitters to match all of the *_any patterns.
5036 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
5038 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
5039 (SVE_COND_FP_BINARY_REV): Remove.
5040 (sve_int_op_rev, sve_fp_op_rev): New.
5041 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
5042 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
5043 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
5044 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
5045 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
5046 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
5047 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
5048 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
5049 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
5051 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
5052 Remove match_dup 1 from the inner unspec.
5053 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
5055 * config/aarch64/aarch64.md (movprfx): New attr.
5056 (length): Default movprfx to 8.
5057 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
5058 (*madd<SVE_I>, *msub<SVE_I): Likewise.
5059 (*<su>mul<SVE_I>3_highpart): Likewise.
5060 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
5061 (*v<ASHIFT><SVE_I>3): Likewise.
5062 (*<su><MAXMIN><SVE_I>3): Likewise.
5063 (*<su><MAXMIN><SVE_F>3): Likewise.
5064 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
5065 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
5066 (*div<SVE_F>4): Likewise.
5068 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
5070 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
5073 2018-07-02 Richard Biener <rguenther@suse.de>
5075 PR tree-optimization/86363
5076 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
5077 memset argument refers to a non-variable address.
5079 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
5081 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
5082 {PLUS,MINUS}_EXPR code to...
5083 (adjust_symbolic_bound): ...here,
5084 (combine_bound): ...here,
5085 (set_value_range_with_overflow): ...and here.
5087 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
5089 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
5091 (extract_range_from_abs_expr): ...here.
5093 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
5095 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
5096 -fno-omit-frame-pointer when not optimizing.
5098 2018-07-02 Martin Liska <mliska@suse.cz>
5101 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
5102 (propagate_nothrow): Likewise.
5104 2018-07-02 Martin Liska <mliska@suse.cz>
5107 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
5109 2018-07-02 David Malcolm <dmalcolm@redhat.com>
5111 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
5112 function in r262149, changing "loc" param from source_location to
5113 const dump_location_t &.
5114 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
5115 declaration, as above.
5117 2018-07-01 Paul Koning <ni1d@arrl.net>
5119 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
5120 -munit-asm, -mgnu-asm, -mdec-asm.
5121 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
5122 (pdp11_output_labelref): New.
5123 (pdp11_output_def): New.
5124 (pdp11_output_addr_vec_elt): New.
5125 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
5126 %# and %@ format codes.
5127 (pdp11_option_override): New.
5128 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
5129 (pdp11_output_ident): New.
5130 (pdp11_asm_named_section): New.
5131 (pdp11_asm_init_sections): New.
5132 (pdp11_file_start): New.
5133 (pdp11_file_end): New.
5134 (output_ascii): Use .ascii/.asciz for -mdec-asm.
5135 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
5136 %o, like %c but octal.
5137 (pdp11_option_override): New.
5138 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
5140 (DATA_SECTION_ASM_OP): Ditto.
5141 (READONLY_DATA_SECTION_ASM_OP): New.
5142 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
5143 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
5144 (ASM_OUTPUT_LABELREF): Ditto.
5145 (ASM_OUTPUT_DEF): Ditto.
5146 (ASM_OUTPUT_EXTERNAL): New.
5147 (ASM_OUTPUT_SOURCE_FILENAME): New.
5148 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
5149 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
5150 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
5151 %# and %@ format codes.
5152 * config/pdp11/pdp11.opt (mgnu-asm): New.
5153 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
5154 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
5155 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
5157 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
5159 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
5160 dereferencing path[] beyond its length.
5162 (debug_all_paths): New.
5163 (rewire_first_differing_edge): New.
5164 (adjust_paths_after_duplication): New.
5165 (duplicate_thread_path): Call adjust_paths_after_duplication.
5167 (thread_through_all_blocks): Add new argument to
5168 duplicate_thread_path.
5170 2018-06-30 Jim Wilson <jimw@sifive.com>
5172 * config/riscv/predicates.md (p2m1_shift_operand): New.
5173 (high_mask_shift_operand): New.
5174 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
5175 pattern using p2m1_shift_operand.
5176 (lshsi3_zero_extend_3+2): New combiner pattern using
5177 high_mask_shift_operand.
5179 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5181 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
5183 (vect_recog_rotate_pattern): ...here.
5184 (vect_convert_input): Try to insert casts of invariants in the
5186 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
5187 preheader to be empty.
5189 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5191 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
5192 vector type. If given, install it in the new statement's
5194 (vect_element_precision): New function.
5195 (vect_unpromoted_value): New struct.
5196 (vect_unpromoted_value::vect_unpromoted_value): New function.
5197 (vect_unpromoted_value::set_op): Likewise.
5198 (vect_look_through_possible_promotion): Likewise.
5199 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
5200 (vect_widened_op_tree, vect_convert_input): Likewise.
5201 (vect_convert_inputs, vect_convert_output): Likewise.
5202 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
5203 to handle the optional cast of the multiplication result and
5204 vect_widened_op_tree to detect the widened multiplication itself.
5205 Do not require the input and output of promotion casts to have
5206 the same sign, but base the signedness of the operation on the
5207 input rather than the result. If the pattern includes two
5208 promotions, check that those promotions have the same sign.
5209 Do not restrict the MULT_EXPR handling to a double-width result;
5210 handle quadruple-width results and wider. Use vect_convert_inputs
5211 to convert the inputs to the common type.
5212 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
5213 to handle the optional cast of the ABS result. Also allow a sign
5214 change or a sign extension between the ABS and MINUS.
5215 Use vect_widened_op_tree to detect the widened subtraction and use
5216 vect_convert_inputs to convert the inputs to the common type.
5217 (vect_handle_widen_op_by_const): Delete.
5218 (vect_recog_widen_op_pattern): New function.
5219 (vect_recog_widen_mult_pattern): Use it.
5220 (vect_recog_widen_shift_pattern): Likewise.
5221 (vect_recog_widen_sum_pattern): Use
5222 vect_look_through_possible_promotion to handle the promoted
5225 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5227 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
5228 the containing gimple_seq *.
5229 * gimple-iterator.h (gsi_for_stmt): Declare it.
5230 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
5231 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
5232 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
5233 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
5234 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
5236 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
5238 (vect_mark_pattern_stmts): ...here. Handle cases in which the
5239 statement being replaced is part of an existing pattern
5240 definition sequence, inserting the new pattern statements before
5242 (vect_pattern_recog_1): Don't return a bool. If the statement
5243 is already part of a pattern, instead apply pattern matching
5244 to the pattern definition statements. Don't clear the
5245 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
5246 (vect_pattern_recog): Don't break after the first match;
5247 continue processing the pattern definition statements instead.
5248 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
5250 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5252 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5253 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
5254 (vect_recog_widen_sum_pattern): Use it.
5256 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5258 * tree-vect-loop.c (vectorizable_reduction): Assert that the
5259 phi is not a pattern statement and has not been replaced by
5260 a pattern statement.
5261 * tree-vect-patterns.c (type_conversion_p): Don't check
5262 STMT_VINFO_IN_PATTERN_P.
5263 (vect_recog_vector_vector_shift_pattern): Likewise.
5264 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
5265 the pattern statement rather than the original statement; check
5266 directly for a WIDEN_MULT_EXPR here.
5267 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
5268 vect_is_simple_use to return the pattern statement rather
5269 than the original statement; use is_pattern_stmt_p to check
5270 for such a pattern statement.
5271 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
5272 to return the pattern statement rather than the original statement;
5273 don't do the same transformation here.
5274 (vect_is_simple_use): If the defining statement has been replaced
5275 by a pattern statement, return the pattern statement instead.
5276 Remove the corresponding (local) transformation from the vectype
5279 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
5281 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
5282 end and default to null.
5283 * tree-vect-loop.c (vect_create_epilog_for_reduction)
5284 (vectorizable_reduction): Update calls accordingly, dropping the
5285 gimple ** argument if the passed-back statement isn't needed.
5286 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
5287 (vect_recog_rotate_pattern): Likewise.
5288 (vect_recog_mask_conversion_pattern): Likewise.
5289 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
5290 (vect_mask_constant_operand_p): Likewise.
5291 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
5292 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
5293 (get_group_load_store_type, get_load_store_type): Likewise.
5294 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
5295 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
5296 (vectorizable_conversion, vectorizable_assignment): Likewise.
5297 (vectorizable_shift, vectorizable_operation): Likewise.
5298 (vectorizable_store, vect_is_simple_cond): Likewise.
5299 (vectorizable_condition, vectorizable_comparison): Likewise.
5300 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
5301 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
5302 and move it to the end. Cope with null def_stmt_outs.
5304 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
5306 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
5308 2018-06-29 Jeff Law <law@redhat.com>
5310 * config/v850/v850.c (v850_legitimate_address_p): Handle large
5311 displacements for TARGET_V850E2V3 and newer.
5312 (TARGET_LRA_P): Remove. Defaults to LRA now.
5313 * config/v850/v850.md (sign23byte_load): Remove.
5314 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
5315 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
5317 2018-06-29 Martin Liska <mliska@suse.cz>
5320 * coverage.c (coverage_init): Mangle full path name.
5321 * doc/invoke.texi: Document the change.
5322 * gcov-io.c (mangle_path): New.
5323 * gcov-io.h (mangle_path): Likewise.
5324 * gcov.c (mangle_name): Use mangle_path for path mangling.
5326 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5328 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
5329 if starting source register is not even.
5331 2018-06-29 Martin Liska <mliska@suse.cz>
5333 PR tree-optimization/86263
5334 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
5335 Make edge redirection.
5337 2018-06-29 David Malcolm <dmalcolm@redhat.com>
5339 * dumpfile.c (dump_loc): Add indentation based on scope depth.
5340 (dump_scope_depth): New variable.
5341 (get_dump_scope_depth): New function.
5342 (dump_begin_scope): New function.
5343 (dump_end_scope): New function.
5344 * dumpfile.h (get_dump_scope_depth): New declaration.
5345 (dump_begin_scope): New declaration.
5346 (dump_end_scope): New declaration.
5347 (class auto_dump_scope): New class.
5348 (AUTO_DUMP_SCOPE): New macro.
5349 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
5352 2018-06-29 Richard Biener <rguenther@suse.de>
5354 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
5355 compute_all_dependences succeeds.
5356 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
5357 exceed --param loop-max-datarefs-for-datadeps.
5359 2018-06-29 Jakub Jelinek <jakub@redhat.com>
5361 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
5363 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
5366 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
5367 alternative 0 in preferred_for_speed attribute.
5369 2018-06-28 Paul Koning <ni1d@arrl.net>
5371 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
5372 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
5373 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
5374 * config/pdp11/pdp11.md: Correct "length" attribute calculation
5375 for shift insn patterns.
5377 2018-06-28 David Malcolm <dmalcolm@redhat.com>
5379 * cgraph.c (cgraph_node::get_body): Replace assignments to
5380 "dump_file" with calls to set_dump_file.
5381 * dumpfile.c (alt_dump_file): Make static, and group with...
5382 (alt_flags): ...this definition.
5383 (dumps_are_enabled): New variable.
5384 (refresh_dumps_are_enabled): New function.
5385 (set_dump_file): New function.
5386 (set_alt_dump_file): New function.
5387 (gcc::dump_manager::dump_start): Replace assignments to
5388 "dump_file" and "alt_dump_file" with calls to set_dump_file and
5390 (gcc::dump_manager::dump_finish): Likewise.
5391 * dumpfile.h (alt_dump_file): Delete decl.
5392 (dumps_are_enabled): New variable decl.
5393 (set_dump_file): New function decl.
5394 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
5396 * tree-nested.c (lower_nested_functions): Replace assignments to
5397 "dump_file" with calls to set_dump_file.
5399 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
5401 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
5402 goto_locus of each outgoing edge of each basic block.
5404 2018-06-28 Richard Biener <rguenther@suse.de>
5406 * dwarf2out.c (decl_scope_table): Remove.
5407 (push_decl_scope): Likewise.
5408 (pop_decl_scope): Likewise.
5409 (gen_type_die_for_member): Do not call push/pop_decl_scope.
5410 (gen_struct_or_union_type_die): Likewise.
5411 (gen_tagged_type_die): Likewise.
5412 (dwarf2out_init): Do not initialize decl_scope_table.
5413 (dwarf2out_c_finalize): Do not free it.
5415 2018-06-28 Richard Biener <rguenther@suse.de>
5417 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
5418 deciding whether to not re-use a DIE.
5420 2018-06-28 Richard Biener <rguenther@suse.de>
5422 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
5423 DW_AT_abstract_origin attribute.
5425 2018-06-28 Martin Liska <mliska@suse.cz>
5427 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
5428 Use newly introduced constants.
5429 * tree-switch-conversion.h (struct jump_table_cluster):
5430 Define max_ratio_for_size and max_ratio_for_speed.
5432 2018-06-28 Martin Liska <mliska@suse.cz>
5434 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
5435 Add new checking assert to catch invalid state.
5436 (jump_table_cluster::can_be_handled): Handle single case
5438 (jump_table_cluster::is_beneficial): Bail out for such case.
5439 (bit_test_cluster::find_bit_tests):
5440 Add new checking assert to catch invalid state.
5441 (bit_test_cluster::can_be_handled): Handle single case
5443 (bit_test_cluster::is_beneficial): Bail out for such case.
5444 (switch_decision_tree::analyze_switch_statement):
5447 2018-06-28 Martin Liska <mliska@suse.cz>
5449 * common.opt: Introduce -completion option.
5450 * gcc.c (driver_handle_option): Handle it.
5451 (driver::main): Print completions if completion
5453 * opt-suggestions.c (option_proposer::get_completions):
5455 (option_proposer::suggest_completion): Likewise.
5456 (option_proposer::find_param_completions): Likewise.
5457 (verify_autocompletions): Likewise.
5458 (test_completion_valid_options): Likewise.
5459 (test_completion_valid_params): Likewise.
5460 (in_completion_p): Likewise.
5461 (empty_completion_p): Likewise.
5462 (test_completion_partial_match): Likewise.
5463 (test_completion_garbage): Likewise.
5464 (opt_proposer_c_tests): Likewise.
5465 * opt-suggestions.h: Declare new functions.
5466 * opts.c (common_handle_option): Handle OPT__completion_.
5467 * selftest-run-tests.c (selftest::run_tests): Add
5468 opt_proposer_c_tests.
5469 * selftest.c (assert_str_startswith): New.
5470 * selftest.h (assert_str_startswith): Likewise.
5471 (opt_proposer_c_tests): New.
5472 (ASSERT_STR_STARTSWITH): Likewise.
5474 2018-06-28 Martin Liska <mliska@suse.cz>
5476 * Makefile.in: Add opt-suggestions.o.
5477 * gcc-main.c: Include opt-suggestions.h.
5478 * gcc.c (driver::driver): Likewise.
5479 (driver::~driver): Remove m_option_suggestions.
5480 (driver::build_option_suggestions): Moved to option_proposer.
5481 (driver::suggest_option): Likewise.
5482 (driver::handle_unrecognized_options): Use option_proposer.
5483 * gcc.h (class driver): Add new memver m_option_proposer.
5484 * opt-suggestions.c: New file.
5485 * opt-suggestions.h: New file.
5487 2018-06-28 Martin Liska <mliska@suse.cz>
5489 * vec.h (class auto_string_vec): New (moved from auto_argvec).
5490 (auto_string_vec::~auto_string_vec): Likewise.
5492 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
5494 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
5495 prevent_decl_creation_for_types fields up and add reset_location field.
5496 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
5497 statement if id->reset_location is true.
5498 (copy_edges_for_bb): Do not set goto_locus on the new edges if
5499 id->reset_location is true.
5500 (copy_phis_for_bb): Force input_location on the arguments if
5501 id->reset_location is true.
5502 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
5503 is set on the function to be inlined.
5505 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
5507 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
5509 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
5511 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
5512 registers for Pmode.
5513 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
5514 hard registers for the clobbered pseudo.
5516 2018-06-27 Paul Koning <ni1d@arrl.net>
5518 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
5519 mutually exclusive options.
5520 * config/pdp11/constraints.md (h): New constraint.
5521 (O): Update definition to match shift code generation.
5522 (D): New constraint.
5523 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
5525 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
5527 (output_jump): Change arguments.
5528 (pdp11_fixed_cc_regs): New function.
5529 (pdp11_cc_mode): Ditto.
5530 (pdp11_expand_shift): Ditto.
5531 (pdp11_assemble_shift): Ditto.
5532 (pdp11_small_shift): Ditto.
5533 (pdp11_branch_cost): Remove.
5534 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
5536 (pdp11_register_move_cost): Update for CC registers.
5537 (pdp11_rtx_costs): Add case for LSHIFTRT.
5538 (pdp11_output_jump): Add CCNZ mode conditional branches.
5539 (notice_update_cc_on_set): Remove.
5540 (pdp11_cc_mode): New function.
5541 (simple_memory_operand): Correct pre/post decrement case.
5542 (no_side_effect_operand): New function.
5543 (pdp11_regno_reg_class): Add CC_REGS class.
5544 (pdp11_fixed_cc_regs): New function.
5545 (pdp11_small_shift): New function.
5546 (pdp11_expand_shift): New function to expand shift insns.
5547 (pdp11_assemble_shift): New function to output shifts.
5548 (pdp11_branch_cost): Remove.
5549 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
5550 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
5551 (WCHAR_TYPE): Ditto.
5552 (PTRDIFF_TYPE): Ditto.
5553 (ADJUST_INSN_LENGTH): New macro.
5554 (FIXED_REGISTERS): Add CC registers.
5555 (CALL_USED_REGISTERS): Ditto.
5557 (REG_CLASS_NAMES): Ditto.
5558 (REG_CLASS_CONTENTS): Ditto.
5559 (SELECT_CC_MODE): Use new function.
5560 (TARGET_FLAGS_REGNUM): New macro.
5561 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
5562 (cc0_reg_rtx): Remove.
5563 (CC_STATUS_MDEP): Remove.
5564 (CC_STATUS_MDEFP_INIT): Remove.
5565 (CC_IN_FPU): Remove.
5566 (NOTICE_UPDATE_CC): Remove.
5567 (REGISTER_NAMES): Add CC registers.
5568 (BRANCH_COST): Change to constant 1.
5569 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
5571 * config/pdp11/pdp11.opt (mbcopy): Remove.
5572 (mbcopy-builtin): Remove.
5573 (mbranch-cheap): Remove.
5574 (mbranch-expensive): Remove.
5575 * config/pdp11/predicates.md (expand_shift_operand): Update to
5576 match shift code generation.
5577 (ccnz_operator): New predicate.
5578 * doc/invoke.texi (PDP-11 Options): Remove deleted options
5579 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
5580 Remove non-existent option -mabshi, -mno-abshi. Document mutually
5582 * doc/md.texi (PDP-11): Document new D and h constraints. Update
5583 description of O constraint.
5585 2018-06-27 Jeff Law <law@redhat.com>
5586 Austin Law <austinklaw@gmail.com>
5588 * config/v850/v850.md (addsi3_set_flags): New pattern.
5589 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
5590 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
5591 (zero_extendhisi2_v850_set_flags): Likewise.
5592 (zero_extendqisi2_v850_set_flags): Likewise.
5593 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
5594 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
5595 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
5597 * config/v850/v850-protos.h (notice_update_cc): Remove.
5598 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
5599 (v850_print_operand): Handle 'D' and "d".
5600 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
5601 Add handling of arithmetic/logical operations compared against zero.
5602 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
5603 Do not look at v850_compare_op, instead get mode from last argument.
5604 (v850_gen_compare): Remove
5605 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
5606 after reload for prologue insns.
5607 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
5609 (construct_save_jarl): Likewise.
5610 (TARGET_FLAGS_REGNUM): Define.
5611 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
5612 (NOTICE_UPDATE_CC): Remove.
5613 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
5614 than cc0. Conditionalize on reload_completed.
5615 (cmpsi_insn, setfcc_insn): Likewise.
5616 (tst1 splitter): Turn into define_and_split which sets the flags
5618 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
5619 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
5620 (cstoresf4, cstoredf4): Clobber the flags.
5621 (cmpsi, cmpsf, cmpdf): Remove expanders.
5622 (setf_insn): Remove pattern.
5623 (addsi3): Turn into define_and_split which clobbers the flags after
5624 reload and a suitable pattern (addsi3_clobber_flags) for use after
5626 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
5627 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
5628 (ashrsi3, ashrsi3_v850e2): Likewise.
5629 (bins): Clobber the flags.
5630 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
5631 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
5632 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
5633 (call_value_internal_short, call_value_internal_long): Likewise.
5634 (callt_save_interrupt, callt_return_interrupt): Likewise.
5635 (save_interrupt, return_interrupt): Likewise.
5636 (callt_save_all_interrupt, save_all_interrupt): Likewise.
5637 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
5638 (restore_all_interrupt, _restore_all_interrupt): Likewise.
5639 (All FP comparisons): Only allow after reload has completed.
5641 (divh, divhu): Tweak output template.
5642 (branch_z_normal, branch_z_invert): Remove
5643 (branch_nz_normal, branch_nz_invert): Likewise.
5644 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
5646 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
5647 * config/v850/v850.c (notice_update_cc): Remove.
5648 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
5649 (CC_NO_CARRY): Likewise.
5650 (NOTICE_UPDATE_CC): Define to nothing.
5651 * config/v850/v850.md: Remove block comment on cc0 handling
5652 Remove "cc" attribute from all patterns. Remove cc_status handling
5653 from all patterns. Minor formatting fixes.
5655 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5657 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
5658 (cortex-a76.cortex-a55): Likewise.
5659 * config/aarch64/aarch64-tune.md: Regenerate.
5660 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
5661 cortex-a76.cortex-a55.
5663 2018-06-27 Jeff Law <law@redhat.com>
5665 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
5666 (MULTILIB_DIRNAMES): Similarly.
5668 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
5670 * gimple.h (gimple_return_retbnd): Delete.
5671 (gimple_return_set_retbnd): Likewise.
5672 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
5673 gimple_return_set_retbnd.
5674 * gimple-pretty-print.c (dump_gimple_return): Remove call to
5675 gimple_return_retbnd and adjust.
5676 * tree-inline.h (struct copy_body_data): Remove retbnd field.
5677 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
5678 Explicitly return NULL in a couple more cases. Move assertion
5679 on debug statements and remove unreachable code.
5680 (reset_debug_binding): Do not test id->retbnd.
5681 (expand_call_inline): Do not set it.
5683 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
5685 * configure.ac: Add --disable-gcov option.
5686 * configure: Regenerate.
5687 * Makefile.in: Honour @enable_gcov@.
5688 * doc/install.texi: Document --disable-gcov.
5690 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5692 * config/arm/arm-cpus.in (cortex-a76): New entry.
5693 (cortex-a76.cortex-a55): Likewise.
5694 * config/arm/arm-tables.opt: Regenerate.
5695 * config/arm/arm-tune.md: Likewise.
5696 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
5697 * doc/invoke.texi (ARM Options): Document cortex-a76 and
5698 cortex-a76.cortex-a55.
5700 2018-06-27 Tamar Christina <tamar.christina@arm.com>
5703 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
5705 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
5707 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
5709 (EPILOGUE_USES): Likewise.
5711 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
5713 * tree-inline.c (remap_location): New function extracted from...
5714 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
5715 (copy_phis_for_bb): ...here. Call remap_location.
5716 (copy_cfg_body): Adjust call to copy_edges_for_bb.
5718 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
5720 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
5721 unaligned vsx for 16B memset.
5723 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
5726 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
5727 ieee128_float_type_node to long_double_type_node unless
5728 TARGET_LONG_DOUBLE_128 is set.
5730 2018-06-26 David Malcolm <dmalcolm@redhat.com>
5732 * cfgloop.c (get_loop_location): Convert return type from
5733 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
5734 by implicit construction from rtx_insn *, and using
5735 dump_user_location_t::from_function_decl for the fallback case.
5736 * cfgloop.h (get_loop_location): Convert return type from
5737 location_t to dump_user_location_t.
5738 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
5739 dump_printf_loc to pass in a dump_location_t rather than a
5740 location_t, via the gimple stmt.
5741 * coverage.c (get_coverage_counts): Update calls to
5742 dump_printf_loc to pass in dump_location_t rather than a
5744 * doc/optinfo.texi (Dump types): Convert example of
5745 dump_printf_loc from taking "locus" to taking "insn". Update
5746 description of the "_loc" calls to cover dump_location_t.
5747 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
5749 (dump_user_location_t::dump_user_location_t): New constructors,
5750 from gimple *stmt and rtx_insn *.
5751 (dump_user_location_t::from_function_decl): New function.
5752 (dump_loc): Make static.
5753 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
5754 const dump_location_t &.
5755 (dump_generic_expr_loc): Delete.
5756 (dump_printf_loc): Convert param "loc" from location_t to
5757 const dump_location_t &.
5758 (selftest::test_impl_location): New function.
5759 (selftest::dumpfile_c_tests): New function.
5760 * dumpfile.h: Include "profile-count.h".
5761 (class dump_user_location_t): New class.
5762 (struct dump_impl_location_t): New struct.
5763 (class dump_location_t): New class.
5764 (dump_printf_loc): Convert 2nd param from source_location to
5765 const dump_location_t &.
5766 (dump_generic_expr_loc): Delete.
5767 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
5768 const dump_location_t &.
5769 * gimple-fold.c (fold_gimple_assign): Update call to
5770 dump_printf_loc to pass in a dump_location_t rather than a
5771 location_t, via the gimple stmt.
5772 (gimple_fold_call): Likewise.
5773 * gimple-loop-interchange.cc
5774 (loop_cand::analyze_iloop_reduction_var): Update for change to
5775 check_reduction_path.
5776 (tree_loop_interchange::interchange): Update for change to
5778 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
5779 change in return-type of find_loop_location.
5780 (graphite_regenerate_ast_isl): Likewise.
5781 * graphite-optimize-isl.c (optimize_isl): Likewise.
5782 * graphite.c (graphite_transform_loops): Likewise.
5783 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
5784 pass in a dump_location_t rather than a location_t, via the
5786 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5787 * ipa.c (walk_polymorphic_call_targets): Likewise.
5788 * loop-unroll.c (report_unroll): Convert "locus" param from
5789 location_t to dump_location_t.
5790 (decide_unrolling): Update for change to get_loop_location's
5792 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
5793 location_t to dump_user_location_t.
5794 (grid_find_single_omp_among_assignments_1): Updates calls to
5795 dump_printf_loc to pass in a dump_location_t rather than a
5796 location_t, via the gimple stmt.
5797 (grid_parallel_clauses_gridifiable): Convert "tloc" from
5798 location_t to dump_location_t. Updates calls to dump_printf_loc
5799 to pass in a dump_location_t rather than a location_t, via the
5801 (grid_inner_loop_gridifiable_p): Likewise.
5802 (grid_dist_follows_simple_pattern): Likewise.
5803 (grid_gfor_follows_tiling_pattern): Likewise.
5804 (grid_target_follows_gridifiable_pattern): Likewise.
5805 (grid_attempt_target_gridification): Convert initialization
5806 of local "grid" from memset to zero-initialization; FIXME: does
5807 this require C++11? Update call to dump_printf_loc to pass in a
5808 optinfo_location rather than a location_t, via the gimple stmt.
5809 * profile.c (read_profile_edge_counts): Updates call to
5810 dump_printf_loc to pass in a dump_location_t rather than a
5812 (compute_branch_probabilities): Likewise.
5813 * selftest-run-tests.c (selftest::run_tests): Call
5815 * selftest.h (dumpfile_c_tests): New decl.
5816 * tree-loop-distribution.c (pass_loop_distribution::execute):
5817 Update for change in return type of find_loop_location.
5818 * tree-parloops.c (parallelize_loops): Likewise.
5819 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
5820 "locus" from location_t to dump_user_location_t.
5821 (canonicalize_loop_induction_variables): Likewise.
5822 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
5823 for change in return type of find_loop_location.
5824 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
5825 to dump_printf_loc to pass in a dump_location_t rather than a
5826 location_t, via the stmt.
5827 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
5829 * tree-vect-loop-manip.c (find_loop_location): Convert return
5830 type from source_location to dump_user_location_t.
5831 (vect_do_peeling): Update for above change.
5832 (vect_loop_versioning): Update for change in type of
5834 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
5835 from location_t to dump_user_location_t.
5836 (vect_estimate_min_profitable_iters): Update for change in type
5838 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
5839 location_t to dump_location_t.
5840 (vect_slp_bb): Update for change in type of vect_location.
5841 * tree-vectorizer.c (vect_location): Convert from source_location
5842 to dump_user_location_t.
5843 (try_vectorize_loop_1): Update for change in vect_location's type.
5844 (vectorize_loops): Likewise.
5845 (increase_alignment): Likewise.
5846 * tree-vectorizer.h (vect_location): Convert from source_location
5847 to dump_user_location_t.
5848 (find_loop_location): Convert return type from source_location to
5849 dump_user_location_t.
5850 (check_reduction_path): Convert 1st param from location_t to
5851 dump_user_location_t.
5852 * value-prof.c (check_counter): Update call to dump_printf_loc to
5853 pass in a dump_user_location_t rather than a location_t; update
5854 call to error_at for change in type of "locus".
5855 (check_ic_target): Update call to dump_printf_loc to
5856 pass in a dump_user_location_t rather than a location_t, via the
5859 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
5861 * config/s390/s390.h (enum processor_flags): Do not use
5862 default tune parameter when -march was specified.
5864 2018-06-26 Jakub Jelinek <jakub@redhat.com>
5867 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
5868 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
5871 2018-06-26 Richard Biener <rguenther@suse.de>
5873 PR tree-optimization/86287
5875 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
5876 (vect_analyze_loop): Initialize n_stmts.
5878 2018-06-26 Richard Biener <rguenther@suse.de>
5881 * fold-const.c (fold_convertible_p): Pointer extension
5884 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
5887 * dwarf2out.c (loc_list_has_views): Adjust comments.
5888 (dw_loc_list): Split single cross-partition range with
5891 2018-06-25 Jeff Law <law@redhat.com>
5893 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
5894 on -mbig-switch by default.
5896 * config/v850/predicates.md (const_float_1_operand): Fix match_code
5898 (const_float_0_operand): Remove unused predicate.
5899 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
5900 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
5901 (recipsf2): New expander. Original pattern now called
5903 (recipdf2, recipdf2_insn): Similarly.
5904 (rsqrtsf2, rsqrtsf2_insn): Similarly
5905 (rsqrtdf2, rsqrtdf2_insn): Similarly
5907 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
5909 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
5910 Simplify logic for FreeBSD (twice).
5912 2018-06-25 Martin Sebor <msebor@redhat.com>
5914 PR tree-optimization/86204
5915 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
5916 a strnlen result if it's less than the length of the string.
5918 2018-06-25 Martin Sebor <msebor@redhat.com>
5920 PR tree-optimization/85700
5921 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
5922 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
5923 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
5925 2018-06-25 Martin Sebor <msebor@redhat.com>
5927 * doc/extend.texi (Zero-length arrays): Update and clarify.
5929 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
5931 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
5932 added IEEE/IBM long double multilib support on PowerPC little
5933 endian Linux systems.
5934 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
5935 (MULTILIB_DEFAULTS): Likewise.
5936 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5938 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
5939 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
5940 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
5942 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
5945 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
5946 (REORDER_45): Likewise.
5948 2018-06-25 Jeff Law <law@redhat.com>
5950 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
5951 dividend to 32 bits. Adjust length.
5952 (udivmodhi4): Cleanup output template. Fix length.
5954 2018-06-25 Carl Love <cel@us.ibm.com>
5956 * config/rs6000/vsx.md: Change word selector to prefered location.
5958 2018-06-25 Richard Biener <rguenther@suse.de>
5960 PR tree-optimization/86304
5961 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
5962 epilogue-if-converted loops as well.
5964 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
5966 * lto-section-out.c (lto_begin_section): Do not print section
5967 name for noaddr and unnumbered dumps.
5969 2018-06-25 Richard Biener <rguenther@suse.de>
5971 * tree-vectorizer.h (struct vec_info_shared): New structure
5972 with parts split out from struct vec_info and loop_nest from
5973 struct _loop_vec_info.
5974 (struct vec_info): Adjust accordingly.
5975 (struct _loop_vec_info): Likewise.
5976 (LOOP_VINFO_LOOP_NEST): Adjust.
5977 (LOOP_VINFO_DATAREFS): Likewise.
5978 (LOOP_VINFO_DDRS): Likewise.
5979 (struct _bb_vec_info): Likewise.
5980 (BB_VINFO_DATAREFS): Likewise.
5981 (BB_VINFO_DDRS): Likewise.
5982 (struct _stmt_vec_info): Add dr_aux member.
5983 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
5984 (DR_MISALIGNMENT_UNINITIALIZED): New.
5985 (set_dr_misalignment): Adjust.
5986 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
5987 (vect_analyze_loop): Adjust prototype.
5988 (vect_analyze_loop_form): Likewise.
5989 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
5990 Compute dependences lazily.
5991 (vect_record_base_alignments): Use shared datarefs/ddrs.
5992 (vect_verify_datarefs_alignment): Likewise.
5993 (vect_analyze_data_refs_alignment): Likewise.
5994 (vect_analyze_data_ref_accesses): Likewise.
5995 (vect_analyze_data_refs): Likewise.
5996 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
5997 constructor parameter for shared part.
5998 (vect_analyze_loop_form): Pass in shared part and adjust.
5999 (vect_analyze_loop_2): Pass in storage for the number of
6000 stmts. Move loop nest finding to the caller. Compute
6002 (vect_analyze_loop): Pass in shared part.
6003 (vect_transform_loop): Verify shared datarefs are unchanged.
6004 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
6005 constructor parameter for shared part.
6006 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
6007 (vect_slp_bb): Verify shared datarefs are unchanged before
6009 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
6011 (new_stmt_vec_info): Initialize DR_AUX misalignment to
6012 DR_MISALIGNMENT_UNINITIALIZED.
6013 * tree-vectorizer.c (vec_info::vec_info): Add constructor
6014 parameter for shared part.
6015 (vec_info::~vec_info): Adjust.
6016 (vec_info_shared::vec_info_shared): New.
6017 (vec_info_shared::~vec_info_shared): Likewise.
6018 (vec_info_shared::save_datarefs): Likewise.
6019 (vec_info_shared::check_datarefs): Likewise.
6020 (try_vectorize_loop_1): Construct shared part live for analyses
6021 of a single loop for multiple vector sizes.
6022 * tree-parloops.c (gather_scalar_reductions): Adjust.
6024 2018-06-25 Richard Biener <rguenther@suse.de>
6026 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
6027 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
6028 (vect_analyze_data_refs): Remove similar code from here and
6029 simplify accordingly.
6031 2018-06-25 Richard Biener <rguenther@suse.de>
6033 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
6034 for reverse storage order accesses rather than asserting
6035 they cannot happen here.
6037 2018-06-25 Tom de Vries <tdevries@suse.de>
6040 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
6041 Use data16 instead of .byte for insn prefix.
6043 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
6046 * parser.c (make_char_string_pack): Pass this literal chars
6047 through cpp_interpret_string.
6048 (cp_parser_userdef_numeric_literal): Check the result of
6049 make_char_string_pack.
6051 2018-06-24 Maya Rashish <coypu@sdf.org>
6053 * ginclude/stddef.h: Simplify conditions around avoiding
6054 re-definition of __size_t.
6056 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6058 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
6059 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
6061 2018-06-22 Maya Rashish <coypu@sdf.org>
6063 * doc/invoke.texi (mno-fancy-math-387): Update for changes
6064 made to OpenBSD and NetBSD through the years.
6066 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
6068 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
6069 behavior of vec_pack (vector double, vector double) to match
6070 behavior of vec_float2 (vector double, vector double).
6072 2018-06-22 Olivier Hainque <hainque@adacore.com>
6074 * gimplify.c (gimplify_function_tree): Prevent creation
6075 of a trampoline for the address of the current function
6076 passed to entry/exit instrumentation hooks.
6078 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
6081 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
6084 2018-06-22 Martin Liska <mliska@suse.cz>
6086 PR tree-optimization/86263
6087 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6088 Bail out if is_enabled is false.
6089 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
6091 (jump_table_cluster::is_enabled): New function.
6093 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6095 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6096 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
6097 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
6098 (lto_input_ts_binfo_tree_pointers): Likewise.
6099 * tree-streamer-out.c (streamer_write_tree_bitfields,
6100 write_ts_binfo_tree_pointers): Likewise.
6101 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
6103 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
6105 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
6107 2018-06-22 Martin Liska <mliska@suse.cz>
6109 * symbol-summary.h (get): Make it pure and inline move
6110 functionality from ::get function.
6111 (get): Remove and inline into ::get and ::get_create.
6112 (get_create): Move code from ::get function.
6114 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6117 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
6118 -x assembler-with-cpp.
6120 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6122 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
6123 _FILE_OFFSET_BITS=64 for C++.
6125 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
6127 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
6128 conversion insn that shows up when pr85657-3.c is compiled using
6129 IEEE 128-bit long double.
6130 (neg<mode>2_internal): Use the correct mode to check whether the
6131 mode is IBM extended.
6132 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
6133 multiply and divide external functions from being created more
6136 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6138 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
6140 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
6141 the edge can be forwarded.
6142 (cfg_layout_merge_blocks): Likewise.
6144 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6146 * except.c (finish_eh_generation): Commit edge insertions only after
6147 the EH edges have been redirected from post-landing to landing pads.
6149 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6151 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
6152 create_tmp_var_for to create the FRAME decl.
6153 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
6155 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
6157 * tree-inline.c (copy_edges_for_bb): Minor tweak.
6158 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
6159 debug statement when resetting its value.
6160 (expand_call_inline): Copy the locus of the call onto the assignment
6161 of the return value, if any. Use local variable in more cases.
6163 2018-06-21 Martin Liska <mliska@suse.cz>
6165 * ipa-pure-const.c (propagate_nothrow): Use
6166 funct_state_summaries->get.
6167 (dump_malloc_lattice): Likewise.
6168 (propagate_malloc): Likewise.
6170 2018-06-21 Richard Biener <rguenther@suse.de>
6172 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
6173 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
6174 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
6176 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
6177 BLOCK_ABSTRACT_ORIGIN unconditionally.
6179 2018-06-21 David Malcolm <dmalcolm@redhat.com>
6181 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
6183 * ipa-reference.c (ipa_reference_c_finalize): Delete
6184 ipa_ref_opt_sum_summaries and set it to NULL.
6186 2018-06-21 Tom de Vries <tdevries@suse.de>
6188 PR tree-optimization/85859
6189 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
6190 test with comment from bb_no_side_effects_p.
6192 2018-06-21 Richard Biener <rguenther@suse.de>
6194 PR tree-optimization/86232
6195 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
6196 max for constant niter.
6198 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
6200 * config/aarch64/aarch64-simd.md
6201 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
6203 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
6205 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
6206 Make opernads of the unspec commutative.
6208 2018-06-21 Richard Biener <rguenther@suse.de>
6210 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
6211 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
6212 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
6213 (vect_analyze_data_ref_dependence): Re-order checks to deal with
6215 (vect_record_base_alignments): Do not record base alignment
6216 for gathers or scatters.
6217 (vect_compute_data_ref_alignment): Drop return value that is always
6218 true. Bail out early for gathers or scatters.
6219 (vect_enhance_data_refs_alignment): Bail out early for gathers
6221 (vect_find_same_alignment_drs): Likewise.
6222 (vect_analyze_data_refs_alignment): Remove dead code.
6223 (vect_slp_analyze_and_verify_node_alignment): Likewise.
6224 (vect_analyze_data_refs): For possible gathers or scatters do
6225 not create an alternate DR, just check their possible validity
6226 and mark them. Adjust DECL_NONALIASED handling to not rely
6228 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
6229 update inits of gathers or scatters.
6230 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
6231 Also copy gather/scatter flag to pattern vinfo.
6233 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
6235 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
6236 behavior of vec_packsu (vector unsigned long long, vector unsigned
6237 long long) to match behavior of vec_packs with same signature.
6239 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
6240 Thomas Schwinge <thomas@codesourcery.com>
6241 Cesar Philippidis <cesar@codesourcery.com>
6243 * gimplify.c (gimplify_scan_omp_clauses): Add support for
6244 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
6245 (gimplify_adjust_omp_clauses): Likewise.
6246 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
6247 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
6248 (gimplify_omp_target_update): Update handling of acc update and
6250 * omp-low.c (install_var_field): Remove unused parameter
6251 base_pointers_restrict.
6252 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
6253 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
6255 (omp_target_base_pointers_restrict_p): Delete.
6256 (scan_omp_target): Update call to scan_sharing_clauses.
6257 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
6259 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
6260 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
6261 (convert_local_omp_clauses): Likewise.
6262 * tree-pretty-print.c (dump_omp_clause): Likewise.
6263 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
6265 (omp_clause_code_name): Likewise.
6267 2018-06-20 Jakub Jelinek <jakub@redhat.com>
6270 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
6273 PR tree-optimization/86231
6274 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
6275 anti-range don't overwrite *vr0min before using it to compute *vr0max.
6277 2018-06-20 Tom de Vries <tdevries@suse.de>
6279 PR tree-optimization/86097
6280 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
6281 iv type if signedness of iv type is not the same as that of *nit.
6283 2018-06-20 Jakub Jelinek <jakub@redhat.com>
6285 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
6286 EDGE_EH edges, verify they are all EDGE_EH.
6288 2018-06-20 Maya Rashish <coypu@sdf.org>
6290 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
6292 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6294 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
6295 * config/aarch64/aarch64.c (xgene1_tunings): Add
6296 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
6297 (aarch64_mode_valid_for_sched_fusion_p):
6298 Allow 16-byte modes.
6299 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
6300 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
6302 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
6304 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
6305 * config/aarch64/iterators.md (VQ2): New mode iterator.
6307 2018-06-20 Martin Liska <mliska@suse.cz>
6309 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
6310 Change default ratio from 10 to 8.
6312 2018-06-20 Martin Liska <mliska@suse.cz>
6314 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
6316 (bit_test_cluster::find_bit_tests): Likewise.
6317 (switch_decision_tree::analyze_switch_statement): Find clusters.
6318 * tree-switch-conversion.h (struct jump_table_cluster): Document
6321 2018-06-20 Martin Liska <mliska@suse.cz>
6323 * tree-switch-conversion.c (switch_conversion::collect):
6324 Record m_uniq property.
6325 (switch_conversion::expand): Bail out for special conditions.
6326 (group_cluster::~group_cluster): New.
6327 (group_cluster::group_cluster): Likewise.
6328 (group_cluster::dump): Likewise.
6329 (jump_table_cluster::emit): New.
6330 (switch_decision_tree::fix_phi_operands_for_edges): New.
6331 (struct case_node): Remove struct.
6332 (jump_table_cluster::can_be_handled): New.
6333 (case_values_threshold): Moved to header.
6334 (reset_out_edges_aux): Likewise.
6335 (jump_table_cluster::is_beneficial): New.
6336 (bit_test_cluster::can_be_handled): Likewise.
6337 (add_case_node): Remove.
6338 (bit_test_cluster::is_beneficial): New.
6339 (case_bit_test::cmp): New.
6340 (bit_test_cluster::emit): New.
6341 (expand_switch_as_decision_tree_p): Remove.
6342 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
6343 (fix_phi_operands_for_edge): Likewise.
6344 (switch_decision_tree::analyze_switch_statement): New.
6345 (compute_cases_per_edge): Move ...
6346 (switch_decision_tree::compute_cases_per_edge): ... here.
6347 (try_switch_expansion): Likewise.
6348 (switch_decision_tree::try_switch_expansion): Likewise.
6349 (record_phi_operand_mapping): Likewise.
6350 (switch_decision_tree::record_phi_operand_mapping): Likewise.
6351 (emit_case_decision_tree): Likewise.
6352 (switch_decision_tree::emit): Likewise.
6353 (balance_case_nodes): Likewise.
6354 (switch_decision_tree::balance_case_nodes): Likewise.
6355 (dump_case_nodes): Likewise.
6356 (switch_decision_tree::dump_case_nodes): Likewise.
6357 (emit_jump): Likewise.
6358 (switch_decision_tree::emit_jump): Likewise.
6359 (emit_cmp_and_jump_insns): Likewise.
6360 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
6361 (emit_case_nodes): Likewise.
6362 (switch_decision_tree::emit_case_nodes): Likewise.
6363 (conditional_probability): Remove.
6364 * tree-switch-conversion.h (enum cluster_type): New.
6366 (struct cluster): Likewise.
6367 (cluster::cluster): Likewise.
6368 (struct simple_cluster): Likewise.
6369 (simple_cluster::simple_cluster): Likewise.
6370 (struct group_cluster): Likewise.
6371 (struct jump_table_cluster): Likewise.
6372 (struct bit_test_cluster): Likewise.
6373 (struct min_cluster_item): Likewise.
6374 (struct case_tree_node): Likewise.
6375 (case_tree_node::case_tree_node): Likewise.
6376 (jump_table_cluster::case_values_threshold): Likewise.
6377 (struct case_bit_test): Likewise.
6378 (struct switch_decision_tree): Likewise.
6379 (struct switch_conversion): Likewise.
6380 (switch_decision_tree::reset_out_edges_aux): Likewise.
6382 2018-06-20 Martin Liska <mliska@suse.cz>
6384 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
6385 (hoist_edge_and_branch_if_true): Likewise.
6386 (expand_switch_using_bit_tests_p): Likewise.
6387 (struct case_bit_test): Likewise.
6388 (case_bit_test_cmp): Likewise.
6389 (emit_case_bit_tests): Likewise.
6390 (switch_conversion::switch_conversion): New class.
6391 (struct switch_conv_info): Remove old struct.
6392 (collect_switch_conv_info): More to ...
6393 (switch_conversion::collect): ... this.
6394 (check_range): Likewise.
6395 (switch_conversion::check_range): Likewise.
6396 (check_all_empty_except_final): Likewise.
6397 (switch_conversion::check_all_empty_except_final): Likewise.
6398 (check_final_bb): Likewise.
6399 (switch_conversion::check_final_bb): Likewise.
6400 (create_temp_arrays): Likewise.
6401 (switch_conversion::create_temp_arrays): Likewise.
6402 (free_temp_arrays): Likewise.
6403 (gather_default_values): Likewise.
6404 (switch_conversion::gather_default_values): Likewise.
6405 (build_constructors): Likewise.
6406 (switch_conversion::build_constructors): Likewise.
6407 (constructor_contains_same_values_p): Likewise.
6408 (switch_conversion::contains_same_values_p): Likewise.
6409 (array_value_type): Likewise.
6410 (switch_conversion::array_value_type): Likewise.
6411 (build_one_array): Likewise.
6412 (switch_conversion::build_one_array): Likewise.
6413 (build_arrays): Likewise.
6414 (switch_conversion::build_arrays): Likewise.
6415 (gen_def_assigns): Likewise.
6416 (switch_conversion::gen_def_assigns): Likewise.
6417 (prune_bbs): Likewise.
6418 (switch_conversion::prune_bbs): Likewise.
6419 (fix_phi_nodes): Likewise.
6420 (switch_conversion::fix_phi_nodes): Likewise.
6421 (gen_inbound_check): Likewise.
6422 (switch_conversion::gen_inbound_check): Likewise.
6423 (process_switch): Use the newly created class.
6424 (switch_conversion::expand): New.
6425 (switch_conversion::~switch_conversion): New.
6426 * tree-switch-conversion.h: New file.
6428 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6430 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
6431 tree-vect-patterns.c.
6432 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
6433 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
6434 (vect_recog_sad_pattern): Likewise.
6435 (vect_recog_widen_sum_pattern): Likewise.
6436 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
6437 (vect_recog_widen_shift_pattern): Remove the type_in argument.
6438 (vect_recog_rotate_pattern): Likewise.
6439 (vect_recog_mult_pattern): Likewise.
6440 (vect_recog_vector_vector_shift_pattern): Likewise.
6441 (vect_recog_divmod_pattern): Likewise.
6442 (vect_recog_mixed_size_cond_pattern): Likewise.
6443 (vect_recog_bool_pattern): Likewise.
6444 (vect_recog_mask_conversion_pattern): Likewise.
6445 (vect_try_gather_scatter_pattern): Likewise.
6446 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
6447 (vect_recog_over_widening_pattern): Likewise.
6448 (vect_recog_gather_scatter_pattern): Likewise.
6449 (vect_recog_func_ptr): Move from tree-vectorizer.h
6450 (vect_vect_recog_func_ptrs): Move further down the file.
6451 (vect_recog_func): Likewise. Remove the third argument.
6452 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
6453 (vect_pattern_recog_1): Expect the pattern function to do any
6454 necessary target tests. Also expect it to provide a vector type.
6455 Remove the type_in handling.
6457 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6459 * tree-vect-patterns.c (vect_pattern_detected): New function.
6460 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
6461 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
6462 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
6463 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
6464 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
6465 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
6466 (vect_recog_mask_conversion_pattern)
6467 (vect_try_gather_scatter_pattern): Likewise.
6469 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6471 * tree-vect-patterns.c (vect_get_internal_def): New function.
6472 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
6473 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
6474 (search_type_for_mask_1): Use it.
6476 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6478 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6479 redundant WIDEN_SUM_EXPR handling.
6480 (vect_recog_sad_pattern): Likewise.
6482 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6484 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
6485 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
6486 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
6487 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
6488 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
6490 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6492 * tree-vect-stmts.c (vectorizable_call): Make sure that we
6493 use the stmt_vec_info of the original bb statement for the
6494 new zero assignment, even if the call is part of a pattern.
6496 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6498 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
6499 that the sequence is attached to the original statement rather
6500 than the pattern statement.
6501 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
6502 PATTERN_DEF_SEQ from the original statement rather than
6503 the main pattern statement.
6504 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
6505 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
6506 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
6508 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
6510 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
6511 definition statements before the early exit for statements that aren't
6513 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
6515 (vect_transform_loop): ...here. Process pattern definition
6516 statements without first checking whether the main pattern
6517 statement is live or relevant.
6519 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6521 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
6522 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
6524 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
6526 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
6527 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
6528 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
6529 (expand_block_compare): Change select_block_compare_mode call.
6530 (expand_strncmp_align_check): Use new functions, fix comment.
6531 (emit_final_str_compare_gpr): New function.
6532 (expand_strn_compare): Refactor and clean up code.
6533 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
6535 2018-06-19 Tony Reix <tony.reix@atos.com>
6536 Damien Bergamini <damien.bergamini@atos.com>
6537 David Edelsohn <dje.gcc@gmail.com>
6539 * collect2.c (static_obj): New variable.
6540 (static_libs): New variable.
6541 (is_in_list): Uncomment declaration.
6542 (main): Track AIX libraries linked statically.
6543 (is_in_list): Uncomment definition.
6544 (scan_prog_file): Don't add AIX shared libraries initializer
6545 to constructor list if linking statically.
6547 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
6549 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
6551 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
6553 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
6555 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
6558 2018-06-19 Martin Liska <mliska@suse.cz>
6560 * config/i386/i386.c (ix86_can_inline_p): Do not use
6561 ipa_fn_summaries::get_create.
6562 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
6564 (devirtualization_time_bonus): Likewise.
6565 (ipcp_propagate_stage): Likewise.
6566 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
6567 (edge_set_predicate): Likewise.
6568 (evaluate_conditions_for_known_args): Likewise.
6569 (evaluate_properties_for_edge): Likewise.
6570 (ipa_call_summary::reset): Tranform to ...
6571 (ipa_call_summary::~ipa_call_summary): ... this.
6572 (ipa_fn_summary::reset): Transform to ...
6573 (ipa_fn_summary::~ipa_fn_summary): ... this.
6574 (ipa_fn_summary_t::remove): Rename to ...
6575 (ipa_fn_summary_t::remove_callees): ... this.
6576 (ipa_fn_summary_t::duplicate): Use placement new
6577 instead of memory copy.
6578 (ipa_call_summary_t::duplicate): Likewise.
6579 (ipa_call_summary_t::remove): Remove.
6580 (dump_ipa_call_summary): Change get_create to get.
6581 (ipa_dump_fn_summary): Dump only when summary exists.
6582 (analyze_function_body): Use symbol_summary::get instead
6584 (compute_fn_summary): Likewise.
6585 (estimate_edge_devirt_benefit): Likewise.
6586 (estimate_edge_size_and_time): Likewise.
6587 (inline_update_callee_summaries): Likewise.
6588 (remap_edge_change_prob): Likewise.
6589 (remap_edge_summaries): Likewise.
6590 (ipa_merge_fn_summary_after_inlining): Likewise.
6591 (write_ipa_call_summary): Likewise.
6592 (ipa_fn_summary_write): Likewise.
6593 (ipa_free_fn_summary): Likewise.
6594 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
6595 (struct ipa_call_summary): Likewise.
6596 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
6598 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
6599 (estimate_size_after_inlining): Likewise.
6600 (estimate_growth): Likewise.
6601 (growth_likely_positive): Likewise.
6602 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
6603 (inline_call): Likewise.
6604 * ipa-inline.c (caller_growth_limits): Likewise.
6605 (can_inline_edge_p): Likewise.
6606 (can_inline_edge_by_limits_p): Likewise.
6607 (compute_uninlined_call_time): Likewise.
6608 (compute_inlined_call_time): Likewise.
6609 (want_inline_small_function_p): Likewise.
6610 (edge_badness): Likewise.
6611 (update_caller_keys): Likewise.
6612 (update_callee_keys): Likewise.
6613 (inline_small_functions): Likewise.
6614 (inline_to_all_callers_1): Likewise.
6615 (dump_overall_stats): Likewise.
6616 (early_inline_small_functions): Likewise.
6617 (early_inliner): Likewise.
6618 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
6619 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6620 * ipa-pure-const.c (malloc_candidate_p): Likewise.
6621 * ipa-split.c (execute_split_functions): Likewise.
6622 * symbol-summary.h: Likewise.
6623 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
6625 2018-06-19 Richard Biener <rguenther@suse.de>
6627 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
6628 (vectorize_loops): ... here. Fix dbgcnt handling.
6629 (try_vectorize_loop): Wrap try_vectorize_loop_1.
6631 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
6634 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
6635 ieee128 argument takes up only one (vector) register, not two (floating
6638 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
6640 * gimplify.c (gimplify_init_constructor): Really never clear for an
6641 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
6643 2018-06-19 Richard Biener <rguenther@suse.de>
6645 PR tree-optimization/86179
6646 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
6647 after failed recognition.
6649 2018-06-18 Martin Sebor <msebor@redhat.com>
6652 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
6653 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
6654 Handle integer subtraction.
6655 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
6656 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
6658 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6660 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
6661 param from rtx to rtx_insn *.
6662 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
6664 (frv_ifcvt_modify_insn): Likwise.
6665 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
6666 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
6667 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
6668 as_a <rtx_insn *> cast to local "unprotected_region" once
6669 it's been established that it's not NULL or pc_rtx.
6670 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
6671 param "sethi" from rtx to rtx_insn *.
6672 (nds32_group_float_insns): Likewise for param "insn".
6673 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
6675 (vax_output_int_subtract): Likewise.
6676 * config/vax/vax.c (vax_output_int_add): Likewise for param
6678 (vax_output_int_subtract): Likewise.
6679 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
6680 (emit_pattern_after): Likewise for param "after".
6681 (emit_insn_after): Likewise.
6682 (emit_jump_insn_after): Likewise.
6683 (emit_call_insn_after): Likewise.
6684 (emit_debug_insn_after): Likewise.
6685 (emit_pattern_before): Likewise for param "before".
6686 (emit_insn_before): Likewise.
6687 (emit_jump_insn_before): Likewise.
6688 * final.c (get_insn_template): Likewise for param "insn", removing
6690 * output.h (get_insn_template): Likewise for 2nd param.
6691 * rtl.h (emit_insn_before): Likewise.
6692 (emit_jump_insn_before): Likewise.
6693 (emit_debug_insn_before_noloc): Likewise.
6694 (emit_insn_after): Likewise.
6695 (emit_jump_insn_after): Likewise.
6696 (emit_call_insn_after): Likewise.
6697 (emit_debug_insn_after): Likewise.
6698 (set_insn_deleted): Likewise for param.
6700 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
6703 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
6704 floating point modes, so that IFmode is numerically greater than
6705 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
6706 to declare the ordering. This prevents IFmode from being
6707 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
6708 machine. Include rs6000-modes.h to share the fractional values
6709 between genmodes* and the rest of the compiler.
6713 * config/rs6000/rs6000-modes.h: New file.
6714 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
6715 meaning of rs6000_long_double_size so that 126..128 selects an
6716 appropriate 128-bit floating point type.
6717 (rs6000_option_override_internal): Likewise.
6718 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
6719 (TARGET_LONG_DOUBLE_128): Change the meaning of
6720 rs6000_long_double_size so that 126..128 selects an appropriate
6721 128-bit floating point type.
6722 (LONG_DOUBLE_TYPE_SIZE): Update comment.
6723 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
6724 source and destination to match the standard usage.
6725 (truncifkf2): Likewise.
6726 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
6727 ISA 2.07 to use an explicit clobber, instead of passing in a
6729 (copysign<mode>3_soft): Likewise.
6731 2018-06-18 David Malcolm <dmalcolm@redhat.com>
6733 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
6734 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
6735 (vect_slp_analyze_instance_dependence): Likewise.
6736 (vect_enhance_data_refs_alignment): Likewise.
6737 (vect_analyze_data_refs_alignment): Likewise.
6738 (vect_slp_analyze_and_verify_instance_alignment
6739 (vect_analyze_data_ref_accesses): Likewise.
6740 (vect_prune_runtime_alias_test_list): Likewise.
6741 (vect_analyze_data_refs): Likewise.
6742 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
6743 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6744 (vect_analyze_scalar_cycles_1): Likewise.
6745 (vect_get_loop_niters): Likewise.
6746 (vect_analyze_loop_form_1): Likewise.
6747 (vect_update_vf_for_slp): Likewise.
6748 (vect_analyze_loop_operations): Likewise.
6749 (vect_analyze_loop): Likewise.
6750 (vectorizable_induction): Likewise.
6751 (vect_transform_loop): Likewise.
6752 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
6753 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6754 (vect_make_slp_decision): Likewise.
6755 (vect_detect_hybrid_slp): Likewise.
6756 (vect_slp_analyze_operations): Likewise.
6757 (vect_slp_bb): Likewise.
6758 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
6759 (vectorizable_bswap): Likewise.
6760 (vectorizable_call): Likewise.
6761 (vectorizable_simd_clone_call): Likewise.
6762 (vectorizable_conversion): Likewise.
6763 (vectorizable_assignment): Likewise.
6764 (vectorizable_shift): Likewise.
6765 (vectorizable_operation): Likewise.
6766 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
6768 2018-06-18 Martin Sebor <msebor@redhat.com>
6770 PR tree-optimization/81384
6771 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
6772 * builtins.c (expand_builtin_strnlen): New function.
6773 (expand_builtin): Call it.
6774 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
6775 * builtins.def (BUILT_IN_STRNLEN): New.
6776 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
6777 Warn for bounds in excess of maximum object size.
6778 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
6779 single-value ranges. Handle strnlen.
6780 (handle_builtin_strlen): Handle strnlen.
6781 (strlen_check_and_optimize_stmt): Same.
6782 * doc/extend.texi (Other Builtins): Document strnlen.
6784 2018-06-18 Maya Rashish <coypu@sdf.org>
6786 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
6787 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
6788 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
6790 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
6792 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
6794 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6796 * tree.c (escaped_string::escape): Replace cast to char * by
6797 const_cast<char *> (unescaped).
6799 2018-06-18 Nick Clifton <nickc@redhat.com>
6802 * tree.c (escaped_string): New class. Converts an unescaped
6803 string into its escaped equivalent.
6804 (warn_deprecated_use): Use the new class to convert the
6805 deprecation message, if present.
6806 (test_escaped_strings): New self test.
6807 (test_c_tests): Add test_escaped_strings.
6808 * doc/extend.texi (deprecated): Add a note that the
6809 deprecation message is affected by the -fmessage-length
6810 option, and that control characters will be escaped.
6811 (#pragma GCC error): Document this pragma.
6812 (#pragma GCC warning): Likewise.
6813 * doc/invoke.texi (-fmessage-length): Document this option's
6814 effect on the #warning and #error preprocessor directives and
6815 the deprecated attribute.
6817 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
6819 * tree.c (decl_value_expr_lookup): Revert latest change.
6820 (decl_value_expr_insert): Likewise.
6822 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
6824 * gimplify.c (nonlocal_vlas): Delete.
6825 (nonlocal_vla_vars): Likewise.
6826 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
6828 (gimplify_body): Do not create and destroy nonlocal_vlas.
6829 * tree-nested.c: Include diagnostic.h.
6830 (use_pointer_in_frame): Tweak.
6831 (lookup_field_for_decl): Add assertion and declare the transformation.
6832 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
6833 internal error when the reference is in a wrong context. Do not
6834 create a debug decl by default.
6835 (note_nonlocal_block_vlas): Delete.
6836 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
6837 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
6838 create a debug decl by default.
6839 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
6840 call is in a wrong context.
6841 (fixup_vla_decls): New function.
6842 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
6843 debug variables were created.
6844 * tree.c (decl_value_expr_lookup): Add checking assertion.
6845 (decl_value_expr_insert): Likewise.
6847 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6850 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
6851 * tree-scalar-evolution.c (interpret_expr): Likewise.
6852 (expression_expensive_p): Likewise.
6853 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
6854 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
6855 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
6856 (ssa_defined_by_minus_one_stmt_p): New.
6858 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
6861 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
6862 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
6863 * dojump.c (do_jump): Likewise.
6864 * expr.c (expand_expr_real_2): Check operand type's sign.
6865 * fold-const.c (const_unop): Handle ABSU_EXPR.
6866 (fold_abs_const): Likewise.
6867 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
6868 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
6869 (strip_sign_op_1): Likesise.
6870 * match.pd: Add new pattern to generate ABSU_EXPR.
6871 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
6872 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6873 * tree-eh.c (operation_could_trap_helper_p): Likewise.
6874 * tree-inline.c (estimate_operator_cost): Likewise.
6875 * tree-pretty-print.c (dump_generic_node): Likewise.
6876 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
6877 * tree.def (ABSU_EXPR): New.
6879 2018-06-16 Jakub Jelinek <jakub@redhat.com>
6882 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
6883 documented as preserved for backward compatibility only.
6884 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
6886 PR rtl-optimization/86108
6887 * bb-reorder.c (create_forwarder_block): Renamed to ...
6888 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
6889 jump from new landing pad to the second part.
6890 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
6893 2018-06-15 Jakub Jelinek <jakub@redhat.com>
6896 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
6897 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
6898 Only call store_expr for halves if the mode is the same.
6901 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
6902 Fix up comment formatting.
6904 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
6906 * typed-splay-tree.h (typed_splay_tree::remove): New function.
6907 (typed_splay_tree::closure,
6908 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
6909 (typed_splay_tree::typed_splay_tree,
6910 typed_splay_tree::operator =): Declared private.
6911 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
6912 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
6913 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
6914 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
6915 typed_splay_tree::splay_tree_splay,
6916 typed_splay_tree::splay_tree_foreach_helper,
6917 typed_splay_tree::splay_tree_insert,
6918 typed_splay_tree::splay_tree_remove,
6919 typed_splay_tree::splay_tree_lookup,
6920 typed_splay_tree::splay_tree_predecessor,
6921 typed_splay_tree::splay_tree_successor,
6922 typed_splay_tree::splay_tree_min,
6923 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
6924 (typed_splay_tree::root, typed_splay_tree::comp,
6925 typed_splay_tree::delete_key,
6926 typed_splay_tree::delete_value): New data members.
6927 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
6928 typed_splay_tree::remove.
6930 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
6932 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
6933 -mginv and -mno-ginv to the assembler.
6934 * config/mips/mips.opt (-mcrc): New option.
6936 * doc/invoke.text (-mcrc): Document.
6939 2018-06-15 Nick Clifton <nickc@redhat.com>
6942 * tree.c (escaped_string): New class. Converts an unescaped
6943 string into its escaped equivalent.
6944 (warn_deprecated_use): Use the new class to convert the
6945 deprecation message, if present.
6946 (test_escaped_strings): New self test.
6947 (test_c_tests): Add test_escaped_strings.
6948 * doc/extend.texi (deprecated): Add a note that the
6949 deprecation message is affected by the -fmessage-length
6950 option, and that control characters will be escaped.
6951 (#pragma GCC error): Document this pragma.
6952 (#pragma GCC warning): Likewise.
6953 * doc/invoke.texi (-fmessage-length): Document this option's
6954 effect on the #warning and #error preprocessor directives and
6955 the deprecated attribute.
6957 2018-06-15 Richard Biener <rguenther@suse.de>
6959 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
6960 here, also noting vector size used.
6961 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
6962 size used in MSG_OPTIMIZED_LOCATIONS dump.
6963 (pass_slp_vectorize::execute): Adjust.
6965 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
6968 * config/arc/arc.c (arc_return_address_register): Fix
6971 2018-06-15 Richard Biener <rguenther@suse.de>
6974 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
6975 leave useless conversion stripping to force_gimple_operand_gsi.
6976 (gimplify_build2): Likewise.
6977 (gimplify_build1): Likewise.
6979 2018-06-15 Richard Biener <rguenther@suse.de>
6982 * tree-cfg.c (move_stmt_op): unshare invariant addresses
6983 before adjusting their block.
6985 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
6987 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
6988 multilibs for *-*-rtems*.
6989 * config/riscv/t-rtems: New file.
6991 2018-06-14 Jakub Jelinek <jakub@redhat.com>
6994 * match.pd ((A +- CST1) +- CST2): Punt if last resort
6995 unsigned_type_for returns NULL.
6998 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
6999 subregs of multi-word pseudos unless the float mode has word size.
7001 2018-06-14 Richard Biener <rguenther@suse.de>
7004 * tree-vect-generic.c (build_word_mode_vector_type): Remove
7005 duplicate and harmful type_hash_canon.
7006 * tree.c (type_hash_canon): Assert we didn't find ourselves.
7008 2018-06-14 Richard Biener <rguenther@suse.de>
7011 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
7014 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
7016 * config/rtems.h (STDINT_LONG32): Define.
7018 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
7019 Prachi Godbole <prachi.godbole@imgtec.com>
7021 * config/mips/mips-cpus.def: Define P6600.
7022 * config/mips/mips-tables.opt: Regenerate.
7023 * config/mips/mips.c (mips_ucbranch_type): New enum.
7024 (mips_rtx_cost_data): Add support for P6600.
7025 (mips_issue_rate): Likewise.
7026 (mips_multipass_dfa_lookahead): Likewise.
7027 (mips_avoid_hazard): Likewise.
7028 (mips_reorg_process_insns): Likewise.
7029 (mips_classify_branch_p6600): New function.
7030 * config/mips/mips.h (TUNE_P6600): New define.
7031 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
7032 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
7033 * config/mips/mips.md: Include p6600.md.
7034 (processor): Add p6600.
7035 * config/mips/p6600.md: New file.
7036 * doc/invoke.texi: Add p6600 to supported architectures.
7038 2018-06-13 Martin Sebor <msebor@redhat.com>
7040 PR tree-optimization/86114
7041 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
7043 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
7045 2018-06-13 Richard Biener <rguenther@suse.de>
7047 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
7048 Properly set vector type of the intermediate stmt.
7049 * tree-vect-stmts.c (vectorizable_operation): The destination
7050 var always has vectype_out type.
7052 2018-06-13 Jeff Law <law@redhat.com>
7054 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
7055 integer 0 for argument to print_rtl_with_bb.
7056 (rl78_reorg): Likewise.
7058 2018-06-13 David Malcolm <dmalcolm@redhat.com>
7060 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
7061 from rtx to rtx_insn *.
7062 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
7064 (add_sched_insns_for_speculation): Likewise for local "target",
7065 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
7066 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
7067 from rtx_insn ** to rtx_code_label **.
7068 (reorg_emit_nops): Likewise.
7069 (c6x_reorg): Likewise for local "call_labels".
7070 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
7072 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
7073 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
7074 the loops over LABEL_REFS.
7075 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
7077 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
7078 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
7079 (split_branches): Strengthen local "olabel" from rtx to
7080 rtx_insn *, adding a safe_as_a cast.
7081 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
7083 (add_insn_after): Likewise for first two params.
7084 (add_insn_before): Likewise.
7085 (remove_insn): Likewise for param.
7086 (emit_pattern_before_noloc): Likewise for second and third params.
7087 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
7088 (emit_call_insn_before_noloc): Likewise.
7089 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
7091 (emit_barrier_before): Likewise.
7092 (emit_label_before): Strengthen "label" param from "rtx" to
7093 "rtx_code_label *". Strengthen "before" param from "rtx" to
7095 (emit_insn_after_1): Strengthen "after" param from "rtx" to
7097 (emit_pattern_after_noloc): Likewise.
7098 (emit_insn_after_noloc): Likewise.
7099 (emit_jump_insn_after_noloc): Likewise.
7100 (emit_call_insn_after_noloc): Likewise.
7101 (emit_debug_insn_after_noloc): Likewise.
7102 (emit_barrier_after): Likewise.
7103 (emit_label_after): Likewise for both params.
7104 (emit_pattern_after_setloc): Likewise for "after" param. Convert
7105 "loc" param from "int" to "location_t".
7106 (emit_insn_after_setloc): Likewise.
7107 (emit_jump_insn_after_setloc): Likewise.
7108 (emit_call_insn_after_setloc): Likewise.
7109 (emit_debug_insn_after_setloc): Likewise.
7110 (emit_pattern_before_setloc): Likewise for "before" param. Convert
7111 "loc" param from "int" to "location_t".
7112 (emit_pattern_before): Convert NULL_RTX to NULL.
7113 (emit_insn_before_setloc): Convert "loc" param from "int" to
7115 (emit_jump_insn_before_setloc): Likewise.
7116 (emit_call_insn_before_setloc): Likewise.
7117 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
7118 rtx_insn *. Convert "loc" param from "int" to "location_t".
7119 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
7120 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
7121 Convert 3rd param from "int" to "location_t".
7122 (emit_barrier_before, emit_barrier_after, next_real_insn):
7123 Strengthen param from rtx to rtx_insn *.
7124 (emit_label_before): Strengthen 1st param from "rtx" to
7125 "rtx_code_label *". Strengthen 2nd param from "rtx" to
7127 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
7128 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
7129 Strengthen 2nd param from "rtx" to "rtx_insn *".
7130 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
7131 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
7132 Likewise. Convert 3rd param from "int" to "location_t".
7133 (emit_label_after): Strengthen 1st param from "rtx" to
7135 (next_real_insn, remove_insn): Strengthen param from "rtx" to
7137 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
7138 from "rtx" to "rtx_insn *".
7140 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
7142 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
7143 bodies streamed in with -Q.
7144 * dumpfile.c (dump_files): Add lto-stream-out dump file.
7145 * dumpfile.h (tree_dump_index): Add lto_stream_out.
7146 * gimple-streamer-out.c: Include gimple-pretty-print.h
7147 (output_bb): Dump stmts streamed.
7148 * lto-section-out.c: Include print-tree.h
7149 (lto_begin_section): Dump sections created.
7150 (lto_output_decl_index): Dump decl encoded.
7151 * lto-streamer-out.c: Include print-tree.h
7152 (create_output_block): Dump output block created.
7153 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
7154 (output_function): Dump function output.
7155 (output_constructor): Dump constructor streamed.
7156 (write_global_stream): Output indexes encoded.
7157 (produce_asm_for_decls): Dump streams encoded.
7158 * lto-streamer.c (streamer_dump_file): New global var.
7159 * lto-streamer.h (streamer_dump_file): Declare.
7160 * passes.c (ipa_write_summaries): Initialize streamer dump.
7161 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
7164 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
7167 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
7168 offsets for register save directives. Emit a second batch of save
7169 directives, if need be, when the function accesses prior frames.
7171 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7173 * config/arc/fpu.md (fmasf4): Force operand to register.
7174 (fnmasf4): Likewise.
7176 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7178 * config/arc/arc-protos.h (arc_pad_return): Remove.
7179 * config/arc/arc.c (machine_function): Remove force_short_suffix
7181 (arc_print_operand): Adjust printing of '&'.
7182 (arc_verify_short): Remove conditional printing of short suffix.
7183 (arc_final_prescan_insn): Remove reference to size_reason.
7184 (pad_return): New function.
7185 (arc_reorg): Call pad_return.
7186 (arc_pad_return): Remove.
7187 (arc_init_machine_status): Remove reference to force_short_suffix.
7188 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
7189 (attr length): When attribute iscompact is true force to 2
7190 regardless; in the case of maybe check if we want to force the
7191 instruction to have 4 bytes length.
7192 (nopv): Change it to generate 4 byte long nop as well.
7193 (blockage): New pattern.
7194 (simple_return): Remove call to arc_pad_return.
7195 (p_return_i): Likewise.
7197 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7199 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
7201 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7203 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
7206 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7208 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
7209 for ARC700 and ARCv2.
7211 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
7214 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
7215 operands[2] instead of operands[1].
7218 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
7220 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
7221 case, check whether the outer register overlaps an unallocatable
7222 register, not just whether it fits the required class.
7224 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
7226 * poly-int.h (can_div_trunc_p): Add new overload in which all values
7228 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
7229 (memrefs_conflict_p): Likewise.
7230 (init_alias_analysis): Likewise.
7231 * cfgexpand.c (expand_debug_expr): Likewise.
7232 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
7233 * cse.c (fold_rtx): Likewise.
7234 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
7235 * expr.c (emit_block_move_hints): Likewise.
7236 (clear_storage_hints, push_block, emit_push_insn): Likewise.
7237 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
7238 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
7239 (emit_group_store): Likewise.
7240 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
7241 to read the PRE/POST_MODIFY increment.
7242 * calls.c (store_one_arg): Use strip_offset.
7243 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
7245 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
7247 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
7248 (simplify_binary_operation_1): Extend CONST_INT handling to
7250 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
7251 than a HOST_WIDE_INT.
7252 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
7254 (adjust_mems, add_stores): Update accodingly.
7255 (vt_canonicalize_addr): Track polynomial offsets.
7256 (emit_note_insn_var_location): Likewise.
7257 (vt_add_function_parameter): Likewise.
7258 (vt_initialize): Likewise.
7260 2018-06-12 Jeff Law <law@redhat.com>
7262 * config.gcc (alpha*-*-freebsd*): Remove.
7263 * config/alpha/freebsd.h: Remove.
7265 2018-06-12 David Malcolm <dmalcolm@redhat.com>
7268 * spellcheck-tree.c (levenshtein_distance): Rename to...
7269 (get_edit_distance): ...this, and update for underlying renaming.
7270 * spellcheck-tree.h (levenshtein_distance): Rename to...
7271 (get_edit_distance): ...this.
7272 * spellcheck.c (levenshtein_distance): Rename to...
7273 (get_edit_distance): ...this. Convert from Levenshtein distance
7274 to Damerau-Levenshtein distance by supporting transpositions of
7275 adjacent characters. Rename "v1" to "v_next" and "v0" to
7277 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
7278 (selftest::test_edit_distance_unit_test_oneway): ...this, and
7279 update for underlying renaming.
7280 (selftest::levenshtein_distance_unit_test): Rename to...
7281 (selftest::test_get_edit_distance_unit): ...this, and update for
7282 underlying renaming.
7283 (selftest::test_find_closest_string): Add example from PR 69968
7284 where transposition helps
7285 (selftest::test_metric_conditions): Update for renaming.
7286 (selftest::test_metric_conditions): Likewise.
7287 (selftest::spellcheck_c_tests): Likewise.
7288 * spellcheck.h (levenshtein_distance): Rename both overloads to...
7289 (get_edit_distance): ...this.
7290 (best_match::consider): Update for renaming.
7292 2018-06-12 Martin Sebor <msebor@redhat.com>
7294 PR tree-optimization/85259
7295 * builtins.c (compute_objsize): Handle constant offsets.
7296 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
7297 true iff a warning has been issued.
7298 * gimple.h (gimple_nonartificial_location): New function.
7299 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
7300 gimple_nonartificial_location and handle -Wno-system-headers.
7301 (handle_builtin_stxncpy): Same.
7303 2018-06-12 Martin Sebor <msebor@redhat.com>
7306 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
7308 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
7310 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
7311 BUILTIN_VEC_XST entries for pointer to double and long long.
7313 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
7316 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
7318 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
7321 2018-06-12 Martin Liska <mliska@suse.cz>
7323 * doc/options.texi: Document IntegerRange.
7325 2018-06-12 Martin Liska <mliska@suse.cz>
7327 * config/i386/i386.opt: Make MPX-related options as Deprecated.
7328 * opt-functions.awk: Handle Deprecated flag.
7329 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
7331 (read_cmdline_option): Report warning for a deprecated option.
7332 * opts.h (struct cl_option): Add new field cl_deprecated.
7333 (CL_ERR_DEPRECATED): New.
7335 2018-06-12 Martin Liska <mliska@suse.cz>
7337 * doc/options.texi: Document Deprecated option flag.
7339 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7341 * config/arc/arc-arch.h (arc_extras): New enum.
7342 (arc_cpu_t):Add field extra.
7343 (arc_cpu_types): Consider the extras.
7344 * config/arc/arc-cpus.def: Add extras info.
7345 * config/arc/arc-opts.h (processor_type): Consider extra field.
7346 * config/arc/arc.c (arc_override_options): Handle extra field.
7348 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
7350 * config/arc/arc-arch.h: Update ARC_OPTX macro.
7351 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
7353 * config/arc/arc.c (arc_init): Update pic warning.
7354 (irq_range): Update irq range parsing warnings.
7355 (arc_override_options): Update various warning messages.
7356 (arc_handle_aux_attribute): Likewise.
7358 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
7360 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
7362 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7364 * doc/sourcebuild.texi: Document usage of line number 0 in verify
7365 compiler messages directives.
7367 2018-06-12 Matthew Fortune <mfortune@gmail.com>
7369 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
7370 * config/mips/mips-tables.opt: Regenerate.
7371 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
7373 * doc/invoke.texi: Document -march=i6500.
7375 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
7377 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
7378 (i6400_gpmul): Add cpu_unit.
7379 (i6400_gpdiv): Likewise.
7380 (i6400_msa_add_d): Update reservations.
7381 (i6400_msa_int_add) Likewise.
7382 (i6400_msa_short_logic3) Likewise.
7383 (i6400_msa_short_logic2) Likewise.
7384 (i6400_msa_short_logic) Likewise.
7385 (i6400_msa_move) Likewise.
7386 (i6400_msa_cmp) Likewise.
7387 (i6400_msa_short_float2) Likewise.
7388 (i6400_msa_div_d) Likewise.
7389 (i6400_msa_long_logic1) Likewise.
7390 (i6400_msa_long_logic2) Likewise.
7391 (i6400_msa_mult) Likewise.
7392 (i6400_msa_long_float2) Likewise.
7393 (i6400_msa_long_float4) Likewise.
7394 (i6400_msa_long_float5) Likewise.
7395 (i6400_msa_long_float8) Likewise.
7396 (i6400_fpu_fadd): Include frint type.
7397 (i6400_fpu_store): New define_insn_reservation.
7398 (i6400_fpu_load): Likewise.
7399 (i6400_fpu_move): Likewise.
7400 (i6400_fpu_fcmp): Likewise.
7401 (i6400_fpu_fmadd): Likewise.
7402 (i6400_int_mult): Include imul3nc type and update reservation.
7403 (i6400_int_div): Include idiv3 type and update reservation.
7404 (i6400_int_load): Update to check type not move_type.
7405 (i6400_int_store): Likewise.
7406 (i6400_int_prefetch): Set zero latency.
7408 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
7410 * gcc.c: Document new %@{...} sequence.
7411 (LINK_COMMAND_SPEC): Use it for the -L switches.
7412 (cpp_unique_options): Use it for the -I switches.
7413 (at_file_argbuf): New global variable.
7414 (in_at_file): Likewise.
7415 (alloc_args): Create at_file_argbuf.
7416 (clear_args): Truncate at_file_argbuf.
7417 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
7418 (open_at_file): New function.
7419 (close_at_file): Likewise.
7420 (create_at_file): Delete.
7421 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
7424 (validate_switches_from_spec): Deal with %@{...} sequence.
7425 (validate_switches): Likewise.
7426 (driver::finalize): Call clear_args.
7428 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
7430 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
7432 2018-06-11 Martin Sebor <msebor@redhat.com>
7434 * doc/invoke.texi (-Wall): List -Wc++17-compat.
7435 (Wno-class-memaccess): Add @opindex.
7436 (Wno-templates, Wno-multiple-inheritance): Same.
7437 (Wno-virtual-inheritance, Wno-namespaces): Same.
7438 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
7439 (Wno-format-overflow, Wno-format-truncation): Same.
7440 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
7441 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
7442 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
7443 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
7444 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
7445 (Wno-misspelled-isr): Same.
7447 2018-06-11 Martin Sebor <msebor@redhat.com>
7449 * PR tree-optimization/86083
7450 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
7452 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
7454 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
7456 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
7459 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
7460 on the correct operand.
7461 (*movdi_internal64): Ditto.
7463 2018-06-11 Martin Liska <mliska@suse.cz>
7465 PR tree-optimization/86089
7466 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
7468 2018-06-11 Julia Koval <julia.koval@intel.com>
7470 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
7471 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
7472 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
7474 2018-06-11 Olivier Hainque <hainque@adacore.com>
7476 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
7477 for Ada with strict dwarf2.
7479 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
7482 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
7485 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
7487 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
7489 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
7491 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
7494 2018-06-08 Martin Liska <mliska@suse.cz>
7496 * tree-cfg.h (debug_function): Fix argument type to match
7499 2018-06-08 Martin Liska <mliska@suse.cz>
7501 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
7502 Remove usage of MPX-related (and removed) fields.
7503 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
7505 2018-06-08 David Malcolm <dmalcolm@redhat.com>
7507 * cfg.c (debug): Use TDF_NONE rather than 0.
7508 * cfghooks.c (debug): Likewise.
7509 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
7510 (struct dump_option_value_info): Convert to...
7511 (struct kv_pair): ...this template type.
7512 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
7514 (optinfo_verbosity_options): Likewise.
7515 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
7517 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
7518 than int for "optgroup_flags" param.
7519 (dump_generic_expr_loc): Use dump_flags_t rather than int for
7521 (dump_dec): Likewise.
7522 (dump_finish): Use TDF_NONE rather than 0.
7523 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
7524 rather than int for "optgroup_flags" param. Use TDF_NONE rather
7525 than 0. Update for change to option_ptr.
7526 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
7527 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
7528 0. Update for changes to optinfo_verbosity_options and
7530 (opt_info_switch_p): Convert optgroup_flags from int to
7532 (dump_basic_block): Use dump_flags_t rather than int
7533 for "dump_kind" param.
7534 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
7535 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
7536 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
7537 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
7538 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
7539 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
7540 TDF_NONE): Convert from macros to...
7541 (enum dump_flag): ...this new enum.
7542 (dump_flags_t): Update to use enum.
7543 (operator|, operator&, operator~, operator|=, operator&=):
7544 Implement for dump_flags_t.
7545 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
7546 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
7547 Convert from macros to...
7548 (enum optgroup_flag): ...this new enum.
7549 (optgroup_flags_t): New typedef.
7550 (operator|, operator|=): Implement for optgroup_flags_t.
7551 (struct dump_file_info): Convert field "alt_flags" to
7552 dump_flags_t. Convert field "optgroup_flags" to
7554 (dump_basic_block): Use dump_flags_t rather than int for param.
7555 (dump_generic_expr_loc): Likewise.
7556 (dump_dec): Likewise.
7557 (dump_register): Convert param "optgroup_flags" to
7559 (opt_info_enable_passes): Likewise.
7560 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
7562 * gimple-pretty-print.c (debug): Likewise.
7563 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
7564 (merged_store_group::apply_stores): Likewise.
7565 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
7566 * gimple.c (verify_gimple_pp): Likewise.
7567 * graphite-poly.c (print_pbb_body): Likewise.
7568 * passes.c (pass_manager::register_one_dump_file): Convert
7569 local "optgroup_flags" to optgroup_flags_t.
7570 * print-tree.c (print_node): Use TDF_NONE rather than 0.
7572 (debug_body): Likewise.
7573 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
7574 to optgroup_flags_t.
7575 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
7577 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
7578 (convert_mult_to_fma): Likewise.
7579 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
7580 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
7581 * tree-vect-data-refs.c (dump_lower_bound): Convert param
7582 "dump_kind" to dump_flags_t.
7584 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7586 * config/rs6000/rs6000.c (min, max): Delete.
7588 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
7590 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
7591 -mabi=spe and -mabi=no-spe.
7593 2018-06-08 Martin Liska <mliska@suse.cz>
7595 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
7596 where we expect an existing summary.
7598 2018-06-08 Martin Liska <mliska@suse.cz>
7600 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
7601 * ipa-inline.h (estimate_edge_growth): Likewise.
7603 2018-06-08 Martin Liska <mliska@suse.cz>
7605 * cgraph.c (function_version_hasher::hash): Use
7606 cgraph_node::get_uid ().
7607 (function_version_hasher::equal):
7608 * cgraph.h (cgraph_node::get_uid): New method.
7609 * ipa-inline.c (update_caller_keys): Use
7610 cgraph_node::get_uid ().
7611 (update_callee_keys): Likewise.
7612 * ipa-utils.c (searchc): Likewise.
7613 (ipa_reduced_postorder): Likewise.
7614 * lto-cgraph.c (input_node): Likewise.
7615 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
7616 * symbol-summary.h (symtab_insertion): Likewise.
7617 (symtab_removal): Likewise.
7618 (symtab_duplication): Likewise.
7619 * tree-pretty-print.c (dump_function_header): Likewise.
7620 * tree-sra.c (convert_callers_for_node): Likewise.
7622 2018-06-08 Martin Liska <mliska@suse.cz>
7624 * cgraph.c (symbol_table::create_edge): Always assign a new
7626 (symbol_table::free_edge): Do not recycle numbers.
7627 * cgraph.h (cgraph_edge::get): New method.
7628 * symbol-summary.h (symtab_removal): Use it.
7629 (symtab_duplication): Likewise.
7630 (call_summary::hashable_uid): Remove.
7632 2018-06-08 Martin Liska <mliska@suse.cz>
7634 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
7635 (initialize_growth_caches): Remove.
7636 (free_growth_caches): Likewise.
7637 (do_estimate_edge_time): Use edge_growth_cache.
7638 (do_estimate_edge_size): Likewise.
7639 (do_estimate_edge_hints): Likewise.
7640 * ipa-inline.c (reset_edge_caches): Likewise.
7641 (recursive_inlining): Likewise.
7642 (inline_small_functions): Likewise.
7643 * ipa-inline.h (initialize_growth_caches): Remove.
7644 (estimate_edge_size): Likewise.
7645 (estimate_edge_time): Likewise.
7646 (estimate_edge_hints): Likewise.
7647 (reset_edge_growth_cache): Likewise.
7648 * symbol-summary.h (call_summary::remove): New method.
7650 2018-06-08 Martin Liska <mliska@suse.cz>
7652 * ipa-cp.c (class edge_clone_summary): New summary.
7653 (grow_edge_clone_vectors): Remove.
7654 (ipcp_edge_duplication_hook): Remove.
7655 (class edge_clone_summary_t): New call_summary class.
7656 (ipcp_edge_removal_hook): Remove.
7657 (edge_clone_summary_t::duplicate): New function.
7658 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
7659 (create_specialized_node): Likewise.
7660 (ipcp_driver): Initialize edge_clone_summaries and do not
7663 2018-06-08 Martin Liska <mliska@suse.cz>
7665 * symbol-summary.h (get): New function.
7666 (call_summary::m_initialize_when_cloning): New class member.
7668 2018-06-08 Martin Liska <mliska@suse.cz>
7670 * cgraph.c (cgraph_node::remove): Do not recycle uid.
7671 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
7672 (symbol_table::allocate_cgraph_symbol): Do not set uid.
7673 * passes.c (uid_hash_t): Record removed_nodes by their uids.
7674 (remove_cgraph_node_from_order): Use the removed_nodes set.
7675 (do_per_function_toporder): Likwise.
7676 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
7677 instead of summary_uid.
7678 (symtab_removal): Likewise.
7679 (symtab_duplication): Likewise.
7681 2018-06-08 Martin Liska <mliska@suse.cz>
7683 * ipa-cp.c (ipcp_store_bits_results): Use
7684 ipcp_transformation_sum.
7685 (ipcp_store_vr_results): Likewise.
7686 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
7688 (ipcp_transformation_initialize): ... this.
7689 (ipa_set_node_agg_value_chain):
7690 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
7691 (write_ipcp_transformation_info): Likewise.
7692 (read_ipcp_transformation_info): Likewise.
7693 (ipcp_update_bits): Likewise.
7694 (ipcp_update_vr): Likewise.
7695 (ipcp_transform_function): Likewise.
7696 * ipa-prop.h: Rename ipcp_transformation_summary to
7697 ipcp_transformation.
7698 (class ipcp_transformation_t): New function summary.
7699 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
7700 (ipa_get_agg_replacements_for_node): Likewise.
7702 2018-06-08 Martin Liska <mliska@suse.cz>
7704 * ipa-pure-const.c (struct funct_state_d): Do it class instead
7706 (class funct_state_summary_t): New function_summary class.
7707 (has_function_state): Remove.
7708 (get_function_state): Likewise.
7709 (set_function_state): Likewise.
7710 (add_new_function): Likewise.
7711 (funct_state_summary_t::insert): New function.
7712 (duplicate_node_data): Remove.
7713 (remove_node_data): Remove.
7714 (funct_state_summary_t::duplicate): New function.
7715 (register_hooks): Create new funct_state_summaries.
7716 (pure_const_generate_summary): Use it.
7717 (pure_const_write_summary): Likewise.
7718 (pure_const_read_summary): Likewise.
7719 (propagate_pure_const): Likewise.
7720 (propagate_nothrow): Likewise.
7721 (dump_malloc_lattice): Likewise.
7722 (propagate_malloc): Likewise.
7723 (execute): Do not register hooks, just remove summary
7725 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
7728 2018-06-08 Martin Liska <mliska@suse.cz>
7730 * ipa-reference.c (remove_node_data): Remove.
7731 (duplicate_node_data): Likewise.
7732 (class ipa_ref_var_info_summary_t): New class.
7733 (class ipa_ref_opt_summary_t): Likewise.
7734 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
7735 (get_reference_optimization_summary): Use
7736 ipa_ref_opt_sum_summaries.
7737 (set_reference_vars_info): Remove.
7738 (set_reference_optimization_summary): Likewise.
7739 (ipa_init): Create summaries.
7740 (init_function_info): Use function summary.
7741 (ipa_ref_opt_summary_t::duplicate): New function.
7742 (ipa_ref_opt_summary_t::remove): New function.
7743 (get_read_write_all_from_node): Fix GNU coding style.
7744 (propagate): Use function summary.
7745 (write_node_summary_p): Fix GNU coding style.
7746 (stream_out_bitmap): Likewise.
7747 (ipa_reference_read_optimization_summary): Use function summary.
7748 (ipa_reference_c_finalize): Do not release hooks.
7750 2018-06-08 Martin Liska <mliska@suse.cz>
7752 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
7753 (analyze_function_body): Extract multiple calls of get_create.
7754 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7755 * ipa-inline.c (recursive_inlining): Use ::get method.
7756 * ipa-inline.h (estimate_edge_growth): Likewise.
7758 2018-06-08 Martin Liska <mliska@suse.cz>
7760 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
7762 (hsa_function_summary::hsa_function_summary): Use the new enum
7764 (hsa_gpu_implementation_p): Use hsa_summaries::get.
7765 * hsa-gen.c (hsa_get_host_function): Likewise.
7766 (get_brig_function_name): Likewise.
7767 * ipa-hsa.c (process_hsa_functions): Likewise.
7768 (ipa_hsa_write_summary): Likewise.
7769 * symbol-summary.h (symtab_duplication): Use ::get function/
7770 (get): New function.
7772 2018-06-08 Martin Liska <mliska@suse.cz>
7774 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
7776 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
7777 (hsa_register_kernel): Likewise.
7778 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
7779 * hsa-gen.c (hsa_get_host_function): Likewise.
7780 (get_brig_function_name): Likewise.
7781 (generate_hsa): Likewise.
7782 (pass_gen_hsail::execute): Likewise.
7783 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
7784 (devirtualization_time_bonus): Likewise.
7785 (ipcp_propagate_stage): Likewise.
7786 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
7787 (edge_set_predicate): Likewise.
7788 (evaluate_conditions_for_known_args): Likewise.
7789 (evaluate_properties_for_edge): Likewise.
7790 (ipa_fn_summary::reset): Likewise.
7791 (ipa_fn_summary_t::duplicate): Likewise.
7792 (dump_ipa_call_summary): Likewise.
7793 (ipa_dump_fn_summary): Likewise.
7794 (analyze_function_body): Likewise.
7795 (compute_fn_summary): Likewise.
7796 (estimate_edge_devirt_benefit): Likewise.
7797 (estimate_edge_size_and_time): Likewise.
7798 (estimate_calls_size_and_time): Likewise.
7799 (estimate_node_size_and_time): Likewise.
7800 (inline_update_callee_summaries): Likewise.
7801 (remap_edge_change_prob): Likewise.
7802 (remap_edge_summaries): Likewise.
7803 (ipa_merge_fn_summary_after_inlining): Likewise.
7804 (ipa_update_overall_fn_summary): Likewise.
7805 (read_ipa_call_summary): Likewise.
7806 (inline_read_section): Likewise.
7807 (write_ipa_call_summary): Likewise.
7808 (ipa_fn_summary_write): Likewise.
7809 (ipa_free_fn_summary): Likewise.
7810 * ipa-hsa.c (process_hsa_functions): Likewise.
7811 (ipa_hsa_write_summary): Likewise.
7812 (ipa_hsa_read_section): Likewise.
7813 * ipa-icf.c (sem_function::merge): Likewise.
7814 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
7815 (do_estimate_edge_time): Likewise.
7816 (estimate_size_after_inlining): Likewise.
7817 (estimate_growth): Likewise.
7818 (growth_likely_positive): Likewise.
7819 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
7820 (inline_call): Likewise.
7821 * ipa-inline.c (caller_growth_limits): Likewise.
7822 (can_inline_edge_p): Likewise.
7823 (can_inline_edge_by_limits_p): Likewise.
7824 (compute_uninlined_call_time): Likewise.
7825 (compute_inlined_call_time): Likewise.
7826 (want_inline_small_function_p): Likewise.
7827 (edge_badness): Likewise.
7828 (update_caller_keys): Likewise.
7829 (update_callee_keys): Likewise.
7830 (recursive_inlining): Likewise.
7831 (inline_small_functions): Likewise.
7832 (inline_to_all_callers_1): Likewise.
7833 (dump_overall_stats): Likewise.
7834 (early_inline_small_functions): Likewise.
7835 (early_inliner): Likewise.
7836 * ipa-inline.h (estimate_edge_growth): Likewise.
7837 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
7838 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7839 * ipa-prop.h (IPA_NODE_REF): Likewise.
7840 (IPA_EDGE_REF): Likewise.
7841 * ipa-pure-const.c (malloc_candidate_p): Likewise.
7842 (propagate_malloc): Likewise.
7843 * ipa-split.c (execute_split_functions): Likewise.
7844 * symbol-summary.h: Rename get to get_create.
7846 (get_create): Likewise.
7847 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
7849 2018-06-08 Martin Liska <mliska@suse.cz>
7851 * symbol-summary.h (release): Move definition out of class
7853 (symtab_removal): Likewise.
7854 (symtab_duplication): Likewise.
7856 2018-06-08 Martin Liska <mliska@suse.cz>
7858 * symbol-summary.h (function_summary): Move constructor
7859 implementation out of class declaration.
7860 (release): Likewise.
7861 (symtab_insertion): Likewise.
7862 (symtab_removal): Likewise.
7863 (symtab_duplication): Likewise.
7866 2018-06-08 Martin Liska <mliska@suse.cz>
7868 * Makefile.in: Remove support for MPX (macros, related functions,
7869 fields in cgraph_node, ...).
7870 * builtin-types.def (BT_BND): Likewise.
7871 (BT_FN_BND_CONST_PTR): Likewise.
7872 (BT_FN_CONST_PTR_BND): Likewise.
7873 (BT_FN_VOID_PTR_BND): Likewise.
7874 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
7875 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
7876 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
7877 (expand_builtin_mempcpy_with_bounds): Likewise.
7878 (expand_builtin_memset_with_bounds): Likewise.
7879 (expand_builtin_memset_args): Likewise.
7880 (std_expand_builtin_va_start): Likewise.
7881 (expand_builtin): Likewise.
7882 (expand_builtin_with_bounds): Likewise.
7883 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
7884 (DEF_LIB_BUILTIN_CHKP): Likewise.
7885 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
7886 (DEF_CHKP_BUILTIN): Likewise.
7887 (BUILT_IN_MEMCPY): Likewise.
7888 (BUILT_IN_MEMMOVE): Likewise.
7889 (BUILT_IN_MEMPCPY): Likewise.
7890 (BUILT_IN_MEMSET): Likewise.
7891 (BUILT_IN_STPCPY): Likewise.
7892 (BUILT_IN_STRCAT): Likewise.
7893 (BUILT_IN_STRCHR): Likewise.
7894 (BUILT_IN_STRCPY): Likewise.
7895 (BUILT_IN_STRLEN): Likewise.
7896 (BUILT_IN_MEMCPY_CHK): Likewise.
7897 (BUILT_IN_MEMMOVE_CHK): Likewise.
7898 (BUILT_IN_MEMPCPY_CHK): Likewise.
7899 (BUILT_IN_MEMSET_CHK): Likewise.
7900 (BUILT_IN_STPCPY_CHK): Likewise.
7901 (BUILT_IN_STRCAT_CHK): Likewise.
7902 (BUILT_IN_STRCPY_CHK): Likewise.
7903 * calls.c (store_bounds): Likewise.
7904 (emit_call_1): Likewise.
7905 (special_function_p): Likewise.
7906 (maybe_warn_nonstring_arg): Likewise.
7907 (initialize_argument_information): Likewise.
7908 (finalize_must_preallocate): Likewise.
7909 (compute_argument_addresses): Likewise.
7910 (expand_call): Likewise.
7911 * cfgexpand.c (expand_call_stmt): Likewise.
7912 (expand_return): Likewise.
7913 (expand_gimple_stmt_1): Likewise.
7914 (pass_expand::execute): Likewise.
7915 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
7916 (cgraph_node::remove): Likewise.
7917 (cgraph_node::dump): Likewise.
7918 (cgraph_node::verify_node): Likewise.
7919 * cgraph.h (chkp_function_instrumented_p): Likewise.
7920 (symtab_node::get_alias_target): Likewise.
7921 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
7922 (cgraph_local_p): Likewise.
7923 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
7924 (cgraph_edge::rebuild_references): Likewise.
7925 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
7926 (walk_polymorphic_call_targets): Likewise.
7927 (cgraph_node::expand_thunk): Likewise.
7928 (symbol_table::output_weakrefs): Likewise.
7929 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
7930 (ix86_handle_option): Likewise.
7931 * config/i386/constraints.md: Likewise.
7932 * config/i386/i386-builtin-types.def (BND): Likewise.
7936 * config/i386/i386-builtin.def (BDESC_END): Likewise.
7937 (BDESC_FIRST): Likewise.
7939 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
7940 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
7941 * config/i386/i386.c (enum reg_class): Likewise.
7942 (ix86_target_string): Likewise.
7943 (ix86_option_override_internal): Likewise.
7944 (ix86_conditional_register_usage): Likewise.
7945 (ix86_valid_target_attribute_inner_p): Likewise.
7946 (ix86_set_indirect_branch_type): Likewise.
7947 (ix86_set_current_function): Likewise.
7948 (ix86_function_arg_regno_p): Likewise.
7949 (init_cumulative_args): Likewise.
7950 (ix86_function_arg_advance): Likewise.
7951 (ix86_function_arg): Likewise.
7952 (ix86_pass_by_reference): Likewise.
7953 (ix86_function_value_regno_p): Likewise.
7954 (ix86_function_value_1): Likewise.
7955 (ix86_function_value_bounds): Likewise.
7956 (ix86_return_in_memory): Likewise.
7957 (ix86_setup_incoming_vararg_bounds): Likewise.
7958 (ix86_va_start): Likewise.
7959 (indirect_thunk_need_prefix): Likewise.
7960 (print_reg): Likewise.
7961 (ix86_print_operand): Likewise.
7962 (ix86_expand_call): Likewise.
7963 (ix86_output_function_return): Likewise.
7964 (reg_encoded_number): Likewise.
7965 (BDESC_VERIFYS): Likewise.
7966 (ix86_init_mpx_builtins): Likewise.
7967 (ix86_init_builtins): Likewise.
7968 (ix86_emit_cmove): Likewise.
7969 (ix86_emit_move_max): Likewise.
7970 (ix86_expand_builtin): Likewise.
7971 (ix86_builtin_mpx_function): Likewise.
7972 (ix86_get_arg_address_for_bt): Likewise.
7973 (ix86_load_bounds): Likewise.
7974 (ix86_store_bounds): Likewise.
7975 (ix86_load_returned_bounds): Likewise.
7976 (ix86_store_returned_bounds): Likewise.
7977 (ix86_class_likely_spilled_p): Likewise.
7978 (ix86_hard_regno_mode_ok): Likewise.
7979 (x86_order_regs_for_local_alloc): Likewise.
7980 (ix86_mitigate_rop): Likewise.
7981 (ix86_bnd_prefixed_insn_p): Likewise.
7982 (ix86_mpx_bound_mode): Likewise.
7983 (ix86_make_bounds_constant): Likewise.
7984 (ix86_initialize_bounds): Likewise.
7985 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
7986 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
7987 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
7988 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
7989 (TARGET_CHKP_BOUND_MODE): Likewise.
7990 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
7991 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
7992 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
7993 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
7994 * config/i386/i386.h (TARGET_MPX): Likewise.
7995 (TARGET_MPX_P): Likewise.
7996 (VALID_BND_REG_MODE): Likewise.
7997 (FIRST_BND_REG): Likewise.
7998 (LAST_BND_REG): Likewise.
7999 (enum reg_class): Likewise.
8000 (BND_REG_P): Likewise.
8001 (BND_REGNO_P): Likewise.
8002 (BNDmode): Likewise.
8003 (ADJUST_INSN_LENGTH): Likewise.
8004 * config/i386/i386.md: Likewise.
8005 * config/i386/i386.opt: Likewise.
8006 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
8007 (defined): Likewise.
8008 (LINK_MPX): Likewise.
8009 (MPX_SPEC): Likewise.
8010 (LIBMPX_SPEC): Likewise.
8011 (LIBMPXWRAPPERS_SPEC): Likewise.
8012 (CHKP_SPEC): Likewise.
8013 * config/i386/predicates.md: Likewise.
8014 * dbxout.c (dbxout_type): Likewise.
8015 * doc/extend.texi: Likewise.
8016 * doc/invoke.texi: Likewise.
8017 * doc/md.texi: Likewise.
8018 * doc/tm.texi: Likewise.
8019 * doc/tm.texi.in: Likewise.
8020 * dwarf2out.c (is_base_type): Likewise.
8021 (gen_formal_types_die): Likewise.
8022 (gen_subprogram_die): Likewise.
8023 (gen_type_die_with_usage): Likewise.
8024 (gen_decl_die): Likewise.
8025 (dwarf2out_late_global_decl): Likewise.
8026 * expr.c (expand_assignment): Likewise.
8027 (emit_storent_insn): Likewise.
8028 (store_expr_with_bounds): Likewise.
8029 (store_expr): Likewise.
8030 (expand_expr_real_1): Likewise.
8031 * expr.h (store_expr_with_bounds): Likewise.
8032 * function.c (use_register_for_decl): Likewise.
8033 (struct bounds_parm_data): Likewise.
8034 (assign_parms_augmented_arg_list): Likewise.
8035 (assign_parm_find_entry_rtl): Likewise.
8036 (assign_parm_is_stack_parm): Likewise.
8037 (assign_parm_load_bounds): Likewise.
8038 (assign_bounds): Likewise.
8039 (assign_parms): Likewise.
8040 (expand_function_start): Likewise.
8041 * gcc.c (CHKP_SPEC): Likewise.
8042 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
8043 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
8044 (wrestrict_dom_walker::check_call): Likewise.
8045 * gimple.c (gimple_build_call_from_tree): Likewise.
8046 * gimple.h (enum gf_mask): Likewise.
8047 (gimple_call_with_bounds_p): Likewise.
8048 (gimple_call_set_with_bounds): Likewise.
8049 * gimplify.c (gimplify_init_constructor): Likewise.
8050 * ipa-cp.c (initialize_node_lattices): Likewise.
8051 (propagate_constants_across_call): Likewise.
8052 (find_more_scalar_values_for_callers_subset): Likewise.
8053 * ipa-hsa.c (process_hsa_functions): Likewise.
8054 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
8055 * ipa-icf.c (sem_function::merge): Likewise.
8056 * ipa-inline.c (early_inliner): Likewise.
8057 * ipa-pure-const.c (warn_function_noreturn): Likewise.
8058 (warn_function_cold): Likewise.
8059 (propagate_pure_const): Likewise.
8060 * ipa-ref.h (enum GTY): Likewise.
8061 * ipa-split.c (find_retbnd): Likewise.
8062 (consider_split): Likewise.
8063 (split_function): Likewise.
8064 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8065 * ipa.c (walk_polymorphic_call_targets): Likewise.
8066 (symbol_table::remove_unreachable_nodes): Likewise.
8067 (process_references): Likewise.
8068 (cgraph_build_static_cdtor_1): Likewise.
8069 * lto-cgraph.c (lto_output_node): Likewise.
8070 (output_refs): Likewise.
8071 (compute_ltrans_boundary): Likewise.
8072 (input_overwrite_node): Likewise.
8073 (input_node): Likewise.
8074 (input_cgraph_1): Likewise.
8075 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
8076 * passes.c (pass_manager::execute_early_local_passes): Likewise.
8077 (class pass_chkp_instrumentation_passes): Likewise.
8078 (make_pass_chkp_instrumentation_passes): Likewise.
8079 * passes.def: Likewise.
8080 * rtl.h (struct GTY): Likewise.
8081 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
8082 * stor-layout.c (layout_type): Likewise.
8083 * symtab.c: Likewise.
8084 * target.def: Likewise.
8085 * targhooks.c (default_chkp_bound_type): Likewise.
8086 (default_chkp_bound_mode): Likewise.
8087 (default_builtin_chkp_function): Likewise.
8088 (default_chkp_function_value_bounds): Likewise.
8089 (default_chkp_make_bounds_constant): Likewise.
8090 (default_chkp_initialize_bounds): Likewise.
8091 * targhooks.h (default_chkp_bound_type): Likewise.
8092 (default_chkp_bound_mode): Likewise.
8093 (default_builtin_chkp_function): Likewise.
8094 (default_chkp_function_value_bounds): Likewise.
8095 (default_chkp_make_bounds_constant): Likewise.
8096 (default_chkp_initialize_bounds): Likewise.
8097 * toplev.c (compile_file): Likewise.
8098 (process_options): Likewise.
8099 * tree-core.h (DEF_BUILTIN): Likewise.
8100 (DEF_BUILTIN_CHKP): Likewise.
8101 * tree-inline.c (declare_return_variable): Likewise.
8102 (remap_gimple_stmt): Likewise.
8103 (copy_bb): Likewise.
8104 (initialize_inlined_parameters): Likewise.
8105 (expand_call_inline): Likewise.
8106 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
8107 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
8108 (make_pass_ipa_chkp_produce_thunks): Likewise.
8109 (make_pass_chkp): Likewise.
8110 (make_pass_chkp_opt): Likewise.
8111 (make_pass_chkp_instrumentation_passes): Likewise.
8112 * tree-pretty-print.c (dump_generic_node): Likewise.
8113 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
8114 * tree-ssa-dce.c (propagate_necessity): Likewise.
8115 (eliminate_unnecessary_stmts): Likewise.
8116 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
8117 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
8118 * tree-ssa-sccvn.h: Likewise.
8119 * tree-ssa-strlen.c (get_string_length): Likewise.
8120 (valid_builtin_call): Likewise.
8121 (adjust_last_stmt): Likewise.
8122 (handle_builtin_strchr): Likewise.
8123 (handle_builtin_strcpy): Likewise.
8124 (handle_builtin_stxncpy): Likewise.
8125 (handle_builtin_memcpy): Likewise.
8126 (handle_builtin_strcat): Likewise.
8127 (strlen_check_and_optimize_stmt): Likewise.
8128 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
8129 * tree-streamer-in.c: Likewise.
8130 * tree-streamer.c (record_common_node): Likewise.
8131 * tree.c (tree_code_size): Likewise.
8132 (wide_int_to_tree_1): Likewise.
8133 (type_contains_placeholder_1): Likewise.
8134 (build_common_tree_nodes): Likewise.
8135 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
8136 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
8137 (POINTER_BOUNDS_P): Likewise.
8138 (BOUNDED_TYPE_P): Likewise.
8139 (BOUNDED_P): Likewise.
8140 (CALL_WITH_BOUNDS_P): Likewise.
8141 (pointer_bounds_type_node): Likewise.
8142 * value-prof.c (gimple_ic): Likewise.
8143 * var-tracking.c (vt_add_function_parameters): Likewise.
8144 * varasm.c (make_decl_rtl): Likewise.
8145 (assemble_start_function): Likewise.
8146 (output_constant): Likewise.
8147 (maybe_assemble_visibility): Likewise.
8148 * varpool.c (ctor_for_folding): Likewise.
8149 * chkp-builtins.def: Remove.
8150 * ipa-chkp.c: Remove.
8151 * ipa-chkp.h: Remove.
8152 * rtl-chkp.c: Remove.
8153 * rtl-chkp.h: Remove.
8154 * tree-chkp-opt.c: Remove.
8155 * tree-chkp.c: Remove.
8156 * tree-chkp.h: Remove.
8158 2018-06-07 Carl Love <cel@us.ibm.com>
8160 * config/rs6000/vsx.md (vextract_fp_from_shorth,
8161 vextract_fp_from_shortl): Add BE support.
8163 2018-06-07 Paul Koning <ni1d@arrl.net>
8165 * compare-elim.c (try_merge_compare): Don't merge compare if
8166 address contains a side effect.
8167 (try_eliminate_compare): Likewise.
8169 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
8171 * config.gcc: Support "tremont".
8172 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
8173 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8175 * config/i386/i386.c (m_TREMONT): Define.
8176 (processor_target_table): Add "tremont".
8177 (PTA_TREMONT): Define.
8178 (ix86_lea_outperforms): Add TARGET_TREMONT.
8179 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
8180 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
8181 and M_INTEL_GOLDMONT_PLUS.
8182 (fold_builtin_cpu): Add "tremont".
8183 (ix86_add_stmt_cost): Add TARGET_TREMONT.
8184 (ix86_option_override_internal): Add "tremont".
8185 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
8186 (processor_type): Add PROCESSOR_TREMONT.
8187 * config/i386/x86-tune.def: Add m_TREMONT.
8188 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
8190 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8192 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
8193 symbol defined for msp430i* devices to be lower case.
8195 2018-06-07 Richard Biener <rguenther@suse.de>
8197 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
8198 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
8199 Properly wrap signed arithmetic if overflow wraps.
8201 2018-06-07 Jakub Jelinek <jakub@redhat.com>
8203 PR tree-optimization/69615
8204 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
8205 of a cast from a same precision integral SSA_NAME in a bb dominated
8206 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
8207 cast to utype if rhs2 has already a compatible type.
8209 2018-06-07 Richard Biener <rguenther@suse.de>
8211 PR tree-optimization/85935
8212 * graphite-scop-detection.c (find_params_in_bb): Analyze
8213 condition operands with respect to the correct loop. Assert
8214 the analysis doesn't fail.
8216 2018-06-04 Carl Love <cel@us.ibm.com>
8218 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
8219 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
8220 as it is slightly cheaper.
8221 (first_match_or_eos_index_<mode>):
8222 Calculate index using natural element order.
8223 (first_match_index_<mode>):
8224 Calculate index using natural element order.
8225 (first_match_or_eos_index_<mode>):
8226 Calculate index using natural order.
8227 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
8228 for BE and LE modes.
8229 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
8230 P9V_BUILTIN_VCLZLSBB_V16QI.
8231 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
8234 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
8236 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
8237 indentation and line wrap for many prototypes. Add missing
8238 @smallexample directives around block of prototypes for vec_xl and
8241 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
8243 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
8244 track if we pass or return IEEE 128-bit floating point.
8245 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
8246 C++ mangling that is compatible with GCC 8.1.
8247 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
8248 (init_cumulative_args): Note if we pass or return IEEE 128-bit
8249 floating point types.
8250 (rs6000_function_arg_advance_1): Likewise.
8251 (rs6000_mangle_type): Optionally generate mangled names that match
8252 what GCC 8.1 generated for IEEE 128-bit floating point types.
8253 (rs6000_globalize_decl_name): If we have an external function that
8254 passes or returns IEEE 128-bit types, generate a weak reference
8255 from the mangled name used in GCC 8.1 to the current mangled
8257 (rs6000_init_builtins): Make __ibm128 use the long double type if
8258 long double is IBM extended double. Make __float128 use the long
8259 double type if long double is IEEE 128-bit.
8262 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
8263 macro for __ibm128 built-in functions.
8264 (PACK_IF): Add __ibm128 pack/unpack functions.
8265 (UNPACK_IF): Likewise.
8266 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
8267 enable long double built-in functions if long double is IEEE
8268 128-bit floating point.
8269 (rs6000_invalid_builtin): Update long double built-in function
8271 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
8272 functions, adjust the built-in function to use the long double
8273 built-in function if __ibm128 and long double are the same type.
8274 * doc/extend.texi (PowerPC builtins): Update documention for
8275 __builtin_{,un}pack_longdouble. Add documentation for
8276 __builtin_{,un}pack_ibm128.
8278 2018-06-06 Jim Wilson <jimw@sifive.com>
8280 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
8281 (struct machine_function): New field interrupt_mode.
8282 (riscv_handle_type_attribute): New function. Add forward declaration.
8283 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
8284 (riscv_expand_epilogue): Check interrupt_mode field.
8285 (riscv_set_current_function): Check interrupt attribute args and
8286 set interrupt_mode field.
8287 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
8288 (riscv_sret, riscv_uret): New.
8289 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
8290 new arguments to interrupt attribute.
8292 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
8295 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
8296 Don't handle -mcpu=power8 if -mpower9-vector is also used.
8298 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
8300 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
8301 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
8302 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
8303 several redundant entries.
8305 2018-06-06 David Malcolm <dmalcolm@redhat.com>
8307 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
8308 type from "rtx" to "rtx_insn *".
8309 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
8310 for local "call_insn", removing cast.
8311 (ix86_expand_call): Likewise, introducing a "call_insn" local.
8313 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
8315 PR tree-optimization/86066
8316 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
8317 for BIT_INSERT_EXPR stores.
8319 2018-06-06 Richard Biener <rguenther@suse.de>
8321 PR tree-optimization/86062
8322 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
8323 component refs ontop
8324 of to be offsetted base.
8326 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8328 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
8329 to be static and remove check on interrupt attribute name.
8331 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
8333 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8334 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
8336 2018-06-05 Steve Ellcey <sellcey@cavium.com>
8339 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
8341 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
8342 Remove second argument, change how error is called.
8343 (aarch64_layout_arg): Remove second argument from
8344 aarch64_err_no_fpadvsimd call.
8345 (aarch64_init_cumulative_args): Ditto.
8346 (aarch64_gimplify_va_arg_expr): Ditto.
8347 * config/aarch64/aarch64.md (mov<mode>): Ditto.
8349 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
8351 * config/i386/i386.md (simple_return_indirect_internal): New expander.
8352 (*simple_return_indirect_internal<mode>): Rename from
8353 simple_return_indirect_internal. Use W mode iterator.
8354 (rstorssp): New expander.
8355 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
8356 (clrssbsy): New expander.
8357 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
8359 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
8361 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
8363 (cmse_check_pointed_object): Likewise.
8365 2018-06-05 Martin Liska <mliska@suse.cz>
8367 PR gcov-profile/47618
8368 * doc/invoke.texi: Document how -fprofile-dir format
8371 2018-06-05 Richard Biener <rguenther@suse.de>
8373 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
8374 removal pretend DOM info isn't available so we do not update
8375 it and only remove edges, not dominated blocks. Actually free
8376 DOM info in case we removed something. Remove unreachable blocks.
8377 (mfb_keep_latches): Work with either DOM info or marked backedges.
8378 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
8379 first. Mark backedges if DOM info isn't available.
8380 (Re-)compute DOM info after cleanup_control_flow_pre.
8382 2018-06-05 Richard Biener <rguenther@suse.de>
8384 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
8385 (locus_discrim_hasher::hash): Adjust.
8386 (locus_discrim_hasher::equal): Likewise.
8387 (next_discriminator_for_locus): Work on line directly.
8388 (same_line_p): Pass in expanded locus1 as well.
8389 (assign_discriminators): Avoid redundant location expansions.
8391 2018-06-05 Richard Biener <rguenther@suse.de>
8393 PR tree-optimization/86046
8394 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
8395 if required after clearing TREE_ADDRESSABLE.
8397 2018-06-05 Richard Biener <rguenther@suse.de>
8399 PR tree-optimization/86047
8400 * tree-ssa-loop.c (for_each_index): Glob handling of all
8401 decls and constants and really handle all of them.
8403 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8406 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
8407 qualifier_void_pointer and qualifier_const_void_pointer.
8408 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
8409 (arm_init_builtins): Handle the above.
8410 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
8411 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
8412 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
8415 2018-06-05 Martin Liska <mliska@suse.cz>
8417 * auto-profile.c (read_autofdo_file): Do not use
8418 gcov_ctr_summary struct.
8419 (afdo_callsite_hot_enough_for_early_inline): Likewise.
8420 * coverage.c (struct counts_entry): Likewise.
8421 (read_counts_file): Read just single summary entry.
8422 (get_coverage_counts): Use gcov_summary struct.
8423 * coverage.h (get_coverage_counts): Likewise.
8424 * gcov-dump.c (dump_working_sets): Likewise.
8425 (tag_summary): Dump just single summary.
8426 * gcov-io.c (gcov_write_summary): Write just histogram
8428 (gcov_read_summary): Read just single summary.
8429 (compute_working_sets): Use gcov_summary struct.
8430 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
8431 of GCOV_COUNTERS_SUMMABLE.
8432 (GCOV_COUNTERS_SUMMABLE): Remove.
8433 (GCOV_FIRST_VALUE_COUNTER): Replace with
8434 GCOV_COUNTER_V_INTERVAL.
8435 (struct gcov_ctr_summary): Remove.
8436 (struct gcov_summary): Directly use fields of former
8438 (compute_working_sets): Use gcov_summary struct.
8439 * gcov.c (read_count_file): Do not use ctrs fields.
8440 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
8442 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
8444 * profile.c: Likewise.
8445 * profile.h: Likewise.
8447 2018-06-05 Martin Liska <mliska@suse.cz>
8449 PR gcov-profile/84846
8450 * gcov.c (output_lines): Print working directory only
8451 in intermediate format.
8453 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
8455 * config/s390/s390-builtin-types.def: Add void function type.
8456 * config/s390/s390-builtins.def: Use the function type for the
8459 2018-06-04 Jim Wilson <jimw@sifive.com>
8461 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
8463 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
8464 and maybe_eh_return. Change regno to unsigned int. Use new args to
8465 handle EH_RETURN_DATA_REGNO registers properly.
8466 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
8467 (riscv_expand_epilogue): Update comment. Change argument name and
8468 type. Update code to use new name and type. Pass new args to
8469 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
8471 * config/riscv/riscv.md (NORMAL_RETURN): New.
8472 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
8473 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
8474 (eh_return): Call gen_eh_return_internal and emit barrier.
8475 (eh_return_internal): Call riscv_expand_epilogue.
8477 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
8479 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
8480 bit_insertion field and declare can_be_merged_into method.
8481 (merged_store_group::can_be_merged_into): New method.
8482 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
8483 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
8484 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
8486 2018-06-04 Richard Biener <rguenther@suse.de>
8488 PR tree-optimization/85955
8489 * builtins.c (fold_builtin_sincos): Convert pointers to
8490 destination to appropriate type before dereferencing.
8492 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
8494 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
8496 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
8498 * expr.c (expand_expr_real_1): Force the operand into memory if
8499 its TYPE_MODE is BLKmode and if there is no integer mode for
8500 the number of bits being extracted.
8502 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8506 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
8507 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
8509 2018-06-04 Richard Biener <rguenther@suse.de>
8511 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
8512 (cleanup_tree_cfg_noloop): ... single caller. Do
8513 start_recording_case_labels later.
8515 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
8517 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
8518 to _IMMINTRIN_H_INCLUDED.
8519 * config/i386/pconfigintrin.h: Ditto.
8520 * config/i386/waitpkgintrin.h: Ditto.
8521 * config/i386/immintrin.h: Add includes for sgxintrin.h,
8522 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
8523 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
8524 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
8525 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
8526 waitpkgintrin.h and cldemoteintrin.h.
8528 2018-06-04 Richard Biener <rguenther@suse.de>
8530 PR tree-optimization/86038
8531 * tracer.c (find_best_successor): Check probability for
8532 being initialized, bail out if not.
8534 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
8537 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
8538 of bits to ignore when comparing architectures.
8540 2018-06-04 Jakub Jelinek <jakub@redhat.com>
8542 PR tree-optimization/69615
8543 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
8544 maximum or minimum of the type, try to merge it also as if
8545 range1 is + [-, x - 1] or + [x + 1, -].
8548 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
8550 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
8552 PR tree-optimization/86034
8553 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
8554 the unsigned bitfield type in a bit insertion sequence if it does not
8555 have a larger precision than the bitfield size.
8556 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
8558 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
8560 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
8562 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8564 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
8565 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
8566 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
8567 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
8569 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
8571 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
8572 Disable -fdelete-null-pointer-checks for ELF toolchain.
8574 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8575 Kito Cheng <kito.cheng@gmail.com>
8577 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
8578 (nds32le-*-*, nds32be-*-*): Integrate checking process.
8579 (nds32*-*-*): Add glibc and uclibc conditions.
8580 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
8581 (TARGET_EXCEPT_UNWIND_INFO): Define.
8582 * config/nds32/elf.h: New file.
8583 * config/nds32/linux.h: New file.
8584 * config/nds32/nds32-elf.opt: New file.
8585 * config/nds32/nds32-linux.opt: New file.
8586 * config/nds32/nds32-fp-as-gp.c
8587 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
8588 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
8590 (nds32_asm_file_end): Ditto.
8591 (nds32_print_operand): Ditto.
8592 (nds32_insert_attributes): Ditto.
8593 (nds32_init_libfuncs): New function.
8594 (TARGET_HAVE_TLS): Define.
8595 (TARGET_INIT_LIBFUNCS): Define.
8596 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
8598 (TARGET_ELF): Apply different mcmodel setting.
8599 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
8600 been migrated into elf.h and linux.h files.
8601 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
8602 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
8603 (mcmodel): The content has been migrated into nds32-elf.opt and
8604 nds32-linux.opt files.
8605 * config/nds32/t-elf: New file.
8606 * config/nds32/t-linux: New file.
8608 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
8609 Shiva Chen <shiva0217@gmail.com>
8611 * config/nds32/constants.md (unspec_volatile_element): Add
8612 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
8613 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
8615 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
8616 (make_pass_nds32_fp_as_gp): Declare.
8617 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
8619 (nds32_asm_function_end_prologue): Remove unused asm output.
8620 (nds32_asm_function_begin_epilogue): Remove unused asm output.
8621 (nds32_asm_file_start): Output necessary fp_as_gp information.
8622 (nds32_option_override): Adjust register usage.
8623 (nds32_expand_prologue): Consider fp_as_gp situation.
8624 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
8625 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
8628 (simple_return): Ditto.
8629 (omit_fp_begin): Output special directive for fp_as_gp.
8630 (omit_fp_end): Output special directive for fp_as_gp.
8631 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
8632 mforbid-fp-as-gp): New options.
8634 2018-06-01 Mark Wielaard <mark@klomp.org>
8636 * dwarf2out.c (dwarf2out_finish): Remove generation of
8637 DW_AT_loclists_base.
8639 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
8641 * gimple-ssa-store-merging.c: Include gimple-fold.h.
8642 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
8643 (struct merged_store_group): Add bit_insertion field.
8644 (dump_char_array): Use standard hexadecimal format.
8645 (merged_store_group::merged_store_group): Set bit_insertion to false.
8646 (merged_store_group::apply_stores): Use optimal buffer size. Deal
8647 with BIT_INSERT_EXPR stores. Move up code updating the mask and
8648 also print the mask in the dump file.
8649 (pass_store_merging::gate): Minor tweak.
8650 (imm_store_chain_info::coalesce_immediate): Fix wrong association
8651 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
8652 stores with INTEGER_CST stores.
8653 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
8654 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
8655 and use it throughout. Generate bit insertion sequences if need be.
8656 (pass_store_merging::process_store): Remove redundant condition.
8657 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
8659 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
8661 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
8662 the 128-bit floating point types. Fix function comment.
8664 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8666 * config/aarch64/aarch64-simd.md
8667 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
8669 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
8672 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8674 PR tree-optimization/85989
8675 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
8677 (backprop::intersect_uses): Check it when deciding whether this
8678 is a backedge reference.
8679 (backprop::process_block): Add each phi to m_visited_phis
8680 after visiting it, then clear it at the end.
8682 2018-06-01 Richard Biener <rguenther@suse.de>
8684 * tree-vectorizer.h (vect_dr_stmt): New function.
8685 (vect_get_load_cost): Adjust.
8686 (vect_get_store_cost): Likewise.
8687 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
8688 Use vect_dr_stmt instead of DR_SMTT.
8689 (vect_record_base_alignments): Likewise.
8690 (vect_calculate_target_alignment): Likewise.
8691 (vect_compute_data_ref_alignment): Likewise and make static.
8692 (vect_update_misalignment_for_peel): Likewise.
8693 (vect_verify_datarefs_alignment): Likewise.
8694 (vector_alignment_reachable_p): Likewise.
8695 (vect_get_data_access_cost): Likewise. Pass down
8696 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
8697 (vect_get_peeling_costs_all_drs): Likewise.
8698 (vect_peeling_hash_get_lowest_cost): Likewise.
8699 (vect_enhance_data_refs_alignment): Likewise.
8700 (vect_find_same_alignment_drs): Likewise.
8701 (vect_analyze_data_refs_alignment): Likewise.
8702 (vect_analyze_group_access_1): Likewise.
8703 (vect_analyze_group_access): Likewise.
8704 (vect_analyze_data_ref_access): Likewise.
8705 (vect_analyze_data_ref_accesses): Likewise.
8706 (vect_vfa_segment_size): Likewise.
8707 (vect_small_gap_p): Likewise.
8708 (vectorizable_with_step_bound_p): Likewise.
8709 (vect_prune_runtime_alias_test_list): Likewise.
8710 (vect_analyze_data_refs): Likewise.
8711 (vect_supportable_dr_alignment): Likewise.
8712 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
8713 (vect_gen_prolog_loop_niters): Likewise.
8714 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
8715 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
8717 (vect_recog_mask_conversion_pattern): Likewise.
8718 (vect_try_gather_scatter_pattern): Likewise.
8719 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
8720 to vect_get_store_cost.
8721 (vect_get_store_cost): Get stmt_info instead of DR.
8722 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
8723 (vect_get_load_cost): Get stmt_info instead of DR.
8725 2018-06-01 Richard Biener <rguenther@suse.de>
8728 * gimple-fold.c (var_decl_component_p): Also allow offsetted
8729 vars wrapped in MEM_REFs.
8731 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
8733 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
8734 Fix subreg tests so that we only return a choice between
8735 GENERAL_REGS and FP_REGS if the original classes included both.
8737 2018-06-01 Richard Biener <rguenther@suse.de>
8740 * tree-ssa-structalias.c (get_function_part_constraint):
8741 Handle NULL fi->decl.
8742 (find_func_aliases_for_call): Properly handle indirect
8743 fi from direct call.
8744 (find_func_clobbers): Likewise.
8745 (ipa_pta_execute): Likewise.
8746 (create_variable_info_for): For functions that are ifunc_resolver
8747 resolve to a varinfo that contains the result of the resolver call.
8748 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
8751 2018-05-31 Michael Collison <michael.collison@arm.com>
8753 * config/aarch64/aarch64.md:
8754 (*fix_to_zero_extenddfdi2): New pattern.
8755 * gcc.target/aarch64/fix_extend1.c: New testcase.
8757 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
8761 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
8762 and BUILT_IN_STRNCMP_EQ.
8763 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
8764 BUILT_IN_STRNCMP_EQ.
8765 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
8766 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8767 (gimple_fold_builtin): Likewise.
8768 * tree-ssa-strlen.c (compute_string_length): New function.
8769 (determine_min_obsize): New function.
8770 (handle_builtin_string_cmp): New function to handle calls to
8771 string compare functions.
8772 (strlen_optimize_stmt): Add handling to builtin string compare
8774 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
8775 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
8776 * tree.c (build_common_builtin_nodes): Add new defines of
8777 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
8779 2018-05-31 Jakub Jelinek <jakub@redhat.com>
8782 * bb-reorder.c (pass_partition_blocks::gate): Return false for
8783 functions with naked attribute.
8785 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
8787 * config/i386/sse.md (avx_vec_concat<mode>):
8788 Substitute concat_tg_mode mode attribute with xtg_mode.
8789 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
8790 (concat_tg_mode): Remove mode attribute.
8792 2018-05-31 Martin Sebor <msebor@redhat.com>
8795 * calls.c (alloc_max_size): Correct a logic error/typo.
8796 Treat excessive arguments as infinite. Warn for invalid arguments.
8797 * doc/invoke.texi (-Walloc-size-larger-than): Update.
8799 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
8802 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
8803 and movx for Haswell.
8805 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
8806 Cesar Philippidis <cesar@codesourcery.com>
8809 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
8810 when emitting error on private/firstprivate reductions.
8811 * omp-low.c (lower_omp_target): Avoid reference-type processing
8812 on pointers for firstprivate clause.
8814 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8816 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
8819 * config/aarch64/aarch64-simd.md
8820 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
8821 (aarch64_ld1_x3_<mode>): Likewise
8822 (aarch64_st1x2<VALLDIF:mode>): Likewise
8823 (aarch64_st1_x2_<mode>): Likewise
8824 (aarch64_st1x3<VALLDIF:mode>): Likewise
8825 (aarch64_st1_x3_<mode>): Likewise
8826 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
8827 (vld1_s8_x3): Likewise.
8828 (vld1_u16_x3): Likewise.
8829 (vld1_s16_x3): Likewise.
8830 (vld1_u32_x3): Likewise.
8831 (vld1_s32_x3): Likewise.
8832 (vld1_u64_x3): Likewise.
8833 (vld1_s64_x3): Likewise.
8834 (vld1_f16_x3): Likewise.
8835 (vld1_f32_x3): Likewise.
8836 (vld1_f64_x3): Likewise.
8837 (vld1_p8_x3): Likewise.
8838 (vld1_p16_x3): Likewise.
8839 (vld1_p64_x3): Likewise.
8840 (vld1q_u8_x3): Likewise.
8841 (vld1q_s8_x3): Likewise.
8842 (vld1q_u16_x3): Likewise.
8843 (vld1q_s16_x3): Likewise.
8844 (vld1q_u32_x3): Likewise.
8845 (vld1q_s32_x3): Likewise.
8846 (vld1q_u64_x3): Likewise.
8847 (vld1q_s64_x3): Likewise.
8848 (vld1q_f16_x3): Likewise.
8849 (vld1q_f32_x3): Likewise.
8850 (vld1q_f64_x3): Likewise.
8851 (vld1q_p8_x3): Likewise.
8852 (vld1q_p16_x3): Likewise.
8853 (vld1q_p64_x3): Likewise.
8854 (vst1_s64_x2): Likewise.
8855 (vst1_u64_x2): Likewise.
8856 (vst1_f64_x2): Likewise.
8857 (vst1_s8_x2): Likewise.
8858 (vst1_p8_x2): Likewise.
8859 (vst1_s16_x2): Likewise.
8860 (vst1_p16_x2): Likewise.
8861 (vst1_s32_x2): Likewise.
8862 (vst1_u8_x2): Likewise.
8863 (vst1_u16_x2): Likewise.
8864 (vst1_u32_x2): Likewise.
8865 (vst1_f16_x2): Likewise.
8866 (vst1_f32_x2): Likewise.
8867 (vst1_p64_x2): Likewise.
8868 (vst1q_s8_x2): Likewise.
8869 (vst1q_p8_x2): Likewise.
8870 (vst1q_s16_x2): Likewise.
8871 (vst1q_p16_x2): Likewise.
8872 (vst1q_s32_x2): Likewise.
8873 (vst1q_s64_x2): Likewise.
8874 (vst1q_u8_x2): Likewise.
8875 (vst1q_u16_x2): Likewise.
8876 (vst1q_u32_x2): Likewise.
8877 (vst1q_u64_x2): Likewise.
8878 (vst1q_f16_x2): Likewise.
8879 (vst1q_f32_x2): Likewise.
8880 (vst1q_f64_x2): Likewise.
8881 (vst1q_p64_x2): Likewise.
8882 (vst1_s64_x3): Likewise.
8883 (vst1_u64_x3): Likewise.
8884 (vst1_f64_x3): Likewise.
8885 (vst1_s8_x3): Likewise.
8886 (vst1_p8_x3): Likewise.
8887 (vst1_s16_x3): Likewise.
8888 (vst1_p16_x3): Likewise.
8889 (vst1_s32_x3): Likewise.
8890 (vst1_u8_x3): Likewise.
8891 (vst1_u16_x3): Likewise.
8892 (vst1_u32_x3): Likewise.
8893 (vst1_f16_x3): Likewise.
8894 (vst1_f32_x3): Likewise.
8895 (vst1_p64_x3): Likewise.
8896 (vst1q_s8_x3): Likewise.
8897 (vst1q_p8_x3): Likewise.
8898 (vst1q_s16_x3): Likewise.
8899 (vst1q_p16_x3): Likewise.
8900 (vst1q_s32_x3): Likewise.
8901 (vst1q_s64_x3): Likewise.
8902 (vst1q_u8_x3): Likewise.
8903 (vst1q_u16_x3): Likewise.
8904 (vst1q_u32_x3): Likewise.
8905 (vst1q_u64_x3): Likewise.
8906 (vst1q_f16_x3): Likewise.
8907 (vst1q_f32_x3): Likewise.
8908 (vst1q_f64_x3): Likewise.
8909 (vst1q_p64_x3): Likewise.
8911 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
8913 * config/msp430/msp430.c (msp430_output_labelref): Prepend
8914 user_label_prefix to name.
8916 * tree-core.h: Update comment about the format of NAME string
8917 passed to handler in attribute_spec.
8919 * config/msp430/msp430.md: Remove erroneous subreg expression from
8920 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
8921 zero_extend{q,h}isi2.
8923 2018-05-30 Borislav Petkov <bp@suse.de>
8925 * doc/extend.texi: Document some architecture specific
8926 constraints and sort entries.
8928 2018-05-30 Martin Sebor <msebor@redhat.com>
8931 * builtins.c (expand_builtin_stpcpy_1): New function.
8932 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
8933 only if the former succeeds.
8935 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
8937 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
8940 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8942 * doc/invoke.texi (-flinker-output): Document
8944 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8946 * passes.c (ipa_write_summaries): Only modify statements if body
8948 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
8949 incrementally linking.
8950 (ipa_passes): Likewise.
8951 * lto-cgraph.c (lto_output_node): When incrementally linking do not
8952 pass down resolution info.
8953 * common.opt (flag_incremental_link): Update info.
8954 * gcc.c (plugin specs): Turn flinker-output=* to
8955 -plugin-opt=-linker-output-known
8956 * toplev.c (compile_file): Also cut compilation when doing incremental
8958 * flag-types. (enum lto_partition_model): Add
8959 LTO_LINKER_OUTPUT_NOLTOREL.
8960 (invoke.texi): Add -flinker-output docs.
8961 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
8962 link same way as WPA; do not stream in dead initializers.
8964 * dwarf2out.c (dwarf2out_die_ref_for_decl,
8965 darf2out_register_external_decl): Support incremental link.
8967 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8969 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
8971 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
8973 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
8974 it down to simple_object_copy_lto_debug_sections.
8975 (run_gcc): Determine incremental LTO link time and configure
8976 lto1 into non-wpa mode, disable renaming of debug sections.
8978 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
8980 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
8981 descriptions of various incorrectly documented functions.
8983 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
8986 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
8987 address check not strict.
8989 2018-05-30 Richard Biener <rguenther@suse.de>
8991 PR tree-optimization/85964
8992 * tracer.c (better_p): Drop initialized count check, we only
8993 call the function with initialized counts now.
8994 (find_best_successor): Do find a best edge if one
8995 has uninitialized count.
8996 (find_best_predecessor): Likewise. Do BB frequency check only
8997 if count is initialized.
8999 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
9001 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
9002 (aarch64_ldrstr_offset_compare): New.
9003 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
9004 load/store orderings.
9005 (aarch64_gen_adjusted_ldpstp): Likewise.
9007 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
9009 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
9010 Check for subset of GENERAL_REGS and FP_REGS.
9011 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
9014 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
9016 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
9017 and wi::to_poly_offset. Add the current offset and then check
9018 whether the sum fits, rather than using an unchecked addition of
9019 a checked term. Check for a shwi rather than a uhwi.
9020 * expr.c (get_bit_range): Use tree_to_poly_uint64.
9021 (store_constructor): Use poly_int_tree_p.
9022 (expand_expr_real_1): Likewise.
9023 * function.c (assign_temp): Likewise.
9024 * fold-const.c (const_binop): Use poly_int_tree_p and
9026 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
9028 * ipa-icf-gimple.c (func_checker::compare_operand): Use
9029 to_poly_offset for MEM offsets.
9030 * ipa-icf.c (sem_variable::equals): Likewise.
9031 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
9032 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
9033 wi::to_poly_offset for BIT_FIELD_REF offsets.
9034 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
9036 * var-tracking.c (emit_note_insn_var_location): Use
9037 tree_to_poly_uint64.
9039 2018-05-29 Jim Wilson <jimw@sifive.com>
9041 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
9043 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
9046 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
9047 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
9049 (sse4_1_round<mode>2): Use nonimmediate_operand
9050 for operand 1 predicate.
9052 2018-05-29 Martin Sebor <msebor@redhat.com>
9053 Richard Biener <rguenther@suse.de>
9056 * calls.c (get_size_range): Call determine_value_range instead
9057 of get_value_range..
9058 * tree-vrp.h (determine_value_range): Declared new function.
9059 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
9061 2018-05-29 Richard Biener <rguenther@suse.de>
9063 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
9064 sure to use non-pattern stmts for get_earlier_stmt arguments.
9065 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
9066 called on pattern stmts.
9067 (get_later_stmt): Likewise.
9069 2018-05-29 Martin Liska <mliska@suse.cz>
9071 PR gcov-profile/85759
9072 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
9075 2018-05-29 Jakub Jelinek <jakub@redhat.com>
9077 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
9079 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
9083 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
9084 VEC_PACK_FLOAT_EXPR): New tree codes.
9085 * tree-pretty-print.c (op_code_prio): Handle
9086 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
9087 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
9088 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
9089 * tree-inline.c (estimate_operator_cost): Likewise.
9090 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
9091 * fold-const.c (const_binop): Likewise.
9092 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
9093 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
9094 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9095 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
9096 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
9097 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
9098 * expr.c (expand_expr_real_2): Likewise.
9099 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
9100 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
9101 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
9103 * optabs.c (expand_widen_pattern_expr): For
9104 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
9105 sign from result type rather than operand's type.
9106 (expand_binop_directly): For vec_packu_float_optab and
9107 vec_packs_float_optab allow result type to be different from operand's
9109 * optabs-tree.c (optab_for_tree_code): Handle
9110 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
9111 VEC_PACK_FLOAT_EXPR. Formatting fixes.
9112 * tree-vect-generic.c (expand_vector_operations_1): Handle
9113 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
9114 VEC_PACK_FLOAT_EXPR.
9115 * tree-vect-stmts.c (supportable_widening_operation): Handle
9117 (supportable_narrowing_operation): Handle FLOAT_EXPR.
9118 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
9119 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
9120 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
9121 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
9123 (vec_pack<floatprefix>_float_<mode>): New expander.
9124 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
9126 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
9127 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
9128 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
9129 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
9130 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
9132 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
9133 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
9134 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
9135 VEC_PACK_FLOAT_EXPR): Document.
9137 2018-05-29 Richard Biener <rguenther@suse.de>
9139 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
9141 (stmt_vec_info_vec): Make pointer.
9142 (init_stmt_vec_info_vec): Remove.
9143 (free_stmt_vec_info_vec): Likewise.
9144 (set_stmt_vec_info_vec): New function.
9145 (free_stmt_vec_infos): Likewise.
9146 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
9147 (set_vinfo_for_stmt): Likewise.
9148 (get_earlier_stmt): Likewise.
9149 (get_later_stmt): Likewise.
9150 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
9151 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
9152 (vec_info::~vec_info): Free stmt_vec_infos.
9153 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
9154 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
9155 (pass_slp_vectorize::execute): Likewise.
9156 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
9157 (free_stmt_vec_info_vec): Likewise.
9158 (set_stmt_vec_info_vec): New function.
9159 (free_stmt_vec_infos): Likewise.
9160 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
9161 the global stmt_vec_info_vec.
9162 * tree-parloops.c (gather_scalar_reductions): Use
9163 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
9166 2018-05-29 Richard Biener <rguenther@suse.de>
9168 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
9170 2018-05-29 Martin Liska <mliska@suse.cz>
9171 David Malcolm <dmalcolm@redhat.com>
9173 * vec.c (test_reverse): New.
9174 (vec_c_tests): Add new test.
9175 * vec.h (vl_ptr>::reverse): New function.
9177 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
9179 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
9181 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
9184 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
9186 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
9188 2018-05-28 Richard Biener <rguenther@suse.de>
9190 PR tree-optimization/85933
9191 * tree-vect-data-refs.c (vect_record_base_alignments): Only
9192 look at stmts marked as vectorizable.
9194 2018-05-28 Richard Biener <rguenther@suse.de>
9196 PR tree-optimization/85934
9197 * tree-vect-generic.c (expand_vector_operations_1): Hoist
9198 vector boolean check before scalar optimization.
9200 2018-05-28 Jakub Jelinek <jakub@redhat.com>
9202 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
9205 2018-05-28 Mark Wielaard <mark@klomp.org>
9207 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
9208 if it is an expression containing a minus sign.
9210 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
9212 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
9214 2018-05-27 Paul Koning <ni1d@arrl.net>
9216 * config/pdp11/pdp11.md (truncsihi2): Remove.
9218 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
9219 Chung-Ju Wu <jasonwucj@gmail.com>
9221 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
9223 (unaligned_store_dw): Ditto.
9224 * config/nds32/nds32-memory-manipulation.c
9225 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
9226 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
9227 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
9228 (emit_setmem_word_loop): Rename to ...
9229 (emit_setmem_doubleword_loop): ... this.
9230 (nds32_gen_dup_4_byte_to_word_value): New function.
9231 (nds32_gen_dup_8_byte_to_double_word_value): New function.
9232 (nds32_expand_setmem_loop): Refine implementation.
9233 (nds32_expand_setmem_loop_v3m): Ditto.
9234 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
9237 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
9239 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
9241 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
9243 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
9244 (nds32_init_machine_status): Initialize machine->attr_naked_p and
9245 machine->attr_no_prologue_p.
9246 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
9247 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
9248 (nds32_expand_epilogue): Consider attr_naked_p.
9249 (nds32_expand_epilogue_v3pop): Likewise.
9250 (nds32_can_use_return_insn): Likewise.
9251 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
9252 attr_no_prologue_p fields.
9253 * config/nds32/nds32.opt (mret-in-naked-func): New option.
9255 2018-05-27 Jakub Jelinek <jakub@redhat.com>
9258 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
9260 * config/i386/sse.md
9261 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
9263 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
9265 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
9267 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
9269 (*<floatsuffix>floatv2div2sf2): Rename to ...
9270 (*float<floatunssuffix>v2div2sf2): ... this.
9271 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
9272 (float<floatunssuffix>v2div2sf2_mask): ... this.
9273 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
9274 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
9275 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
9277 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
9279 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
9281 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
9283 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
9285 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
9287 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
9288 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
9289 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
9290 gen_ufix_truncv8dfv8si2.
9291 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
9292 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
9293 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
9294 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
9295 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
9296 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
9297 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
9298 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
9300 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
9304 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
9306 2018-05-25 Jim Wilson <jimw@sifive.com>
9308 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
9309 * config/riscv/riscv.c (struct machine_function): Add
9310 interrupt_handler_p and attribute_checked_p fields.
9311 (riscv_attribute_table): Add interrupt.
9312 (riscv_interrupt_type_p): New.
9313 (riscv_save_reg_p): Save extra regs for interrupt handler.
9314 (riscv_use_save_libcall): Return false for interrupt handler.
9315 (riscv_first_stack_step): Add forward declaration.
9316 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
9317 for interrupt handler with large frame. Use it for saved reg list.
9318 (riscv_expand_prologue): Move flag_stack_usage_info support to
9319 eliminate duplication.
9320 (riscv_expand_epilogue): Generate mret for interrupt handler.
9321 (riscv_epilogue_uses): New.
9322 (riscv_can_use_return_insn): Return false for interrupt handler.
9323 (riscv_function_ok_for_sibcall): Likewise.
9324 (riscv_set_current_function): Add interrupt handler support.
9325 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
9326 * config/riscv/riscv.md (UNSPECV_MRET): New.
9328 (riscv_frflags, riscv_fsflags): Use tab after opcode.
9330 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
9332 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
9334 PR tree-optimization/85712
9335 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
9336 this candidate has already been replaced in-situ by a copy.
9338 2018-05-25 Jason Merrill <jason@redhat.com>
9340 PR c++/80485 - inline function non-zero address.
9341 * symtab.c (nonzero_address): Check DECL_COMDAT.
9343 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
9346 * config/alpha/alpha.md (ashlsi3): New insn pattern.
9347 (*ashlsi_se): Rename from *ashldi_se. Define as sign
9348 extension of SImode operation. Use const123_operand predicate.
9349 (*saddsi_1): Remove.
9350 (*saddl_se_1): Ditto.
9352 (*ssubl_se_1): Ditto.
9353 * config/alpha/predicates.md (const123_operand): New predicate.
9354 * config/alpha/constraints.md (P): Use IN_RANGE.
9356 2018-05-25 Richard Biener <rguenther@suse.de>
9358 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
9360 (ref_maybe_used_by_stmt_p): Likewise.
9361 (stmt_may_clobber_ref_p): Likewise.
9362 (stmt_may_clobber_ref_p_1): Likewise.
9363 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
9365 (ref_maybe_used_by_stmt_p): Likewise.
9366 (stmt_may_clobber_ref_p): Likewise.
9367 (stmt_may_clobber_ref_p_1): Likewise.
9368 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
9369 the alias oracle to disambiguate DRs with stmts DR analysis
9371 (vect_analyze_data_refs): Do not give up on not analyzable
9372 DRs for BB vectorization. Remove code truncating the dataref
9375 2018-05-25 Jakub Jelinek <jakub@redhat.com>
9378 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
9379 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
9380 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
9382 2018-05-25 Richard Biener <rguenther@suse.de>
9384 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
9385 function, combining stmt data ref gathering and fatal analysis
9387 (vect_analyze_data_refs): Remove now redudnant code and simplify.
9388 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
9389 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
9390 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
9391 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
9393 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9395 PR tree-optimization/85720
9396 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
9397 SCC if all partitions are builtins.
9398 (version_loop_by_alias_check): New parameter. Generate cancelable
9399 runtime alias check if all partitions are builtins.
9400 (distribute_loop): Update call to above function.
9402 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9404 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
9405 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
9406 (parm_default_def_partition_arg): Ditto.
9407 (set_parm_default_def_partition): Ditto.
9408 (get_parm_default_def_partitions): Ditto and make it static.
9409 (get_undefined_value_partitions): Ditto and make it static.
9410 (remove_ssa_form): Refactor call to init_var_map here.
9411 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
9412 computation for loop region.
9413 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
9414 (register_default_def): Delete.
9415 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
9416 (parm_default_def_partition_arg): Ditto.
9417 (set_parm_default_def_partition): Ditto.
9418 (get_parm_default_def_partitions): Ditto and make it static.
9419 (get_undefined_value_partitions): Ditto and make it static.
9420 (coalesce_with_default, coalesce_with_default): Update comment.
9421 (create_coalesce_list_for_region): New func factored out from
9422 create_outofssa_var_map.
9423 (populate_coalesce_list_for_outofssa): New func factored out from
9424 create_outofssa_var_map and coalesce_ssa_name.
9425 (create_outofssa_var_map): Delete.
9426 (coalesce_ssa_name): Refactor to support live range computation.
9427 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
9428 (get_parm_default_def_partitions): Delete.
9429 (get_undefined_value_partitions): Ditto.
9430 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
9431 computation for loop region.
9432 (new_tree_live_info, loe_visit_block): Ditto.
9433 (live_worklist, set_var_live_on_entry): Ditto.
9434 (calculate_live_on_exit, verify_live_on_entry): Ditto.
9435 * tree-ssa-live.h (struct _var_map): New fields.
9436 (init_var_map): Change decl.
9437 (region_contains_p): New.
9439 2018-05-25 Bin Cheng <bin.cheng@arm.com>
9441 * tree-ssa-live.h (live_merge_and_clear): Delete.
9443 2018-05-25 Richard Biener <rguenther@suse.de>
9446 * tree-dump.c (dequeue_and_dump): Remove access to removed
9447 operand 2 of a SWITCH_EXPR.
9449 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9451 * doc/sourcebuild.texi (vect_double_cond_arith): Include
9452 multiplication and division.
9453 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
9454 (cond_udiv@var{m}, cond_umod@var{m}): Document.
9455 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
9456 (cond_udiv_optab, cond_umod_optab): New optabs.
9457 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
9458 (IFN_COND_RDIV): New internal functions.
9459 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
9460 TRUNC_MOD_EXPR and RDIV_EXPR.
9461 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
9462 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
9464 (SVE_INT_BINARY): Include mult.
9465 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
9466 (optab, sve_int_op): Handle mult.
9467 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
9469 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
9470 for SVE_INT_BINARY_SD.
9472 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9474 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
9475 (optab, sve_int_op): Handle div and udiv.
9476 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
9477 for SVE_INT_BINARY_SD.
9478 (*<optab><mode>3): New insn for the same.
9480 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9482 * tree-vect-patterns.c: Include predict.h.
9483 (vect_recog_divmod_pattern): Restrict check for division support
9484 to when optimizing for size.
9486 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9488 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
9489 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
9490 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
9491 (gimple_match_op::set_op): Likewise.
9492 (gimple_resimplify4): Declare.
9493 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
9494 (expr::gen_transform): Likewise.
9495 (decision_tree::gen): Generate a simplification routine for 4 operands.
9496 * gimple-match-head.c (gimple_simplify): Add an overload for
9497 4 operands. In the top-level function, handle up to 4 call
9498 arguments and call gimple_resimplify4.
9499 (gimple_resimplify4): New function.
9500 (build_call_internal): Pass a fourth operand.
9501 (maybe_push_to_seq): Likewise.
9502 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
9503 Fold VEC_COND_EXPRs of an operation and a default value into
9504 an IFN_COND_* function if possible.
9505 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
9507 (SVE_COND_FP_BINARY): Include them.
9508 (optab, sve_fp_op): Handle them.
9509 (SVE_INT_BINARY_REV): New code iterator.
9510 (SVE_COND_FP_BINARY_REV): New int iterator.
9511 (commutative): New int attribute.
9512 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
9514 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
9516 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
9517 (*cond_<optab><mode>): New patterns for reversed operands.
9519 2018-05-25 Richard Biener <rguenther@suse.de>
9521 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
9522 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
9523 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
9524 (STMT_VINFO_GROUPED_ACCESS): Adjust.
9525 * tree-vect-data-refs.c (everywhere): Adjust users.
9526 * tree-vect-loop.c (everywhere): Likewise.
9527 * tree-vect-slp.c (everywhere): Likewise.
9528 * tree-vect-stmts.c (everywhere): Likewise.
9529 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
9531 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9533 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
9535 (gcc_cv_as_section_exclude): ... this.
9536 Try Solaris as #exclude syntax.
9537 * configure: Regenerate.
9538 * config.in: Regenerate.
9539 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
9541 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
9542 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
9544 * varasm.c (default_elf_asm_named_section): Don't check if
9545 HAVE_GAS_SECTION_EXCLUDE is defined.
9547 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9549 * doc/md.texi: Update the documentation of the cond_* optabs
9550 to mention the new final operand. Fix GET_MODE_NUNITS call.
9551 Describe the scalar case too.
9552 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
9553 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
9555 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
9556 (get_conditional_internal_fn): Update comment.
9557 * tree-vect-loop.c (vectorizable_reduction): Pass the original
9558 accumulator value as a final argument to conditional functions.
9559 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
9560 a define_expand and add an "else" operand. Assert for now that
9561 the else operand is equal to operand 2. Use SVE_INT_BINARY and
9562 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
9563 (*cond_<optab><mode>): New patterns.
9564 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
9565 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9566 (UNSPEC_COND_EOR): Delete.
9567 (optab): Remove associated mappings.
9568 (SVE_INT_BINARY): New code iterator.
9569 (sve_int_op): Remove int attribute and add "minus" to the code
9571 (SVE_COND_INT_OP): Delete.
9572 (SVE_COND_FP_OP): Rename to...
9573 (SVE_COND_FP_BINARY): ...this.
9575 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
9577 * optabs.c (can_reuse_operands_p): New function.
9578 (maybe_legitimize_operands): Try to reuse the results for
9581 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9583 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
9584 Add {q} suffix to insn mnemonic.
9586 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9588 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
9589 (msp430_warn_func_return): New.
9591 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
9593 * fold-const.c (tree_nonzero_bits): New function.
9594 * fold-const.h (tree_nonzero_bits): Likewise.
9595 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
9596 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
9598 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
9602 * varasm.c (assemble_alias): Check ifunc_resolver only on
9605 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
9608 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
9609 when memory input operand is handled.
9611 2018-05-24 Luis Machado <luis.machado@linaro.org>
9613 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
9615 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
9617 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9619 * match.pd: Delay FMA folds until after vectorization.
9621 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
9624 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
9625 address check not strict.
9627 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
9629 * gimple-match.h (gimple_match_op): New class.
9630 (mprts_hook): Replace parameters with a gimple_match_op *.
9631 (maybe_build_generic_op): Likewise.
9632 (gimple_simplified_result_is_gimple_val): Replace parameters with
9633 a const gimple_match_op *.
9634 (gimple_simplify): Replace code_helper * and tree * parameters with
9635 a gimple_match_op * parameter.
9636 (gimple_resimplify1): Replace code_helper *, tree and tree *
9637 parameters with a gimple_match_op * parameter.
9638 (gimple_resimplify2): Likewise.
9639 (gimple_resimplify3): Likewise.
9640 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
9641 parameters with a gimple_match_op * parameter.
9642 * gimple-match-head.c (gimple_simplify): Change prototypes of
9643 auto-generated functions to take a gimple_match_op * instead of
9644 separate code_helper * and tree * parameters. Make the same
9645 change in the top-level overload and update calls to the
9646 gimple_resimplify routines. Update calls to the auto-generated
9647 functions and to maybe_push_res_to_seq in the publicly-facing
9648 operation-specific gimple_simplify overloads.
9649 (gimple_match_op::MAX_NUM_OPS): Define.
9650 (gimple_resimplify1): Replace rcode and ops with a single res_op
9651 parameter. Update call to gimple_simplify.
9652 (gimple_resimplify2): Likewise.
9653 (gimple_resimplify3): Likewise.
9654 (mprts_hook): Replace parameters with a gimple_match_op *.
9655 (maybe_build_generic_op): Likewise.
9656 (build_call_internal): Replace type, nargs and ops with
9657 a gimple_match_op *.
9658 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
9659 with a single gimple_match_op *. Update calls to mprts_hook,
9660 build_call_internal and gimple_simplified_result_is_gimple_val.
9661 Factor out code that is common to the tree_code and combined_fn cases.
9662 * genmatch.c (expr::gen_transform): Replace tem_code and
9663 tem_ops with a gimple_match_op called tem_op. Update calls
9664 to the gimple_resimplify functions and maybe_push_res_to_seq.
9665 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
9666 res_ops. Update call to the gimple_resimplify functions.
9667 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
9668 (decision_tree::gen): Make the functions take a gimple_match_op *
9669 called res_op instead of separate res_code and res_ops parameters.
9670 Update call accordingly.
9671 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
9672 and ops with a single res_op parameter. Update calls to
9673 maybe_build_generic_op and maybe_push_res_to_seq.
9674 (fold_stmt_1): Update calls to gimple_simplify and
9675 replace_stmt_with_simplification.
9676 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
9677 and gimple_simplified_result_is_gimple_val.
9678 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
9680 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
9681 with a gimple_match_op *.
9682 (vn_nary_build_or_lookup): Likewise. Update call to
9683 vn_nary_build_or_lookup_1.
9684 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
9685 gimple_match_op *. Update calls to the gimple_resimplify routines
9686 and to gimple_simplified_result_is_gimple_val.
9687 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
9688 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
9689 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
9690 (visit_nary_op): Likewise.
9691 (visit_reference_op_load): Likewise.
9693 2018-05-23 Luis Machado <luis.machado@linaro.org>
9695 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
9696 modifier for printing the step amount.
9698 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
9701 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
9704 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
9706 * doc/sourcebuild.texi (Endianness): New subsubsection.
9708 2018-05-23 Luis Machado <luis.machado@linaro.org>
9710 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9711 <prefetch_dynamic_strides>: New const bool field.
9712 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9713 prefetch_dynamic_strides.
9714 (exynosm1_prefetch_tune): Likewise.
9715 (thunderxt88_prefetch_tune): Likewise.
9716 (thunderx_prefetch_tune): Likewise.
9717 (thunderx2t99_prefetch_tune): Likewise.
9718 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
9720 (aarch64_override_options_internal): Update to set
9721 PARAM_PREFETCH_DYNAMIC_STRIDES.
9722 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9723 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9724 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9725 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9726 prefetch-dynamic-strides setting.
9728 2018-05-23 Luis Machado <luis.machado@linaro.org>
9730 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9731 <minimum_stride>: New const int field.
9732 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9733 minimum_stride field defaulting to -1.
9734 (exynosm1_prefetch_tune): Likewise.
9735 (thunderxt88_prefetch_tune): Likewise.
9736 (thunderx_prefetch_tune): Likewise.
9737 (thunderx2t99_prefetch_tune): Likewise.
9738 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
9739 <default_opt_level>: Set to 3.
9740 (aarch64_override_options_internal): Update to set
9741 PARAM_PREFETCH_MINIMUM_STRIDE.
9742 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9743 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9744 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9745 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9746 stride is constant and is below the minimum stride threshold.
9748 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9750 * config/arm/arm-cpus.in (mode26): Delete.
9751 (armv4): Delete mode26 reference.
9752 * config/arm/arm.c (arm_configure_build_target): Delete use of
9755 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9757 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
9759 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
9760 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
9762 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
9763 (floatunsdidf2): Ditto.
9765 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
9767 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
9768 (fixuns_trunc<mode>si2_avx512f): Ditto.
9769 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
9770 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
9771 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
9773 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9775 PR rtl-optimization/79985
9776 * df-scan.c (df_insn_refs_collect): Remove special case for
9777 global registers and asm statements.
9779 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
9781 * extend.texi (Global Register Variables): Rewrite the bullet list.
9782 Note that the register is available for allocation. Note that access
9783 via inline asm must use constraints. Add note about async-signal
9784 handlers. Remove paragraph about automagic register selection.
9786 2018-05-23 Richard Biener <rguenther@suse.de>
9788 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
9789 of fixed offset from memset VN.
9791 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
9793 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
9795 (alloc_cand_and_find_basis): Initialize first_interp field.
9796 (slsr_process_mul): Modify first_interp field.
9797 (slsr_process_add): Likewise.
9798 (slsr_process_cast): Modify first_interp field for each new
9800 (slsr_process_copy): Likewise.
9801 (dump_candidate): Dump first_interp field.
9802 (replace_mult_candidate): Process all interpretations, not just
9804 (replace_rhs_if_not_dup): Likewise.
9805 (replace_one_candidate): Likewise.
9807 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
9809 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
9811 (aarch64_needs_frame_chain): New function.
9812 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
9814 2018-05-23 Sudakshina Das <sudi.das@arm.com>
9817 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
9818 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
9819 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
9820 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
9821 as true for strict-align.
9822 (aarch64_can_inline_p): Perform checks even when callee has no
9823 attributes to check for strict alignment.
9824 * doc/extend.texi (AArch64 Function Attributes): Document
9826 * doc/invoke.texi: (AArch64 Options): Likewise.
9828 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
9830 PR tree-optimization/85853
9831 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
9832 the handling of the root of the node to...
9833 (vect_slp_analyze_node_operations_1): ...this new function,
9834 and run the whole thing with the child nodes' def types
9835 set according to their SLP node's def type.
9837 2018-05-23 Richard Biener <rguenther@suse.de>
9840 * tree-data-ref.c (create_runtime_alias_checks): Defer
9841 and ignore overflow warnings.
9843 2018-05-23 Yury Gribov <tetra2005@gmail.com>
9845 PR tree-optimization/85822
9846 * tree-vrp.c (is_masked_range_test): Fix handling of negative
9849 2018-05-23 Richard Biener <rguenther@suse.de>
9851 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
9852 memset constants via native_interpret_expr.
9854 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
9857 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
9859 (cgraph_node::create_alias): Likewise.
9860 (cgraph_node::get_availability): Check ifunc_resolver instead
9861 of looking up ifunc attribute.
9862 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9863 * varasm.c (do_assemble_alias): Likewise.
9864 (assemble_alias): Likewise.
9865 (default_binds_local_p_3): Likewise.
9866 * cgraph.h (cgraph_node): Add ifunc_resolver.
9867 (cgraph_node::only_called_directly_or_aliased_p): Return false
9869 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
9871 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
9872 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
9873 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
9874 instead of looking up ifunc attribute.
9876 2018-05-22 Luis Machado <luis.machado@linaro.org>
9878 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
9880 2018-05-22 Martin Sebor <msebor@redhat.com>
9883 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
9884 only when expasion succeeds.
9885 (expand_builtin_strcmp): Same.
9886 (expand_builtin_strncmp): Same.
9888 2018-05-22 Martin Sebor <msebor@redhat.com>
9890 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
9892 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9893 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9895 * config/aarch64/aarch64-ldpstp.md: Replace uses of
9896 aarch64_mem_pair_operand with memory_operand and delete operand swapping
9898 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9899 Add check for legitimate_address.
9900 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
9901 (aarch64_swap_ldrstr_operands): New.
9902 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
9905 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
9906 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9908 * config/aarch64/aarch64.md: New patterns to generate stp
9910 (store_pair_sw, store_pair_dw): New patterns to generate stp for
9911 single words and double words.
9912 (load_pair_sw, load_pair_dw): Likewise.
9913 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
9915 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
9917 * config/aarch64/aarch64-ldpstp.md: Modify peephole
9918 for different mode ldpstp and add peephole for merged zero stores.
9920 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
9922 (aarch64_gen_store_pair): Rename calls to match new patterns.
9923 (aarch64_gen_load_pair): Rename calls to match new patterns.
9924 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
9925 (load_pair<DREG:mode><DREG2:mode>): ... This.
9926 (store_pair<mode>): Rename to...
9927 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
9928 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
9930 (V_INT_EQUIV): Handle SImode.
9931 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
9934 2018-05-22 Martin Sebor <msebor@redhat.com>
9937 * calls.c (maybe_warn_nonstring_arg): Use string length to set
9938 or ajust the presumed bound on an operation to avoid unnecessary
9941 2018-05-22 Martin Sebor <msebor@redhat.com>
9943 PR tree-optimization/85826
9944 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
9945 assuming that a DECL necesarily has a constant size.
9947 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
9950 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
9952 2018-05-22 Richard Biener <rguenther@suse.de>
9954 PR tree-optimization/85834
9955 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
9956 non-constant and non-zero memset arguments.
9958 2018-05-22 Martin Liska <mliska@suse.cz>
9961 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
9963 2018-05-22 Richard Biener <rguenther@suse.de>
9965 PR tree-optimization/85863
9966 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
9967 comparisons when vectype is specified.
9968 (vectorizable_condition): Do not specify vectype for
9969 vect_is_simple_cond when SLP vectorizing.
9971 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
9974 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
9975 define __ibm128 as long double.
9976 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
9977 as a distinct type when IEEE 128-bit support is enabled.
9978 (init_float128_ieee): Fix up conversions between IFmode and IEEE
9979 128-bit types to use the correct functions.
9980 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
9981 convert between 128-bit floating point types that have different
9982 modes but the same representation, instead of using gen_lowpart to
9984 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
9986 (IFKF_reg): New attributes to give the register constraints for
9988 (extend<mode>tf2_internal): New insns to mark an explicit
9989 conversion between 128-bit floating point types that have a
9990 different mode but share the same representation.
9992 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
9994 PR tree-optimization/85814
9995 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
9996 a null return from get_strinfo when unsharing the next
9997 strinfo in the chain.
9999 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
10002 * varasm.c (weak_finish): Clean up weak_decls.
10004 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10006 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
10007 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
10008 UNSPEC_UADALP values.
10009 * config/aarch64/iterators.md (ABAL): New int iterator.
10012 (sur): Add mappings for the above.
10013 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
10015 (aarch64_<sur>abal<mode>_4): Likewise.
10016 (aarch64_<sur>adalp<mode>_3): Likewise.
10017 (<sur>sadv16qi): New define_expand.
10019 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
10021 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
10022 (*movdf_internal): Ditto.
10023 (*rcpsf2_sse): Ditto.
10024 (*rsqrtsf2_sse): Ditto.
10025 (*sqrt<mode>2_sse): Ditto.
10027 2018-05-21 Tamar Christina <tamar.christina@arm.com>
10029 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
10031 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
10032 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
10033 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
10034 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
10036 * config/aarch64/arm_neon.h: Likewise.
10037 * config/aarch64/iterators.md (VQ_I): New.
10039 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
10041 * config.gcc: Add arc/t-multilib-linux to tmake_file for
10043 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
10046 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
10048 * config/nds32/constraints.md (S): New constraint.
10049 * config/nds32/nds32.md (call_internal): Use constraint S.
10050 (call_value_internal): Likewise.
10051 (sibcall_internal): Likewise.
10052 (sibcall_value_internal): Likewise.
10054 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
10055 Chung-Ju Wu <jasonwucj@gmail.com>
10057 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
10058 into consideration.
10060 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
10061 Chung-Ju Wu <jasonwucj@gmail.com>
10063 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
10064 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
10065 (nds32_rtx_costs_impl): Simplify.
10066 (nds32_address_cost_impl): Simplify.
10067 (nds32_init_rtx_costs): New function.
10068 (nds32_rtx_costs_speed_prefer): Likewise.
10069 (nds32_rtx_costs_size_prefer): Likewise.
10070 (nds32_address_cost_speed_prefer): Likewise.
10071 (nds32_address_cost_speed_fwprop): Likewise.
10072 (nds32_address_cost_size_prefer): Likewise.
10073 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
10074 * config/nds32/nds32.c (nds32_option_override): Use
10075 nds32_init_rtx_costs function.
10077 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
10079 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
10080 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
10081 (TARGET_PIPELINE_N8): Likewise.
10082 (TARGET_PIPELINE_N10): Likewise.
10083 (TARGET_PIPELINE_N13): Likewise.
10084 (TARGET_PIPELINE_GRAYWOLF): Likewise.
10086 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
10088 * config/nds32/nds32-fpu.md: Update copyright year.
10090 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10092 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
10094 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10096 * config/nds32/nds32.c
10097 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
10098 * config/nds32/nds32.opt (minline-asm-r15): New option.
10100 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10102 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
10104 * config/nds32/nds32.md (abssi2): New pattern.
10106 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
10108 * config/i386/i386.md (rex64namesuffix): New mode attribute.
10109 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
10110 Merge insn pattern from sse_cvtsi2ss<round_name> and
10111 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
10112 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
10113 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
10114 using SWI48 mode iterator.
10115 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
10116 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
10117 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
10118 pattern from sse_cvttss2si<round_saeonly_name>
10119 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
10120 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
10121 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
10122 using SWI48 mode iterator.
10123 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
10124 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
10125 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
10126 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
10127 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
10128 using SWI48 mode iterator.
10129 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
10130 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
10131 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
10132 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
10133 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
10134 SWI48 mode iterator.
10135 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
10136 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
10137 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
10138 pattern from sse_cvttsd2si<round_saeonly_name>
10139 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
10141 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10143 * config/nds32/nds32-md-auxiliary.c
10144 (nds32_valid_smw_lwm_base_p): Refine.
10145 (nds32_output_smw_single_word): Refine.
10146 (nds32_output_smw_double_word): New.
10147 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
10149 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10151 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
10152 (nds32_output_stack_pop): Refine.
10153 (nds32_expand_unaligned_load): Refine.
10154 (nds32_expand_unaligned_store): Refine.
10156 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10157 Chung-Ju Wu <jasonwucj@gmail.com>
10159 * config/nds32/constants.md: Add TP_REGNUM constant.
10160 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
10161 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
10163 * config/nds32/nds32-doubleword.md: Consider flag_pic.
10164 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
10165 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
10166 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
10167 and PIC code generation.
10168 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
10170 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
10172 * config/nds32/nds32.md: Support TLS and PIC.
10173 * config/nds32/nds32.c: Support TLS and PIC.
10174 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
10175 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
10178 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
10180 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
10181 mode with E_ prefix.
10183 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10184 Chung-Ju Wu <jasonwucj@gmail.com>
10186 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
10187 * config/nds32/nds32-md-auxiliary.c
10188 (symbolic_reference_mentioned_p): New.
10189 (nds32_legitimize_ict_address): New.
10190 (nds32_expand_ict_move): New.
10191 (nds32_indirect_call_referenced_p): New.
10192 (nds32_symbol_binds_local_p): Delete.
10193 (nds32_long_call_p): Modify.
10194 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
10195 * config/nds32/nds32-protos.h
10196 (symbolic_reference_mentioned_p): Declare.
10197 (nds32_legitimize_ict_address): Declare.
10198 (nds32_expand_ict_move): Declare.
10199 (nds32_indirect_call_referenced_p): Declare.
10200 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
10201 (nds32_relax_group): Use nds32_ict_const_p as condition.
10202 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
10203 (nds32_asm_file_start): Output ict_model directive in asm code.
10204 (nds32_legitimate_address_p): Consider indirect call.
10205 (nds32_print_operand): Consider indirect call.
10206 (nds32_print_operand_address): Consider indirect call.
10207 (nds32_insert_attributes): Handle "indirect_call" attribute.
10208 (TARGET_LEGITIMATE_ADDRESS_P): Define.
10209 (TARGET_LEGITIMATE_CONSTANT_P): Define.
10210 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
10211 (TARGET_DELEGITIMIZE_ADDRESS): Define.
10212 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
10213 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
10214 (TARGET_ICT_MODEL_SMALL): Define.
10215 (TARGET_ICT_MODEL_LARGE): Define.
10216 * config/nds32/nds32.md (movsi): Consider ict model.
10217 (call, call_value): Consider ict model.
10218 (sibcall, sibcall_value): Consider ict model.
10219 * config/nds32/nds32.opt (mict-model): New option.
10220 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
10223 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
10224 Monk Chiang <sh.chiang04@gmail.com>
10225 Jim Wilson <jimw@sifive.com>
10227 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
10228 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
10229 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
10230 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
10231 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
10232 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
10233 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
10234 compute save_libcall_adjustment properly.
10235 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
10236 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
10237 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
10238 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
10239 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
10240 (ABI_SPEC): Handle mabi=ilp32e.
10241 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
10242 (RVE): Add RVE mask.
10243 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
10244 <-march>: Add rv32e as an example.
10246 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
10249 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
10250 (intra_create_variable_infos): Handle C++ constructors.
10252 2018-05-18 Martin Liska <mliska@suse.cz>
10254 * passes.def: Remove a redundant pass.
10256 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
10259 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
10261 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10263 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
10265 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
10267 (armv2, armv2a, armv3, armv3m): Delete architectures.
10268 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
10269 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
10270 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
10272 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
10273 (*mulsidi3adddi): Likewise.
10274 (mulsidi3): Likewise.
10275 (*mulsidi3_nov6): Likewise.
10276 (umulsidi3): Likewise.
10277 (umulsidi3_nov6): Likewise.
10278 (umaddsidi4): Likewise.
10279 (*umulsidi3adddi): Likewise.
10280 (smulsi3_highpart): Likewise.
10281 (*smulsi3_highpart_nov6): Likewise.
10282 (umulsi3_highpart): Likewise.
10283 (*umulsi3_highpart_nov6): Likewise.
10284 * config/arm/arm.h (arm_arch3m): Delete.
10285 * config/arm/arm.c (arm_arch3m): Delete.
10286 (arm_option_override_internal): Update armv3-related comment.
10287 (arm_configure_build_target): Delete use of isa_bit_mode32.
10288 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
10289 (arm_rtx_costs_internal): Delete check of arm_arch3m.
10290 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
10291 (mulsa3): Likewise.
10292 (mulusa3): Likewise.
10293 * config/arm/arm-protos.h (arm_arch3m): Delete.
10294 * config/arm/arm-tables.opt: Regenerate.
10295 * config/arm/arm-tune.md: Likewise.
10296 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
10297 deleted architectures.
10299 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10301 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
10302 (armv5t, armv5te): New features.
10303 (ARMv5, ARMv5e): Delete fgroups.
10304 (ARMv5t, ARMv5te): Adjust for above changes.
10305 (ARMv6m): Likewise.
10306 (armv5, armv5e): Delete arches.
10307 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
10309 (*call_reg_arm): Likewise.
10310 (*call_value_reg_armv5): Likewise.
10311 (*call_value_reg_arm): Likewise.
10312 (*call_symbol): Likewise.
10313 (*call_value_symbol): Likewise.
10314 (*sibcall_insn): Likewise.
10315 (*sibcall_value_insn): Likewise.
10316 (clzsi2): Likewise.
10317 (prefetch): Likewise.
10318 (define_split and define_peephole2 dependent on arm_arch5):
10320 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
10322 (TARGET_ARM_QBIT): Likewise.
10323 (TARGET_DSP_MULTIPLY): Likewise.
10324 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
10325 (arm_arch5, arm_arch5e): Delete.
10326 (arm_arch5t, arm_arch5te): Declare.
10327 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
10328 (arm_arch5t): Declare.
10329 (arm_option_reconfigure_globals): Update for the above.
10330 (arm_options_perform_arch_sanity_checks): Update comment, replace
10331 use of arm_arch5 with arm_arch5t.
10332 (use_return_insn): Likewise.
10333 (arm_emit_call_insn): Likewise.
10334 (output_return_instruction): Likewise.
10335 (arm_final_prescan_insn): Likewise.
10336 (arm_coproc_builtin_available): Likewise.
10337 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
10338 arm_arch5e with arm_arch5t and arm_arch5te.
10339 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
10340 (arm_arch5t, arm_arch5te): Declare.
10341 * config/arm/arm-tables.opt: Regenerate.
10342 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
10343 * config/arm/t-multilib: Likewise.
10344 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
10345 instead of arm_arch5.
10346 (*call_reg_thumb1): Likewise.
10347 (*call_value_reg_thumb1_v5): Likewise.
10348 (*call_value_reg_thumb1): Likewise.
10349 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
10351 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
10353 2018-05-18 Martin Liska <mliska@suse.cz>
10355 PR gcov-profile/84846
10356 * doc/gcov.texi: Document -t option of gcov tool.
10358 2018-05-18 Martin Liska <mliska@suse.cz>
10360 PR gcov-profile/84846
10361 * gcov.c (print_usage): Add new -t option.
10362 (process_args): Handle the option.
10363 (generate_results): Use stdout as output when requested by
10366 2018-05-18 Martin Liska <mliska@suse.cz>
10368 PR gcov-profile/84846
10369 * coverage.c (coverage_init): Write PWD to .gcno file.
10370 * doc/gcov.texi: Document how working directory is printed.
10371 * gcov-dump.c (dump_gcov_file): Print PWD.
10372 * gcov.c (output_intermediate_file): Likewise.
10373 (read_graph_file): Read PWD string.
10374 (output_lines): Print PWD.
10376 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10378 PR middle-end/85817
10379 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
10380 for retval and return false if all args to phi are zero.
10382 2018-05-18 Richard Biener <rguenther@suse.de>
10384 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
10386 (evrp_dom_walker::before_dom_children): Call it.
10388 2018-05-18 Richard Biener <rguenther@suse.de>
10390 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
10391 results when processing array refs with variable index.
10393 2018-05-18 Toon Moene <toon@moene.org>
10395 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
10396 directly after that of -floop-interchange. Indicate that both
10397 options are enabled by default when specifying -O3.
10399 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10401 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
10402 iterator. Delete separate integer-mode vec_set<mode> expander.
10403 (aarch64_simd_vec_setv2di): Delete.
10404 (vec_setv2di): Delete.
10405 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
10406 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
10407 the "w, r" alternative.
10409 2018-05-18 Martin Liska <mliska@suse.cz>
10411 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
10412 * tree-pass.h (make_pass_lower_switch_O0): New function.
10413 * tree-switch-conversion.c (node_has_low_bound): Remove.
10414 (node_has_high_bound): Likewise.
10415 (node_is_bounded): Likewise.
10416 (class pass_lower_switch): Make it a template type and create
10418 (pass_lower_switch::execute): Add template argument.
10419 (make_pass_lower_switch): New function.
10420 (make_pass_lower_switch_O0): New function.
10421 (do_jump_if_equal): Remove.
10422 (emit_case_nodes): Simplify to just handle all 3 cases and leave
10423 all the hard work to tree optimization passes.
10425 2018-05-18 Martin Liska <mliska@suse.cz>
10427 * dbgcnt.c (limit_low): Renamed from limit.
10428 (limit_high): New variable.
10429 (dbg_cnt_is_enabled): Check for upper limit.
10430 (dbg_cnt): Adjust dumping.
10431 (dbg_cnt_set_limit_by_index): Add new argument for high
10433 (dbg_cnt_set_limit_by_name): Likewise.
10434 (dbg_cnt_process_single_pair): Parse new format.
10435 (dbg_cnt_process_opt): Use strtok.
10436 (dbg_cnt_list_all_counters): Remove 'value' and add
10438 * doc/invoke.texi: Document changes.
10440 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
10442 * doc/sourcebuild.texi (scalar_all_fma): Document.
10443 * tree.def (FMA_EXPR): Delete.
10444 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
10445 * internal-fn.c (ternary_direct): New macro.
10446 (expand_ternary_optab_fn): Likewise.
10447 (direct_ternary_optab_supported_p): Likewise.
10448 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
10449 * builtins.c (fold_builtin_fma): Delete.
10450 (fold_builtin_3): Don't call it.
10451 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
10452 * expr.c (expand_expr_real_2): Likewise.
10453 * fold-const.c (operand_equal_p): Likewise.
10454 (fold_ternary_loc): Likewise.
10455 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
10456 * gimple.c (DEFTREECODE): Likewise.
10457 * gimplify.c (gimplify_expr): Likewise.
10458 * optabs-tree.c (optab_for_tree_code): Likewise.
10459 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
10460 * tree-eh.c (operation_could_trap_p): Likewise.
10461 (stmt_could_throw_1_p): Likewise.
10462 * tree-inline.c (estimate_operator_cost): Likewise.
10463 * tree-pretty-print.c (dump_generic_node): Likewise.
10464 (op_code_prio): Likewise.
10465 * tree-ssa-loop-im.c (stmt_cost): Likewise.
10466 * tree-ssa-operands.c (get_expr_operands): Likewise.
10467 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
10468 * fold-const-call.h (fold_fma): Delete.
10469 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
10470 CFN_FNMA and CFN_FNMS.
10471 (fold_fma): Delete.
10472 * genmatch.c (combined_fn): New enum.
10473 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
10474 (commutative_op): New function.
10475 (commutate): Use it. Handle more than 2 operands.
10476 (dt_operand::gen_gimple_expr): Use commutative_op.
10477 (parser::parse_expr): Allow :c to be used with non-binary
10478 operators if the commutative operand is known.
10479 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
10480 CFN_FMS, CFN_FNMA and CFN_FNMS.
10481 (backprop::process_assign_use): Remove FMA_EXPR handling.
10482 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
10483 (gen_hsa_fma): New function.
10484 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
10485 IFN_FNMA and IFN_FNMS.
10486 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
10487 * gimple-fold.h (follow_all_ssa_edges): Declare.
10488 * gimple-fold.c (follow_all_ssa_edges): New function.
10489 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
10490 gimple_build interface and use follow_all_ssa_edges to fold the result.
10491 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
10492 instead of checking for optabs directly.
10493 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
10494 rather than FMA_EXPRs.
10495 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
10496 call to IFN_FMA instead of an FMA_EXPR.
10498 2018-05-17 Jim Wilson <jimw@sifive.com>
10500 * expr.c (do_tablejump): When converting index to Pmode, if we have a
10501 sign extended promoted subreg, and the range does not have the sign bit
10502 set, then do a sign extend.
10504 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
10505 test, check for sign extended subreg and/or constant operands, and
10506 do a sign extend in that case.
10508 2018-05-17 Steve Ellcey <sellcey@cavium.com>
10510 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
10511 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
10513 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
10514 Change logics_shift_reg to logics_shift_imm.
10515 (thunderx2t99_fp_loadpair_basic): Delete.
10516 (thunderx2t99_fp_storepair_basic): Delete.
10517 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
10518 (thunderx2t99_asimd_polynomial): Delete.
10519 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
10520 and neon_fp_mul_d_scalar_q.
10521 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
10522 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
10523 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
10524 (thunderx2t99_asimd_lut): Add missing tbl types.
10525 (thunderx2t99_asimd_ext): Delete.
10526 (thunderx2t99_asimd_load1_1_mult): Delete.
10527 (thunderx2t99_asimd_load1_2_mult): Delete.
10528 (thunderx2t99_asimd_load1_ldp): New.
10529 (thunderx2t99_asimd_load1): New.
10530 (thunderx2t99_asimd_load2): Add missing *load2* types.
10531 (thunderx2t99_asimd_load3): New.
10532 (thunderx2t99_asimd_load4): New.
10533 (thunderx2t99_asimd_store1_1_mult): Delete.
10534 (thunderx2t99_asimd_store1_2_mult): Delete.
10535 (thunderx2t99_asimd_store2_mult): Delete.
10536 (thunderx2t99_asimd_store2_onelane): Delete.
10537 (thunderx2t99_asimd_store_stp): New.
10538 (thunderx2t99_asimd_store1): New.
10539 (thunderx2t99_asimd_store2): New.
10540 (thunderx2t99_asimd_store3): New.
10541 (thunderx2t99_asimd_store4): New.
10543 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
10545 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
10546 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
10548 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
10549 Segher Boessenkool <segher@kernel.crashing.org>
10552 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
10555 2018-05-17 Richard Biener <rguenther@suse.de>
10557 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
10558 for pruning loop and prune defs feeding only already visited PHIs.
10560 2018-05-17 Richard Biener <rguenther@suse.de>
10562 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
10564 2018-05-17 Bin Cheng <bin.cheng@arm.com>
10565 Richard Biener <rguenther@suse.de>
10567 PR tree-optimization/85793
10568 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
10569 for VMAT_ELEMENTWISE.
10571 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10573 * internal-fn.h (lookup_internal_fn): Declare
10574 * internal-fn.c (lookup_internal_fn): New function.
10575 * gimple.c (gimple_build_call_from_tree): Handle calls to
10576 internal functions.
10577 * gimple-pretty-print.c (dump_gimple_call): Print "." before
10578 internal function names.
10579 * tree-pretty-print.c (dump_generic_node): Likewise.
10580 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
10582 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
10584 * gimple-fold.h (gimple_build): Make the function forms take
10585 combined_fn rather than built_in_function.
10586 (gimple_simplify): Likewise.
10587 * gimple-match-head.c (gimple_simplify): Likewise.
10588 * gimple-fold.c (gimple_build): Likewise.
10589 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
10590 rather than gimple_build_call_internal.
10591 (get_initial_defs_for_reduction): Likewise.
10592 (vect_create_epilog_for_reduction): Likewise.
10593 (vectorizable_live_operation): Likewise.
10595 2018-05-17 Martin Liska <mliska@suse.cz>
10597 * gimple-ssa-sprintf.c (format_directive): Do not use
10598 space in between 'G_' and '('.
10600 2018-05-17 Jakub Jelinek <jakub@redhat.com>
10603 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
10604 even if the mask is not all ones.
10607 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
10609 (ix86_gimple_fold_builtin): Likewise.
10612 * config/i386/i386.c: Include tree-vector-builder.h.
10613 (ix86_vector_shift_count): New function.
10614 (ix86_fold_builtin): Fold shift builtins by scalar count.
10615 (ix86_gimple_fold_builtin): Likewise.
10617 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
10618 _mm512_setzero): New intrinsics.
10620 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
10621 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10623 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
10624 code generation for cases where splatting a value is not useful.
10625 * simplify-rtx.c (simplify_ternary_operation): Simplify
10626 vec_merge across a vec_duplicate and a paradoxical subreg forming
10627 a vector mode to a vec_concat.
10629 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
10631 * config.gcc: Support "goldmont-plus".
10632 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10634 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10635 PROCESSOR_GOLDMONT_PLUS.
10636 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
10637 (processor_target_table): Add "goldmont-plus".
10638 (PTA_GOLDMONT_PLUS): Define.
10639 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
10640 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
10641 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
10642 (fold_builtin_cpu): Add "goldmont-plus".
10643 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
10644 (ix86_option_override_internal): Add "goldmont-plus".
10645 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
10646 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
10647 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
10648 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
10650 2018-05-17 Richard Biener <rguenther@suse.de>
10652 PR tree-optimization/85757
10653 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
10654 remove defs that only feed that PHI from further processing.
10656 2018-05-16 Jim Wilson <jimw@sifive.com>
10658 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
10660 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
10662 2018-05-16 Mark Wielaard <mark@klomp.org>
10664 * dwarf2out.c (count_index_strings): New function.
10665 (output_indirect_strings): Call count_index_strings and generate
10666 header for dwarf_version >= 5.
10668 2018-05-16 Mark Wielaard <mark@klomp.org>
10670 * dwarf2out.c (dwarf_FORM): New function.
10671 (set_indirect_string): Use dwarf_FORM.
10672 (reset_indirect_string): Likewise.
10673 (size_of_die): Likewise.
10674 (value_format): Likewise.
10675 (output_die): Likewise.
10676 (add_skeleton_AT_string): Likewise.
10677 (output_macinfo_op): Likewise.
10678 (index_string): Likewise.
10679 (output_index_string_offset): Likewise.
10680 (output_index_string): Likewise.
10681 (count_index_strings): Likewise.
10683 2018-05-16 Carl Love <cel@us.ibm.com>
10685 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
10686 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
10688 2018-05-16 Martin Jambor <mjambor@suse.cz>
10690 * ipa-prop.c (ipa_free_all_edge_args): Remove.
10691 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
10693 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
10695 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
10696 (fnma<mode>4): Likewise.
10697 (fms<mode>4): Likewise.
10698 (fnms<mode>4): Likewise.
10699 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
10700 (aarch64_fnma<mode>4): Likewise.
10701 (aarch64_fms<mode>4): Likewise.
10702 (aarch64_fnms<mode>4): Likewise.
10703 (aarch64_fnmadd<mode>4): Likewise.
10705 2018-05-16 Jason Merrill <jason@redhat.com>
10707 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
10709 2018-05-16 Richard Biener <rguenther@suse.de>
10711 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
10712 (dump_stmt_cost): Declare.
10713 (add_stmt_cost): Dump cost we add.
10714 (add_stmt_costs): New function.
10715 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
10716 No longer exported.
10717 (vect_analyze_stmt): Adjust prototype.
10718 (vectorizable_condition): Likewise.
10719 (vectorizable_live_operation): Likewise.
10720 (vectorizable_reduction): Likewise.
10721 (vectorizable_induction): Likewise.
10722 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
10723 cost vector to pass to vectorizable_ and record afterwards.
10724 (vect_model_reduction_cost): Take cost vector argument and adjust.
10725 (vect_model_induction_cost): Likewise.
10726 (vectorizable_reduction): Likewise.
10727 (vectorizable_induction): Likewise.
10728 (vectorizable_live_operation): Likewise.
10729 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
10730 SLP_TREE_NUMBER_OF_VEC_STMTS.
10731 (vect_analyze_slp_cost_1): Remove.
10732 (vect_analyze_slp_cost): Likewise.
10733 (vect_slp_analyze_node_operations): Take visited args and
10734 a target cost vector. Avoid processing already visited stmt sets.
10735 (vect_slp_analyze_operations): Use a local cost vector to gather
10736 costs and register those of non-discarded instances.
10737 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
10738 (vect_schedule_slp_instance): Remove copying of
10739 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
10741 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
10742 adding cost. Record cost entry location.
10743 (vect_prologue_cost_for_slp_op): Function to compute cost of
10744 a constant or invariant generated for SLP vect in the prologue,
10745 split out from vect_analyze_slp_cost_1.
10746 (vect_model_simple_cost): Make static. Adjust for SLP costing.
10747 (vect_model_promotion_demotion_cost): Likewise.
10748 (vect_model_store_cost): Likewise, make static.
10749 (vect_model_load_cost): Likewise.
10750 (vectorizable_bswap): Add cost vector arg and adjust.
10751 (vectorizable_call): Likewise.
10752 (vectorizable_simd_clone_call): Likewise.
10753 (vectorizable_conversion): Likewise.
10754 (vectorizable_assignment): Likewise.
10755 (vectorizable_shift): Likewise.
10756 (vectorizable_operation): Likewise.
10757 (vectorizable_store): Likewise.
10758 (vectorizable_load): Likewise.
10759 (vectorizable_condition): Likewise.
10760 (vectorizable_comparison): Likewise.
10761 (can_vectorize_live_stmts): Likewise.
10762 (vect_analyze_stmt): Likewise.
10763 (vect_transform_stmt): Adjust calls to vectorizable_*.
10764 * tree-vectorizer.c: Include gimple-pretty-print.h.
10765 (dump_stmt_cost): New function.
10767 2018-05-16 Richard Biener <rguenther@suse.de>
10769 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
10770 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
10771 * tree-ssa-dse.c: Include tree-ssa-loop.h.
10772 (check_name): New callback.
10773 (dse_classify_store): Track cycles via a visited bitmap of PHI
10774 defs and simplify handling of in-loop and across loop dead stores
10775 and properly fail for loop-variant refs. Handle byte-tracking with
10776 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
10779 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
10781 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
10782 (vect_get_mask_type_for_stmt): Likewise.
10783 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
10785 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
10786 to determine the statement's vector type and the vector type that
10787 should be used for calculating nunits. Deal with cases in which
10788 the type has to be deferred.
10789 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
10790 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
10791 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
10792 (vect_determine_vf_for_stmt): New functions, split out from...
10793 (vect_determine_vectorization_factor): ...here.
10794 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
10795 (vect_get_mask_type_for_stmt): New functions, split out from
10796 vect_determine_vectorization_factor.
10798 2018-05-16 Richard Biener <rguenther@suse.de>
10800 * tree-cfg.c (verify_gimple_assign_ternary): Properly
10801 verify the [VEC_]COND_EXPR embedded comparison.
10803 2018-05-15 Martin Sebor <msebor@redhat.com>
10805 PR tree-optimization/85753
10806 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
10807 RECORD_TYPE in addition to ARRAY_TYPE.
10809 2018-05-15 Martin Sebor <msebor@redhat.com>
10811 PR middle-end/85643
10812 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
10814 2018-05-15 Richard Biener <rguenther@suse.de>
10816 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
10817 add by_clobber_p one. Change algorithm to collect all defs
10818 representing uses we need to walk and try reducing them to
10819 a single one before failing.
10820 (dse_dom_walker::dse_optimize_stmt): Adjust.
10822 2018-05-13 Mark Wielaard <mark@klomp.org>
10824 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
10825 (size_of_loc_descr): Likewise.
10826 (output_loc_operands): Likewise.
10827 (output_loc_operands_raw): Likewise.
10828 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
10829 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
10830 (hash_loc_operands): Likewise.
10831 (compare_loc_operands): Likewise.
10833 2018-05-14 Mark Wielaard <mark@klomp.org>
10835 * dwarf2out.c (count_index_addrs): New function.
10836 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
10838 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10840 PR tree-optimization/83648
10841 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
10842 return value as malloc candidate.
10844 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10847 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
10848 param as true in call to suggest_attribute.
10850 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
10852 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
10853 -mreadonly-in-sdata.
10855 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10857 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
10859 (aarch64_crypto_aesd_fused): Likewise.
10861 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
10863 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
10864 (movsi_aarch64): Likewise.
10865 (load_pairsi): Likewise.
10866 (load_pairdi): Likewise.
10867 (store_pairsi): Likewise.
10868 (store_pairdi): Likewise.
10869 (load_pairsf): Likewise.
10870 (load_pairdf): Likewise.
10871 (store_pairsf): Likewise.
10872 (store_pairdf): Likewise.
10873 (zero_extend): Likewise.
10874 (trunc): Swap alternatives.
10875 (fcvt_target): Add '?' to prefer w over r.
10877 2018-05-14 Jakub Jelinek <jakub@redhat.com>
10880 * config/i386/i386.md: Disallow non-commutative arithmetics in
10881 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
10882 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
10883 in the peephole2 before it.
10885 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
10887 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
10888 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
10889 (ix86_handle_option): Handle -mcldemote.
10890 * config.gcc: New header.
10891 * config/i386/cldemoteintrin.h: New file.
10892 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
10893 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
10895 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10896 OPTION_MASK_ISA_CLDEMOTE.
10897 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
10898 (ix86_valid_target_attribute_inner_p): Ditto.
10899 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
10900 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
10901 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
10902 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
10903 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
10905 * config/i386/i386.opt: Add -mcldemote.
10906 * config/i386/x86intrin.h: New header.
10907 * doc/invoke.texi: Add -mcldemote.
10909 2018-05-14 Richard Biener <rguenther@suse.de>
10911 * doc/match-and-simplify.texi: Adjust :s documentation.
10913 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
10915 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
10916 intended memcpy size.
10917 (REORDER_45): Likewise.
10919 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
10921 * sort.cc: New file.
10922 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
10923 * vec.c (qsort_chk): Use gcc_qsort.
10924 * Makefile.in (OBJS-libcommon): Add sort.o.
10925 (build/sort.o): New target. Use it...
10926 (BUILD_RTL): ... here, and...
10927 (build/gencfn-macros): ... here, and...
10928 (build/genmatch): ... here.
10930 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10931 Chung-Ju Wu <jasonwucj@gmail.com>
10933 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
10934 * config/nds32/nds32-graywolf.md: New file.
10935 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
10936 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
10938 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
10939 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
10940 * config/nds32/nds32.md (pipeline_model): Add graywolf.
10941 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
10942 * config/nds32/pipelines.md: Include n15 settings.
10944 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10945 Chung-Ju Wu <jasonwucj@gmail.com>
10947 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
10948 * config/nds32/nds32-n13.md: New file.
10949 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
10950 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
10952 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
10953 * config/nds32/nds32.md (pipeline_model): Add n13.
10954 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
10955 * config/nds32/pipelines.md: Include n13 settings.
10957 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
10958 Chung-Ju Wu <jasonwucj@gmail.com>
10960 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
10961 * config/nds32/nds32-n10.md: New file.
10962 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
10963 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
10965 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
10966 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
10967 * config/nds32/nds32.md (pipeline_model): Add n10.
10968 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
10969 * config/nds32/pipelines.md: Include n10 settings.
10971 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
10972 Kito Cheng <kito.cheng@gmail.com>
10973 Chung-Ju Wu <jasonwucj@gmail.com>
10975 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
10976 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
10977 Add enum values for DSP extension instructions.
10978 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
10980 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
10981 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
10982 New code iterators.
10983 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
10984 * config/nds32/nds32-dspext.md: New file for DSP implementation.
10985 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
10986 * config/nds32/nds32-intrinsic.md: Likewise.
10987 * config/nds32/nds32_intrinsic.h: Likewise.
10988 * config/nds32/nds32-md-auxiliary.c: Likewise.
10989 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
10990 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
10991 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
10992 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
10993 * config/nds32/nds32-protos.h: New declarations for DSP extension.
10994 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
10995 TYPE_DMAC in switch statement.
10996 * config/nds32/nds32.c: New checking and implementation for DSP
10997 extension instructions.
10998 * config/nds32/nds32.h: Likewise.
10999 * config/nds32/nds32.md: Likewise.
11000 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
11001 * config/nds32/predicates.md: Implement new predicates for DSP
11004 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11006 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
11007 Reformat alternatives and attributes so it is easier to identify
11008 which constraints/attributes go with which instruction.
11009 (mov<mode>_hardfloat32, FMOVE64): Likewise.
11010 (mov<mode>_softfloat32, FMOVE64): Likewise.
11011 (mov<mode>_hardfloat64, FMOVE64): Likewise.
11012 (mov<mode>_softfloat64, FMOVE64): Likewise.
11014 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
11016 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11018 (Basic PowerPC Built-in Functions): The new name of the
11019 subsection previously known as "PowerPC Built-in Functions".
11020 (Basic PowerPC Built-in Functions Available on all Configurations):
11022 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
11023 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
11024 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
11025 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
11027 2018-05-11 Martin Jambor <mjambor@suse.cz>
11030 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
11033 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
11036 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
11038 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
11040 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
11041 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
11042 (ix86_handle_option): Handle -mwaitpkg.
11043 * config.gcc: New header.
11044 * config/i386/cpuid.h (bit_WAITPKG): New bit.
11045 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
11046 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
11048 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11049 OPTION_MASK_ISA_WAITPKG.
11050 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
11051 (ix86_option_override_internal): Add PTA_WAITPKG.
11052 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
11053 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
11054 IX86_BUILTIN_TPAUSE.
11055 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
11056 __builtin_ia32_umwait and __builtin_ia32_tpause.
11057 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
11058 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
11059 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
11060 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
11061 UNSPECV_TPAUSE): New.
11062 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
11063 * config/i386/i386.opt: Add -mwaitpkg.
11064 * config/i386/waitpkgintrin.h: New file.
11065 * config/i386/x86intrin.h: New header.
11066 * doc/invoke.texi: Add -mwaitpkg.
11068 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
11071 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
11073 (cortex-m0): Use armv6s-m isa.
11074 (cortex-m0plus): Likewise.
11075 (cortex-m1): Likewise.
11076 (cortex-m0.small-multiply): Likewise.
11077 (cortex-m0plus.small-multiply): Likewise.
11078 (cortex-m1.small-multiply): Likewise.
11080 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
11081 Jakub Jelinek <jakub@redhat.com>
11083 PR tree-optimization/85692
11084 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
11085 source permute as well.
11087 2018-05-11 Martin Liska <mliska@suse.cz>
11090 * doc/extend.texi: Document LLVM style format for no_sanitize
11093 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11095 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
11096 mode_supports_vsx_dform_quad to mode_supports_dq_form.
11097 (mode_supports_vsx_dform_quad): Likewise.
11098 (mode_supports_vmx_dform): Move these functions to be next to the
11099 other mode_supports functions.
11100 (mode_supports_dq_form): Likewise.
11101 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
11102 mode_supports_dq_form.
11103 (reg_offset_addressing_ok_p): Likewise.
11104 (offsettable_ok_by_alignment): Likewise.
11105 (rs6000_legitimate_offset_address_p): Likewise.
11106 (legitimate_lo_sum_address_p): Likewise.
11107 (rs6000_legitimize_address): Likewise.
11108 (rs6000_legitimize_reload_address): Likewise.
11109 (rs6000_secondary_reload_inner): Likewise.
11110 (rs6000_preferred_reload_class): Likewise.
11111 (rs6000_output_move_128bit): Likewise.
11113 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
11115 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
11116 Generate SImode target register for null target.
11117 <case IX86_BUILTIN_XGETBV>: Ditto.
11118 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
11119 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
11121 2018-05-10 Carl Love <cel@us.ibm.com>
11123 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
11124 dcbtt and dcbtstt if operands[2] is 0.
11126 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
11129 * config/i386/sse.md (usadv64qi): New expander.
11131 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
11133 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
11134 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
11135 -maltivec=be support.
11136 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
11137 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
11138 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
11139 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
11140 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
11141 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
11142 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
11143 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
11144 altivec_vsumsws): Adjust.
11145 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
11146 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
11147 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
11149 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
11150 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
11151 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
11152 (altivec_lve<VI_char>x): Delete expand.
11153 (*altivec_lve<VI_char>x_internal): Rename to...
11154 (altivec_lve<VI_char>x): ... this.
11155 (altivec_lvxl_<mode>): Delete expand.
11156 (*altivec_lvxl_<mode>_internal): Rename to ...
11157 (altivec_lvxl_<mode>): ... this.
11158 (altivec_stvxl_<mode>): Delete expand.
11159 (*altivec_stvxl_<mode>_internal): Rename to ...
11160 (altivec_stvxl_<mode>): ... this.
11161 (altivec_stve<VI_char>x): Delete expand.
11162 (*altivec_stve<VI_char>x_internal): Rename to ...
11163 (altivec_stve<VI_char>x): ... this.
11164 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
11165 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
11166 reduc_plus_scal_<mode>): Adjust.
11167 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
11169 (rs6000_cpu_cpp_builtins): Adjust.
11170 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
11171 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
11172 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
11173 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
11174 -maltivec=be support.
11175 (rs6000_split_vec_extract_var): Adjust.
11176 (rs6000_split_v4si_init): Adjust.
11177 (swap_selector_for_mode): Delete.
11178 (altivec_expand_lvx_be, altivec_expand_stvx_be,
11179 altivec_expand_stvex_be): Delete.
11180 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
11181 -maltivec=be support.
11182 (rs6000_gimple_fold_builtin): Ditto.
11183 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
11185 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
11186 (TARGET_DIRECT_MOVE_64BIT): Adjust.
11187 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
11188 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
11189 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
11190 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
11191 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
11192 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
11193 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
11194 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
11195 anonymous split): Adjust.
11196 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
11197 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
11199 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
11201 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
11202 when --with-gxx-include-dir is also specified.
11203 * configure: Regenerate.
11205 2018-05-09 Jim Wilson <jimw@sifive.com>
11208 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
11209 * config/riscv/t-withmultilib: New.
11210 * config/riscv/withmultilib.h: New.
11211 * doc/install.texi: Document RISC-V --with-multilib-list support.
11213 2018-05-09 Richard Biener <rguenther@suse.de>
11215 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
11217 (vect_bb_vectorization_profitable_p): Adjust. Compute
11218 actual scalar cost using the cost vector and the add_stmt_cost
11221 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11223 PR rtl-optimization/85645
11224 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
11225 in the REG_CFA_REGISTER note for LR, don't leave it empty.
11227 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11229 PR rtl-optimization/85645
11230 * shrink-wrap.c (spread_components): Return a boolean saying if
11231 anything was changed.
11232 (try_shrink_wrapping_separate): Iterate spread_components until
11233 nothing changes anymore.
11235 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11237 PR rtl-optimization/85645
11238 * regrename.c (build_def_use): Also kill the chains that include the
11239 destination of a REG_CFA_REGISTER note.
11241 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
11243 PR rtl-optimization/85645
11244 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
11245 insn that has a REG_CFA_REGISTER note.
11247 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
11249 * cfgexpand.c (expand_clobber): New function.
11250 (expand_gimple_stmt_1): Use it.
11251 * tree-vect-stmts.c (vect_clobber_variable): New function,
11253 (vectorizable_simd_clone_call): ...here.
11254 (vectorizable_store): Emit a clobber either side of an
11255 IFN_STORE_LANES sequence.
11256 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
11258 2018-05-09 Tom de Vries <tom@codesourcery.com>
11261 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
11262 (define_insn "trap_if_false"): Add exit after trap.
11264 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
11266 PR rtl-optimization/85638
11267 * bb-reorder.c: Include common/common-target.h.
11268 (create_forwarder_block): New function extracted from...
11269 (fix_up_crossing_landing_pad): ...here. Rename into...
11270 (dw2_fix_up_crossing_landing_pad): ...this.
11271 (sjlj_fix_up_crossing_landing_pad): New function.
11272 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
11273 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
11274 from both partitions and exit the loop after one iteration.
11276 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
11279 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11281 (Basic PowerPC Built-in Functions): The new name of the
11282 subsection previously known as "PowerPC Built-in Functions".
11283 (Basic PowerPC Built-in Functions Available on all Configurations):
11285 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
11287 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
11289 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
11291 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
11294 2018-05-08 Jim Wilson <jimw@sifive.com>
11296 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
11297 (LD_EMUL_SUFFIX): New.
11298 (LINK_SPEC): Use it.
11300 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
11302 * doc/extend.texi (PowerPC Built-in Functions): Rename this
11304 (Basic PowerPC Built-in Functions): The new name of the
11305 subsection previously known as "PowerPC Built-in Functions".
11306 (Basic PowerPC Built-in Functions Available on all Configurations):
11308 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
11310 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
11312 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
11314 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
11317 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11320 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
11321 after cmpelim optimization.
11323 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
11325 * config.gcc: Support "goldmont".
11326 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
11327 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11328 PROCESSOR_GOLDMONT.
11329 * config/i386/i386.c (m_GOLDMONT): Define.
11330 (processor_target_table): Add "goldmont".
11331 (PTA_GOLDMONT): Define.
11332 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
11333 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
11334 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
11335 (fold_builtin_cpu): Add "goldmont".
11336 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
11337 (ix86_option_override_internal): Add "goldmont".
11338 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
11339 (processor_type): Add PROCESSOR_GOLDMONT.
11340 * config/i386/i386.md: Add CPU "glm".
11341 * config/i386/glm.md: New file.
11342 * config/i386/x86-tune.def: Add m_GOLDMONT.
11343 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
11345 2018-05-08 Jakub Jelinek <jakub@redhat.com>
11348 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
11350 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
11351 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
11352 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
11355 * config/i386/i386.c (ix86_fold_builtin): Handle
11356 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
11359 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
11360 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
11362 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
11365 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
11366 (check_arch): Likewise.
11367 (check_fpu): Return the result rather than printing it.
11368 (end arch): Fix operator precedence.
11369 (end cpu): Likewise.
11370 (END): Print the result from check_fpu.
11372 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
11373 Alan Hayward <alan.hayward@arm.com>
11374 David Sherwood <david.sherwood@arm.com>
11376 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
11377 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
11378 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
11379 (*fcmuo<mode>_and): New patterns.
11381 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
11383 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
11384 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
11385 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
11386 (cmp_op, sve_imm_con): New code attributes.
11387 (SVE_COND_INT_CMP, imm_con): Delete.
11388 (cmp_op): Remove above unspecs from int attribute.
11389 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
11391 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
11392 comparison-specific unspecs.
11393 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
11394 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
11395 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
11396 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
11397 (*vec_fcm<cmp_op><mode>): Rename to...
11398 (*fcm<cmp_op><mode>): ...this and adjust likewise.
11399 (*vec_fcmuo<mode>): Rename to...
11400 (*fcmuo<mode>): ...this and adjust likewise.
11401 (*pred_fcm<cmp_op><mode>): New pattern.
11402 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
11403 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
11405 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
11407 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
11408 (aarch64_emit_sve_predicated_cond): New function.
11409 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
11410 (aarch64_emit_unspec_cond_or): Replace with...
11411 (aarch64_emit_sve_or_conds): ...this new function. Use
11412 aarch64_emit_sve_ptrue_op for the individual comparisons and
11413 aarch64_emit_binop to OR them together.
11414 (aarch64_emit_inverted_unspec_cond): Replace with...
11415 (aarch64_emit_sve_inverted_cond): ...this new function. Use
11416 aarch64_emit_sve_ptrue_op for the comparison and
11417 aarch64_emit_unop to invert the result.
11418 (aarch64_expand_sve_vec_cmp_float): Update after the above
11419 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
11421 2018-05-07 Nathan Sidwell <nathan@acm.org>
11423 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
11424 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
11425 (Backwards Compatibility): Likewise.
11427 2018-05-07 Luis Machado <luis.machado@linaro.org>
11431 2018-05-07 Luis Machado <luis.machado@linaro.org>
11433 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11434 <prefetch_dynamic_strides>: New const bool field.
11435 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11436 prefetch_dynamic_strides.
11437 (exynosm1_prefetch_tune): Likewise.
11438 (thunderxt88_prefetch_tune): Likewise.
11439 (thunderx_prefetch_tune): Likewise.
11440 (thunderx2t99_prefetch_tune): Likewise.
11441 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11443 (aarch64_override_options_internal): Update to set
11444 PARAM_PREFETCH_DYNAMIC_STRIDES.
11445 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11446 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11447 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11448 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11449 prefetch-dynamic-strides setting.
11451 2018-05-07 Luis Machado <luis.machado@linaro.org>
11453 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11454 <minimum_stride>: New const int field.
11455 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11456 minimum_stride field.
11457 (exynosm1_prefetch_tune): Likewise.
11458 (thunderxt88_prefetch_tune): Likewise.
11459 (thunderx_prefetch_tune): Likewise.
11460 (thunderx2t99_prefetch_tune): Likewise.
11461 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11462 (aarch64_override_options_internal): Update to set
11463 PARAM_PREFETCH_MINIMUM_STRIDE.
11464 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11465 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11466 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11467 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11468 stride is constant and is below the minimum stride threshold.
11470 2018-05-07 Luis Machado <luis.machado@linaro.org>
11472 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
11475 2018-05-07 Luis Machado <luis.machado@linaro.org>
11477 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11478 <prefetch_dynamic_strides>: New const bool field.
11479 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11480 prefetch_dynamic_strides.
11481 (exynosm1_prefetch_tune): Likewise.
11482 (thunderxt88_prefetch_tune): Likewise.
11483 (thunderx_prefetch_tune): Likewise.
11484 (thunderx2t99_prefetch_tune): Likewise.
11485 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
11487 (aarch64_override_options_internal): Update to set
11488 PARAM_PREFETCH_DYNAMIC_STRIDES.
11489 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11490 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11491 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11492 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11493 prefetch-dynamic-strides setting.
11495 2018-05-07 Luis Machado <luis.machado@linaro.org>
11497 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11498 <minimum_stride>: New const int field.
11499 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11500 minimum_stride field.
11501 (exynosm1_prefetch_tune): Likewise.
11502 (thunderxt88_prefetch_tune): Likewise.
11503 (thunderx_prefetch_tune): Likewise.
11504 (thunderx2t99_prefetch_tune): Likewise.
11505 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
11506 (aarch64_override_options_internal): Update to set
11507 PARAM_PREFETCH_MINIMUM_STRIDE.
11508 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11509 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11510 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11511 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11512 stride is constant and is below the minimum stride threshold.
11514 2018-05-06 Jakub Jelinek <jakub@redhat.com>
11517 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
11518 the type is addressable. Don't force op into register if it has
11521 2018-05-05 Roland McGrath <mcgrathr@google.com>
11524 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
11525 any section for which we don't know a specific type it should have,
11526 regardless of name. Previously this was done only for the exact
11527 names ".init_array", ".fini_array", and ".preinit_array".
11528 (default_elf_asm_named_section): Add comment about
11529 relationship with default_section_type_flags and SECTION_NOTYPE.
11530 (get_section): Don't consider it a type conflict if one side has
11531 SECTION_NOTYPE and the other doesn't, as long as neither has the
11532 SECTION_BSS et al used in the default_section_type_flags logic.
11534 2018-05-05 Tom de Vries <tom@codesourcery.com>
11537 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
11538 (workaround_barsyncs): New function.
11539 (nvptx_reorg): Use workaround_barsyncs.
11540 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11541 (define_expand "nvptx_membar_cta"): New define_expand.
11542 (define_insn "*nvptx_membar_cta"): New insn.
11544 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
11546 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
11547 To improve optimization opportunities.
11548 * builtin-types.def: The new needed builtin types for the above.
11550 2018-05-04 Richard Biener <rguenther@suse.de>
11552 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
11553 * gimple-ssa-store-merging.c
11554 (imm_store_chain_info::output_merged_store): Remove redundant create,
11555 release split_store vector contents on failure.
11556 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
11557 scalar stmt vector on cache hit.
11559 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
11561 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
11563 * config.gcc (powerpc-xilinx-eabi*): Remove.
11564 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
11566 (fusion_addis_mem_combo_load): Ditto.
11567 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
11569 (rs6000_cpu_cpp_builtins): Ditto.
11570 * config/rs6000/rs6000-linux.c
11571 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
11572 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
11573 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
11575 (rs6000_setup_reg_addr_masks): Ditto.
11576 (rs6000_init_hard_regno_mode_ok): Ditto.
11577 (rs6000_option_override_internal): Ditto.
11578 (legitimate_lo_sum_address_p): Ditto.
11579 (rs6000_legitimize_address): Ditto.
11580 (rs6000_legitimize_reload_address): Ditto.
11581 (rs6000_legitimate_address_p): Ditto.
11582 (abi_v4_pass_in_fpr): Ditto.
11583 (setup_incoming_varargs): Ditto.
11584 (rs6000_gimplify_va_arg): Ditto.
11585 (rs6000_split_multireg_move): Ditto.
11586 (rs6000_savres_strategy): Ditto.
11587 (rs6000_emit_prologue_components): Ditto.
11588 (rs6000_emit_epilogue_components): Ditto.
11589 (rs6000_emit_prologue): Ditto.
11590 (rs6000_emit_epilogue): Ditto.
11591 (rs6000_elf_file_end): Ditto.
11592 (rs6000_function_value): Ditto.
11593 (rs6000_libcall_value): Ditto.
11594 * config/rs6000/rs6000.h: Ditto.
11595 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
11596 (TARGET_MINMAX): ... this. New.
11597 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
11598 * config/rs6000/rs6000.md: Remove Xilinx FP support.
11599 (*movsi_internal1_single): Delete.
11600 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
11601 mfpu=, mxilinx-fpu): Delete.
11602 * config/rs6000/singlefp.h: Delete.
11603 * config/rs6000/sysv4.h: Remove Xilinx FP support.
11604 * config/rs6000/t-rs6000: Ditto.
11605 * config/rs6000/t-xilinx: Delete.
11606 * config/rs6000/titan.md: Adjust for fp_type removal.
11607 * config/rs6000/vsx.md: Remove Xilinx FP support.
11608 (VStype_simple): Delete.
11609 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
11610 * config/rs6000/xfpu.h: Delete.
11611 * config/rs6000/xfpu.md: Delete.
11612 * config/rs6000/xilinx.h: Delete.
11613 * config/rs6000/xilinx.opt: Delete.
11614 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
11615 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
11617 2018-05-04 Tom de Vries <tom@codesourcery.com>
11620 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
11623 2018-05-04 Richard Biener <rguenther@suse.de>
11625 PR middle-end/85627
11626 * tree-complex.c (update_complex_assignment): We are always in SSA form.
11627 (expand_complex_div_wide): Likewise.
11628 (expand_complex_operations_1): Likewise.
11629 (expand_complex_libcall): Preserve EH info of the original stmt.
11630 (tree_lower_complex): Handle removed blocks.
11631 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
11632 on complex multiplication and division libcall builtins.
11634 2018-05-04 Richard Biener <rguenther@suse.de>
11636 PR middle-end/85574
11637 * fold-const.c (negate_expr_p): Restrict negation of operand
11638 zero of a division to when we know that can happen without
11640 (fold_negate_expr_1): Likewise.
11642 2018-05-04 Jakub Jelinek <jakub@redhat.com>
11645 * real.h (real_nextafter): Declare.
11646 * real.c (real_nextafter): New function.
11647 * fold-const-call.c (fold_const_nextafter): New function.
11648 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
11649 CASE_CFN_NEXTTOWARD.
11650 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
11651 even when arg1_mode is different from arg0_mode.
11653 2018-05-03 Nathan Sidwell <nathan@acm.org>
11655 * doc/extend.texi (Deprecated Features): Remove
11656 -ffriend-injection.
11657 (Backwards Compatibility): Likewise.
11658 * doc/invoke.texi (C++ Language Options): Likewise.
11659 (C++ Dialect Options): Likewise.
11661 2018-05-03 Jakub Jelinek <jakub@redhat.com>
11664 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
11665 _mm512_mask_mullox_epi64): New intrinsics.
11667 2018-05-03 Tom de Vries <tom@codesourcery.com>
11670 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11671 dump files): Add offload-tree.
11673 2018-05-03 Richard Biener <rguenther@suse.de>
11675 PR tree-optimization/85615
11676 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
11677 to loops not nested in BBs loop father to avoid creating multi-entry
11680 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11682 PR tree-optimization/70291
11683 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
11684 arguments. Change return type to tree. Emit libcall as a new
11685 statement rather than replacing existing one when inplace_p is true.
11686 (expand_complex_multiplication_components): New function.
11687 (expand_complex_multiplication): Expand floating-point complex
11688 multiplication using the above.
11689 (expand_complex_division): Rename inner_type parameter to type.
11690 Update expand_complex_libcall call-site.
11691 (expand_complex_operations_1): Update expand_complex_multiplication
11692 and expand_complex_division call-sites.
11694 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11697 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11698 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11699 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
11700 the highest significant bit of the shift count mask is clear. In
11701 check whether and[sq]i3 is needed verify that all significant bits
11702 of the shift count other than the highest are set.
11704 2018-05-02 Tom de Vries <tom@codesourcery.com>
11707 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
11708 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
11709 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
11710 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
11711 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
11712 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
11713 __builtin_goacc_parlevel_size.
11715 2018-05-02 Richard Biener <rguenther@suse.de>
11717 PR tree-optimization/85597
11718 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
11719 do not use split vect_get_vec_defs call but call vect_get_slp_defs
11722 2018-05-02 Tom de Vries <tom@codesourcery.com>
11725 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11726 dump files): Add ltrans-tree.
11728 2018-05-02 Tom de Vries <tom@codesourcery.com>
11731 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11732 dump files): Add wpa-ipa.
11734 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
11736 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
11737 powerpc*-*-linux*paired* target.
11738 * config/rs6000/750cl.h: Delete.
11739 * config/rs6000/paired.h: Delete.
11740 * config/rs6000/paired.md: Delete.
11741 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
11743 * config/rs6000/rs6000-builtin.def: Remove paired float support.
11744 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
11745 comment. Remove paired float support.
11746 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
11747 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
11749 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
11750 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
11752 * config/rs6000/rs6000.c: Remove paired float support.
11753 (paired_expand_vector_init, paired_expand_vector_move,
11754 paired_emit_vector_compare, paired_emit_vector_cond_expr,
11755 (paired_expand_lv_builtin, paired_expand_stv_builtin,
11756 paired_expand_builtin, paired_expand_predicate_builtin,
11757 paired_init_builtins): Delete.
11758 * config/rs6000/rs6000.h: Remove paired float support.
11759 * config/rs6000/rs6000.md: Remove paired float support.
11760 (move_from_CR_ov_bit): Delete.
11761 * config/rs6000/rs6000.opt (mpaired): Delete.
11762 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
11763 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
11765 2018-05-02 Richard Biener <rguenther@suse.de>
11767 PR middle-end/85567
11768 * gimplify.c (gimplify_save_expr): When in SSA form allow
11769 SAVE_EXPRs to compute to SSA vars.
11771 2018-05-02 Jakub Jelinek <jakub@redhat.com>
11774 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
11775 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
11776 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
11777 clobber operands[2], instead use a new pseudo. Formatting fixes.
11779 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
11781 PR tree-optimization/85586
11782 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
11783 exit early for statements in the same group if the accesses are
11786 2018-05-02 Tom de Vries <tom@codesourcery.com>
11789 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
11792 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
11794 PR tree-optimization/85143
11795 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
11797 2018-05-01 Tom de Vries <tom@codesourcery.com>
11800 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
11801 not found" error message.
11803 2018-05-01 Tom de Vries <tom@codesourcery.com>
11806 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
11807 * vec.c (test_ordered_remove_if): New function.
11808 (vec_c_tests): Call test_ordered_remove_if.
11809 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
11810 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
11811 * tree-vect-patterns.c (vect_pattern_recog_1): Use
11812 VEC_ORDERED_REMOVE_IF.
11814 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11816 PR tree-optimization/82665
11817 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
11818 pointer subtraction where arguments come from a memchr call.
11820 2018-05-01 Jakub Jelinek <jakub@redhat.com>
11822 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
11823 --push-state --as-needed and --pop-state instead of --as-needed and
11824 --no-as-needed if ld supports it.
11825 * configure: Regenerated.
11828 * doc/install.texi2html: Replace _002d with - and _002a with * in
11829 generated html files using sed.
11831 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11834 * gcc-rich-location.c (blank_line_before_p): New function.
11835 (use_new_line): New function.
11836 (gcc_rich_location::add_fixit_insert_formatted): New function.
11837 * gcc-rich-location.h
11838 (gcc_rich_location::add_fixit_insert_formatted): New function.
11840 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11842 * selftest.c (assert_streq): Rename "expected" and "actual" to
11843 "val1" and "val2". Extend NULL-handling to cover both inputs
11844 symmetrically, while still requiring both to be non-NULL for a pass.
11845 * selftest.h (assert_streq): Rename "expected" and "actual" to
11847 (ASSERT_EQ): Likewise.
11848 (ASSERT_EQ_AT): Likewise.
11849 (ASSERT_KNOWN_EQ): Likewise.
11850 (ASSERT_KNOWN_EQ_AT): Likewise.
11851 (ASSERT_NE): Likewise.
11852 (ASSERT_MAYBE_NE): Likewise.
11853 (ASSERT_MAYBE_NE_AT): Likewise.
11854 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
11855 the assertion to pass.
11856 (ASSERT_STREQ_AT): Likewise.
11858 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11860 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
11861 interaction with -pie.
11863 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11865 * selftest.h: Fix alphabetization of per-source-file selftest
11868 2018-04-30 Jason Merrill <jason@redhat.com>
11870 PR c++/61982 - dead stores to destroyed objects.
11871 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
11874 2018-04-30 Jason Merrill <jason@redhat.com>
11876 * tree.c (build_clobber): New.
11877 * tree.h: Declare it.
11878 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
11880 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11882 * diagnostic-show-locus.c (layout::layout): Update for
11883 location_get_source_line returning a char_span.
11884 (struct char_span): Move to input.h.
11885 (struct correction): Update for fields in char_span becoming
11887 (struct source_line): Update for location_get_source_line
11888 returning a char_span.
11889 (layout::print_line): Likewise.
11890 * edit-context.c (edited_file::print_content): Likewise.
11891 (edited_file::print_diff_hunk): Likewise.
11892 (edited_file::print_run_of_changed_lines): Likewise.
11893 (edited_file::get_num_lines): Likewise.
11894 (edited_line::edited_line): Likewise.
11895 * final.c (asm_show_source): Likewise.
11896 * input.c (location_get_source_line): Convert return type
11897 from const char * to char_span, losing the final "line_len"
11899 (dump_location_info): Update for the above.
11900 (get_substring_ranges_for_loc): Likewise. Use a char_span
11901 when handling the literal within the line.
11902 (test_reading_source_line): Update for location_get_source_line
11903 returning a char_span.
11904 * input.h (class char_span): Move here from
11905 diagnostic-show-locus.c, converting from a struct to a class.
11906 Make data members private.
11907 (char_span::operator bool): New.
11908 (char_span::length): New.
11909 (char_span::get_buffer): New.
11910 (char_span::operator[]): New.
11911 (char_span::subspan): Make const.
11912 (char_span::xstrdup): New.
11913 (location_get_source_line): Convert return type from const char *
11914 to char_span, losing the final "line_size" param.
11916 2018-04-30 Jan Hubicka <jh@suse.cz>
11918 * lto-wrapper.c (ltrans_priorities): New static var.
11919 (cmp_priority): New.
11920 (run_gcc): Read priorities and if doing parallel build order
11921 the Makefile by them.
11923 2018-04-30 David Malcolm <dmalcolm@redhat.com>
11925 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
11927 2018-04-30 Richard Biener <rguenther@suse.de>
11929 * tree-cfg.c (verify_address): Remove base argument, add
11930 flag whether to check TREE_ADDRESSABLE and do that.
11931 (verify_expr): Remove.
11932 (verify_types_in_gimple_reference): Add pieces from verify_expr.
11933 (verify_gimple_assign_single): Likewise.
11934 (verify_gimple_switch): Likewise.
11935 (verify_expr_location_1): Dereference tp once. Add (disabled)
11936 piece from verify_expr.
11937 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
11939 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11941 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
11943 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11945 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
11946 (small_data_pattern): Likewise.
11947 (arc_rewrite_small_data): Likewise.
11948 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
11949 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
11950 (get_symbol_alignment): New function.
11951 (legitimate_small_data_address_p): Likewise.
11952 (legitimate_scaled_address): Update, call
11953 legitimate_small_data_address_p.
11954 (output_sdata): New static variable.
11955 (arc_print_operand): Update how we handle small data operands.
11956 (arc_print_operand_address): Likewise.
11957 (arc_legitimate_address_p): Update, use
11958 legitimate_small_data_address_p.
11959 (arc_rewrite_small_data_p): Remove.
11960 (arc_rewrite_small_data_1): Likewise.
11961 (arc_rewrite_small_data): Likewise.
11962 (small_data_pattern): Likewise.
11963 (compact_sda_memory_operand): Update to use
11964 legitimate_small_data_address_p and get_symbol_alignment.
11965 (prepare_move_operands): Don't rewite sdata pattern.
11966 (prepare_extend_operands): Remove.
11967 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
11969 (zero_extendqisi2): Likewise.
11970 (zero_extendhisi2): Likewise.
11971 (extendqihi2): Likewise.
11972 (extendqisi2): Likewise.
11973 (extendhisi2): Likewise.
11974 (addsi3): Likewise.
11975 (subsi3): Likewise.
11976 (andsi3): Likewise.
11977 * config/arc/constraints.md (Usd): Change it to memory constraint.
11979 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
11981 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
11982 as source of std instructions.
11983 * config/arc/arc.md (movsi_insn): Update pattern predicate to
11984 allow 6-bit constants as source for store instructions.
11985 (movdi_insn): Update instruction pattern to allow 6-bit constants
11986 as source for store instructions.
11988 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
11990 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
11992 2018-04-30 Nathan Sidwell <nathan@acm.org>
11993 Sandra Loosemore <sandra@codesourcery.com>
11995 * dumpfile.c (dump_open): Allow '-' for stdout.
11996 * doc/invoke.texi (Developer Options): Document dump filename
11997 determination early. Document stdin/stdout selection.
11999 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
12001 Microblaze Target: PIC data text relative
12003 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
12004 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
12006 * config/microblaze/microblaze.h (microblaze_constant_address_p):
12007 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
12008 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
12009 New addressing mode for data-text relative position indepenedent code.
12010 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
12011 'ADDRESS_SYMBOLIC_TXT_REL'.
12012 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
12013 (microblaze_legitimate_pic_operand): Exclude function calls from
12014 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
12015 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
12017 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
12018 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
12019 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
12020 for 'address + offset'.
12021 (microblaze_expand_prologue): Add new function prologue call for
12023 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
12024 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
12025 table in case of TARGET_PIC_DATA_TEXT_REL.
12026 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
12027 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
12028 Add new macros 'UNSPEC_TEXT',
12029 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
12030 + exclude function calls from 'UNSPEC_PLT' in case of data text
12032 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
12033 new target hook for generating address diff vector tables in case of
12035 * doc/tm.texi : Regenerate.
12036 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
12037 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
12038 of addr diff vector generation.
12039 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
12040 target hook definition.
12041 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
12042 Add default function for generate_pic_addr_diff_vec -> flag_pic.
12043 * doc/invoke.texi (Add new pic option): Add new microblaze pic
12044 option for data text relative.
12046 2018-04-30 Richard Biener <rguenther@suse.de>
12048 * tree-chrec.h (evolution_function_is_constant_p): Remove
12050 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
12052 2018-04-30 Richard Biener <rguenther@suse.de>
12055 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
12057 2018-04-30 Richard Biener <rguenther@suse.de>
12059 PR tree-optimization/28364
12060 PR tree-optimization/85275
12061 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
12062 copying first exit test.
12064 2018-04-28 Mark Wielaard <mark@klomp.org>
12066 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
12067 dwarf_version >= 5.
12068 (dwarf_AT): Handle DW_AT_addr_base.
12069 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
12071 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
12074 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
12075 (*ashl<dwi>3_doubleword_mask_1): Ditto.
12076 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
12077 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
12079 2018-04-28 Richard Biener <rguenther@suse.de>
12081 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
12082 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
12083 to reflect use. Only add interesting stmts.
12085 2018-04-27 Martin Jambor <mjambor@suse.cz>
12088 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
12089 the jump function allows for passing through aggregate values.
12091 2018-04-27 David Malcolm <dmalcolm@redhat.com>
12093 * input.h (in_system_header_at): Convert from macro to inline
12095 (from_macro_expansion_at): Likewise.
12096 (from_macro_definition_at): Likewise.
12098 2018-04-27 Jeff Law <law@redhat.com>
12100 * config.gcc: Mark tile* targets as deprecated/obsolete.
12102 2018-04-27 Richard Biener <rguenther@suse.de>
12104 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
12107 2018-04-27 Richard Biener <rguenther@suse.de>
12109 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
12111 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
12113 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
12114 with Yd constraint. Set "preferred_for_speed" attribute from
12115 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
12116 with Yd constraint.
12117 (*movdi_internal): Ditto.
12118 (movti_interunit splitters): Remove
12119 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
12120 (movdi_interunit splitters): Ditto.
12121 * config/i386/constraints.md (Ye): Remove.
12122 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
12124 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12127 * config/aarch64/constraints.md (Usg): Limit to 31.
12128 (Usj): Limit to 63.
12130 2018-04-27 Jakub Jelinek <jakub@redhat.com>
12132 PR tree-optimization/85529
12133 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
12134 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
12135 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
12136 zero extension or masking of the MSB bit.
12137 (optimize_range_tests): Add FIRST_BB argument, pass it through
12138 to optimize_range_tests_var_bound.
12139 (maybe_optimize_range_tests, reassociate_bb): Adjust
12140 optimize_range_tests callers.
12142 2018-04-26 Richard Biener <rguenther@suse.de>
12143 Jakub Jelinek <jakub@redhat.com>
12145 * cgraph.h (symbol_table): Just declare debug method here.
12146 * symtab.c (symbol_table::debug): Define.
12148 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
12150 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
12152 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
12154 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
12155 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
12156 (*movdi_internal): Substitute Yi and Yj constraint with x
12157 and Ym and Yn constraint with y constraint. Update "isa"
12158 attribute and set "preferred_for_speed" attribute from
12159 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
12160 (*movsi_internal): Ditto.
12161 (*movdf_internal): Ditto.
12162 (*movsf_internal): Ditto.
12163 (*zero_extendsidi2): Ditto.
12164 * config/i386/sse.md (vec_set<mode>_0): Ditto.
12165 (sse2_loadld): Ditto.
12166 (*vec_extract<ssevecmodelower>_0): Ditto.
12167 (*vec_extractv4si_0_zext_sse4): Ditto.
12168 (vec_concatv2di): Ditto.
12169 (*vec_dup<mode>): Ditto.
12170 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
12171 * config/i386/constraints.md (Yi): Remove.
12176 2018-04-26 Nathan Sidwell <nathan@acm.org>
12178 * dumpfile.c (dump_open): New.
12179 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
12180 (dump_finish): Detect stdio/stderr by value not name.
12182 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
12184 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
12186 2018-04-26 Tom de Vries <tom@codesourcery.com>
12189 * config/nvptx/nvptx.c (verify_neutering_jumps)
12190 (verify_neutering_labels): New function
12191 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
12193 2018-04-26 Tom de Vries <tom@codesourcery.com>
12196 * config/nvptx/nvptx.c (needs_neutering_p): New function.
12197 (nvptx_single): Use needs_neutering_p to skip over insns that do not
12200 2018-04-26 Richard Biener <rguenther@suse.de>
12201 Tom de Vries <tom@codesourcery.com>
12204 * lto-streamer-out.c (output_function): Fixup loops if required to match
12205 discovery done in the reader.
12207 2018-04-26 Richard Biener <rguenther@suse.de>
12209 PR tree-optimization/85116
12210 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
12211 have a loop exit from the single latch predecessor. Remove
12212 case of header with just condition.
12213 (ch_base::copy_headers): Exclude infinite loops from any
12215 (pass_ch::execute): Record exits.
12217 2018-04-26 Richard Biener <rguenther@suse.de>
12219 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
12220 prologue cost vector and pass it to vect_get_load_cost.
12221 (vect_get_peeling_costs_all_drs): Likewise.
12222 (vect_peeling_hash_get_lowest_cost): Likewise.
12223 (vect_enhance_data_refs_alignment): Likewise.
12225 2018-04-26 Richard Biener <rguenther@suse.de>
12227 PR middle-end/85450
12228 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
12229 checking of integer<->pointer conversions.
12230 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
12231 sign-/zero-extending pointer types.
12232 (expand_omp_for_static_chunk): Likewise.
12234 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
12235 Jean Lee <xiaoyur347@gmail.com>
12237 * config/mips/mips.c (mips_asan_shadow_offset): New function.
12238 (TARGET_ASAN_SHADOW_OFFSET): Define.
12239 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
12240 true for -fsanitize=address.
12242 2018-04-25 Mark Wielaard <mark@klomp.org>
12244 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
12247 2018-04-25 Jakub Jelinek <jakub@redhat.com>
12249 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
12250 than "alu", remove explicit "memory" and "imm_disp" attributes.
12251 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
12253 PR middle-end/85414
12254 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
12255 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
12256 gen_lowpart_no_emit.
12258 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
12261 * config/i386/i386.c (ix86_expand_builtin): Change memory
12262 operand to XI, extend p0 to Pmode.
12263 * config/i386/i386.md: Change unspec volatile and operand
12264 1 mode to XI, change operand 0 mode to P.
12266 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12268 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
12269 GET_MODE_MASK before any checking.
12270 (nds32_can_use_bset_p): Likewise.
12271 (nds32_can_use_btgl_p): Likewise.
12273 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12275 * config/nds32/nds32-doubleword.md: New define_split pattern for
12276 illegal register number.
12278 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12280 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
12282 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
12284 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
12286 2018-04-25 Richard Biener <rguenther@suse.de>
12288 * lto-streamer.h (LTO_major_version): Bump to 8.
12290 2018-04-25 Jakub Jelinek <jakub@redhat.com>
12292 * BASE-VER: Set to 9.0.0.
12294 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
12296 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
12297 in __abskf2 and __powikf2.
12299 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12302 * config/aarch64/constraints.md (Usg, Usj): New constraints.
12303 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
12304 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
12305 Use the above on operand 2. Reindent.
12306 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
12308 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
12311 * common/config/i386/i386-common.c (ix86_handle_option): Don't
12313 * config/i386/i386.opt (mcet): Removed.
12314 * doc/install.texi: Remove -mcet documentation.
12315 * doc/invoke.texi: Likewise.
12317 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
12320 * doc/install.texi: Remove -mcet from bootstrap-cet.
12322 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12325 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
12326 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
12330 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
12331 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
12332 containing a CONST_VECTOR.
12334 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
12336 * doc/install.texi: Update newlib dependency for nvptx.
12338 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12341 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
12342 instead of INTVAL when shifting x left.
12344 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
12346 PR tree-optimization/85478
12347 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
12348 vect_grouped_store_supported for single element vectors.
12350 2018-04-24 Richard Biener <rguenther@suse.de>
12353 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
12354 load cost increase to the case of non-constant step.
12356 2018-04-24 Jakub Jelinek <jakub@redhat.com>
12359 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
12360 destination if any_malformed_asm.
12362 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
12364 PR middle-end/85496
12365 * expr.c (store_field): In the bitfield case, if the value comes from
12366 a function call and is returned in registers by means of a PARALLEL,
12367 do not change the mode of the temporary unless BLKmode and VOIDmode.
12369 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
12371 PR rtl-optimization/85423
12372 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
12373 dependencies to debug insns when the previous insn is non-debug.
12375 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
12377 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
12378 enums into a single definition.
12379 (fls): Fix predicates and printing.
12382 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
12384 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
12385 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
12386 and short u6 immediate.
12387 (check_if_valid_sleep_operand): Remove.
12388 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
12390 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12392 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
12393 flag_always_save_lp condition.
12394 * config/nds32/nds32.opt (malways-save-lp): New option.
12396 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12398 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
12399 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
12400 * config/nds32/nds32.h
12401 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
12402 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
12404 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12406 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
12407 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
12409 2018-04-22 Shiva Chen <shiva0217@gmail.com>
12410 Chung-Ju Wu <jasonwucj@gmail.com>
12412 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
12414 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
12415 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
12417 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12419 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
12421 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12423 * config/nds32/nds32-protos.h (nds32_data_alignment,
12424 nds32_local_alignment): Declare.
12425 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
12426 nds32_local_alignment): New functions.
12427 (TARGET_CONSTANT_ALIGNMENT): Define.
12428 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
12430 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12432 * config/nds32/nds32.c
12433 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
12434 (TARGET_MODES_TIEABLE_P): Likewise.
12436 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
12438 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
12439 level Ofast and Og.
12441 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
12442 Chung-Ju Wu <jasonwucj@gmail.com>
12444 * config/nds32/constants.md (unspec_volatile_element): Add enum values
12445 for unaligned access.
12446 * config/nds32/nds32-intrinsic.c: Implementation of expanding
12448 * config/nds32/nds32-intrinsic.md: Likewise.
12449 * config/nds32/nds32_intrinsic.h: Likewise.
12450 * config/nds32/nds32.h (nds32_builtins): Likewise.
12451 * config/nds32/nds32.opt (munaligned-access): New option.
12452 * config/nds32/nds32.c (nds32_asm_file_start): Display
12453 flag_unaligned_access status.
12455 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
12457 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
12458 -mno-relax is present.
12459 * config/riscv/linux.h (LINK_SPEC): Ditto.
12461 2018-04-20 Martin Sebor <msebor@redhat.com>
12464 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
12466 (gimple_fold_builtin_stxcpy_chk): Same.
12467 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
12469 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
12472 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
12473 __powikf2 when long double is IEEE 128-bit.
12475 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
12477 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
12478 step to make sure stack always aligned.
12480 2018-04-20 Carl Love <cel@us.ibm.com>
12483 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
12484 size check for arg0.
12486 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
12487 Tom de Vries <tom@codesourcery.com>
12490 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
12491 Emit insns for calls too.
12492 (nvptx_find_par): Always look for worker-level predecessor insn.
12493 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
12495 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
12496 (nvptx_process_pars): Propagate frames for calls.
12498 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
12501 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
12503 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
12504 (ix86_handle_option): Don't handle OPT_mibt.
12505 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
12507 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
12509 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
12510 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
12511 (ix86_target_macros): Define __CET__ with flag_cf_protection
12512 for -fcf-protection.
12513 * config/i386/i386.c (isa2_opts): Remove -mibt.
12514 * config/i386/i386.h (TARGET_IBT): Removed.
12515 (TARGET_IBT_P): Likewise.
12516 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
12517 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
12518 * config/i386/i386.opt (mcet): Update help message.
12519 (mshstk): Likewise.
12521 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
12522 -mcet as an alias for -mshstk.
12524 2018-04-20 Richard Biener <rguenther@suse.de>
12526 PR middle-end/85475
12527 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
12528 complexity by forcing a single use of the multiply operand.
12530 2018-04-20 Martin Jambor <mjambor@suse.cz>
12533 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
12534 recursion dependency to only apply to non-clones.
12536 2018-04-20 Martin Jambor <mjambor@suse.cz>
12539 * ipa-cp.c (create_specialized_node): Check that clones of
12540 self-recursive edges exist during IPA-CP.
12542 2018-04-19 Toon Moene <toon@moene.org>
12544 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
12547 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12549 PR tree-optimization/85467
12550 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
12551 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
12552 VECTOR_CST element to type.
12554 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12557 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
12558 * config/i386/i386.md (builtin_setjmp_setup): Removed.
12559 (builtin_longjmp): Likewise.
12560 (save_stack_nonlocal): New pattern.
12561 (restore_stack_nonlocal): Likewise.
12563 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12566 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12567 Replace ASM_OUTPUT_LABEL with fprintf.
12569 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
12572 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
12573 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
12574 * config/i386/i386-c.c (ix86_target_macros_internal): Also
12575 define __IBT__ and __SHSTK__ for -fcf-protection.
12576 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
12578 (ix86_trampoline_init): Likewise.
12579 (x86_output_mi_thunk): Likewise.
12580 (ix86_notrack_prefixed_insn_p): Likewise.
12581 (ix86_option_override_internal): Don't disallow -fcf-protection.
12582 * config/i386/i386.md (rdssp<mode>): Also enable for
12584 (incssp<mode>): Likewise.
12585 (nop_endbr): Likewise.
12586 * config/i386/i386.opt (mcet): Change help message to built-in
12589 (mshstk): Likewise.
12590 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
12591 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
12592 enable CET built-in functions.
12594 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
12596 * common/config/i386/i386-common.c
12597 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
12598 OPTION_MASK_ISA_MOVDIRI_UNSET,
12599 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
12600 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
12601 * config.gcc (movdirintrin.h): New header.
12602 * config/i386/cpuid.h (bit_MOVDIRI,
12603 bit_MOVDIR64B): New bits.
12604 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
12606 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
12607 (VOID, PVOID, PCVOID)): New function types.
12608 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
12609 __builtin_ia32_directstoreu_u64,
12610 __builtin_ia32_movdir64b): New builtins.
12611 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
12612 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
12614 (ix86_valid_target_attribute_inner_p): Ditto.
12615 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
12616 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
12617 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
12618 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
12619 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
12620 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
12621 (movdiri<mode>, movdir64b_<mode>): New.
12622 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
12623 * config/i386/immintrin.h: Include movdirintrin.h.
12624 * config/i386/movdirintrin.h: New file.
12625 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
12627 2018-04-19 Richard Biener <rguenther@suse.de>
12629 PR middle-end/85455
12630 * cfg.c (clear_bb_flags): When loop state says we have
12631 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
12633 2018-04-19 Richard Biener <rguenther@suse.de>
12635 PR tree-optimization/84737
12636 * tree-vect-data-refs.c (vect_copy_ref_info): New function
12637 copying restrict info.
12638 (vect_setup_realignment): Use it.
12639 * tree-vectorizer.h (vect_copy_ref_info): Declare.
12640 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
12641 the first DR to all generated stores.
12642 (vectorizable_load): Likewise for loads.
12644 2018-04-19 Jakub Jelinek <jakub@redhat.com>
12646 PR tree-optimization/85446
12647 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
12648 the integral and pointer types to have the same precision.
12650 * doc/install.texi: Document --disable-cet being the default and
12653 2018-04-18 Martin Liska <mliska@suse.cz>
12655 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
12658 2018-04-18 Martin Liska <mliska@suse.cz>
12661 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12664 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12665 arguments if they are comparable.
12667 2018-04-18 Martin Liska <mliska@suse.cz>
12670 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12673 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12676 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
12679 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
12680 ENDBR after calling __morestack.
12682 2018-04-18 David Malcolm <dmalcolm@redhat.com>
12685 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
12686 by using gcc_base_ver to generate a gcc_driver_version, and use
12687 it when generating GCC_DRIVER_NAME.
12688 * configure: Regenerate.
12690 2018-04-18 Jakub Jelinek <jakub@redhat.com>
12693 * config.gcc: Obsolete powerpc*-*-*spe*.
12695 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12698 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
12699 (stabstr_D): Change type of unum from unsigned int to
12700 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
12703 2018-04-17 Jim Wilson <jimw@sifive.com>
12706 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
12707 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
12708 Set arg_pointer_offset after using pretend_args_size.
12710 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12712 PR rtl-optimization/85431
12713 * dse.c (record_store): Ignore zero width stores.
12716 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
12717 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
12718 __builtin_stack_restore rather than after it.
12719 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
12720 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
12721 argument instead of virtual_dynamic_stack_rtx.
12723 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
12725 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
12727 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
12728 Add note to error message to explain internal mapping of overloaded
12729 built-in function name to non-overloaded built-in function name.
12730 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
12733 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12736 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
12737 where the inputs overlap with the output.
12739 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12741 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
12742 (=v, v) alternative and explicit "memory" attribute.
12743 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
12744 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12746 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
12747 "sselog1" type instead of "sselog".
12748 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
12749 "sselog". Remove explicit "memory" attribute.
12750 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
12751 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12753 (vec_extract_hi_v32hi): Merge all alternatives into one, use
12754 "sselog1" type instead of "sselog". Remove explicit "memory"
12756 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
12757 use "sselog1" type instead of "sselog". Remove explicit "memory"
12759 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
12760 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
12762 (vec_extract_hi_v64qi): Merge all alternatives into one, use
12763 "sselog1" type instead of "sselog". Remove explicit "memory"
12765 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
12766 use "sselog1" type instead of "sselog". Remove explicit "memory"
12770 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
12772 PR middle-end/85414
12773 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
12776 2018-04-17 Martin Jambor <mjambor@suse.cz>
12779 * ipa-cp.c (create_specialized_node): Call
12780 expand_all_artificial_thunks if necessary.
12782 2018-04-17 Martin Liska <mliska@suse.cz>
12785 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
12786 in message, remote space in between '_G' and '('.
12788 2018-04-17 Jakub Jelinek <jakub@redhat.com>
12791 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
12792 avx512f_vmcmp<mode>3<round_saeonly_name>,
12793 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
12794 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
12795 avx512f_rndscale<mode><round_saeonly_name>,
12796 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
12797 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
12798 Use %<iptr>2 instead of %2 for -masm=intel.
12799 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
12800 avx512f_vcvttss2usi<round_saeonly_name>,
12801 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
12803 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
12804 avx512f_vcvttsd2usi<round_saeonly_name>,
12805 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
12806 Use %q1 instead of %1 for -masm=intel.
12807 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
12808 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
12809 of %3 for -masm=intel.
12810 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
12812 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
12814 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
12816 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
12817 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
12819 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
12821 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
12822 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
12823 %g1 and one with %0 and %1.
12824 (avx512er_vmrcp28<mode><round_saeonly_name>,
12825 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
12826 %1 for -masm=intel.
12827 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
12828 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
12829 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
12830 of %0 and %{%4%} for -masm=intel.
12831 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
12832 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
12833 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
12834 order of %0 and %{%5%}%{z%} for -masm=intel.
12836 2018-04-17 Jan Hubicka <jh@suse.cz>
12839 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
12841 2018-04-17 Martin Liska <mliska@suse.cz>
12844 * multiple_target.c (create_dispatcher_calls): Set apostrophes
12845 for target_clone error message. Make default implementation
12846 clone to be a local declaration.
12847 (separate_attrs): Add new argument and check for an empty
12849 (expand_target_clones): Handle it.
12850 (ipa_target_clone): Make redirection just for target_clones
12853 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
12854 Tom de Vries <tom@codesourcery.com>
12856 PR middle-end/84955
12857 * omp-expand.c (expand_oacc_for): Add dummy false branch for
12858 tiled basic blocks without omp continue statements.
12860 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
12863 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
12864 vec_extract expression as having side effects to make sure it gets
12867 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
12870 * config/i386/i386.c (get_builtin_code_for_version): Check
12873 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
12876 * config.gcc: Support "skylake".
12877 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12879 * config/i386/i386.c (m_SKYLAKE): Define.
12880 (processor_target_table): Add "skylake".
12881 (ix86_option_override_internal): Add "skylake".
12882 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
12883 PROCESSOR_CANNONLAKE.
12884 (get_builtin_code_for_version): Fix priority for
12885 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
12886 PROCESSOR_SKYLAKE-AVX512.
12887 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
12888 (processor_type): Add PROCESSOR_SKYLAKE.
12890 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
12891 Jason Merrill <jason@redhat.com>
12894 * convert.c (convert_to_integer_1): Use direct recursion for
12895 enumeral types and types with a precision less than the number
12896 of bits in their mode.
12898 2018-04-16 Julia Koval <julia.koval@intel.com>
12901 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
12902 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
12904 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
12907 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
12908 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
12909 and -mno-direct-move.
12911 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
12914 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
12915 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
12916 Ensure negative shifts result in {0}.
12918 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
12920 PR rtl-optimization/79916
12921 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
12922 regs (if any) to define how to gnerate SD moves when LRA is in
12925 2018-04-13 Jakub Jelinek <jakub@redhat.com>
12927 PR rtl-optimization/85393
12928 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
12929 * except.c (expand_dw2_landing_pad_for_region): Make static.
12930 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
12931 a label and unconditional jump to old_bb, rather than
12932 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
12935 PR rtl-optimization/85376
12936 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
12937 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
12938 instead of a specific value.
12940 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12941 Bin Cheng <bin.cheng@arm.com>
12943 PR tree-optimization/82965
12944 PR tree-optimization/83991
12945 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
12946 by_profile_only parameter.
12947 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
12948 information if the loop was predicted to iterate too many times.
12949 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
12951 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
12954 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
12957 2018-04-13 Martin Liska <mliska@suse.cz>
12958 Jakub Jelinek <jakub@redhat.com>
12960 PR middle-end/81657
12961 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
12962 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
12963 * builtins.c (expand_builtin_memory_copy_args): Use
12964 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
12965 handle dest_addr == pc_rtx.
12967 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
12970 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
12971 asked to not generate direct moves.
12972 (fix_trunc<mode>si2_stfiwx): Similar.
12973 (fix_trunc<mode>si2_internal): Similar.
12975 2018-04-12 Jakub Jelinek <jakub@redhat.com>
12978 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
12979 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
12980 lookup if dest in some wider mode is known to be const0_rtx and
12981 if so, record permanent equivalence for it to be ZERO_EXTEND of
12982 the narrower mode destination.
12984 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
12986 * lto-streamer-out.c (output_function): Revert 259346.
12987 * omp-expand.c (expand_oacc_for): Likewise.
12989 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
12991 PR rtl-optimization/85354
12992 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
12993 * sel-sched.c (sel_global_init): ... here.
12995 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
12998 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
12999 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
13000 mode for PE-COFF targets.
13001 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
13002 (i386_pe_asm_lto_end): Likewise.
13003 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
13004 (TARGET_ASM_LTO_END): Likewise.
13005 * config/i386/winnt.c (saved_debug_info_level): New static variable.
13006 (i386_pe_asm_lto_start): New function.
13007 (i386_pe_asm_lto_end): Likewise.
13009 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
13010 Richard Biener <rguenther@suse.de>
13012 PR middle-end/84955
13013 * lto-streamer-out.c (output_function): Fix CFG loop state before
13015 * omp-expand.c (expand_oacc_for): Handle calls to internal
13016 functions like regular functions.
13018 2018-04-12 Richard Biener <rguenther@suse.de>
13021 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
13022 for the early LTO debug to properly generate references to it
13023 during DIE emission. Do not re-use that for the skeleton for
13025 (dwarf2out_early_finish): Likewise.
13027 2018-04-12 Jakub Jelinek <jakub@redhat.com>
13030 * config/i386/sse.md
13031 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
13032 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
13033 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
13034 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
13035 and output is a reg, avoid creating invalid lowpart subreg, but
13036 instead split into a 512-bit move. Don't split if not AVX512VL,
13037 input is xmm16+ reg and output is a mem.
13038 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
13039 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
13040 xmm16+ reg and output is a mem.
13042 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13044 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
13045 also for flag_dwarf2_cfi_asm.
13047 2018-04-12 Jakub Jelinek <jakub@redhat.com>
13049 PR rtl-optimization/85342
13050 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
13051 a bool scalar var inside of the loop instead. Don't try to update
13052 recog_data.operand after failed apply_change_group.
13054 2018-04-12 Tom de Vries <tom@codesourcery.com>
13057 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
13058 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
13059 array with flexible array member as array without given dimension.
13060 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
13061 argument for undefined param to true.
13063 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
13066 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
13067 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
13068 from PowerPC section.
13069 * config/rs6000/sysv4.opt (mcall-): Improve help text.
13070 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
13071 help text that is too long.
13072 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
13073 help text that is too long.
13074 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
13075 help text that is too long.
13077 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
13079 * config/alpha/alpha.md (stack_probe_internal): Rename
13080 from "probe_stack". Update all callers.
13082 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13084 PR rtl-optimization/84566
13085 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
13086 sched_macro_fuse_insns.
13088 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13091 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
13092 (compute_block_dependences): ... from here.
13094 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13096 PR tree-optimization/85331
13097 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
13098 from int to HOST_WIDE_INT.
13100 2018-04-11 Martin Jambor <mjambor@suse.cz>
13103 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
13104 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
13105 not the same as the source val.
13106 (cgraph_edge_brings_value_p): New parameter.
13107 (gather_edges_for_value): Pass destination value to
13108 cgraph_edge_brings_value_p.
13109 (perhaps_add_new_callers): Likewise.
13110 (get_info_about_necessary_edges): Likewise and exclude values brought
13111 only by self-recursive edges.
13112 (create_specialized_node): Redirect only clones of self-calling edges.
13113 (+self_recursive_pass_through_p): New function.
13114 (find_more_scalar_values_for_callers_subset): Use it.
13115 (find_aggregate_values_for_callers_subset): Likewise.
13116 (known_aggs_to_agg_replacement_list): Removed.
13117 (decide_whether_version_node): Re-calculate known constants for all
13118 remaining context clones.
13120 2018-04-11 Richard Biener <rguenther@suse.de>
13123 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
13124 from early DWARF output.
13125 (dwarf2out_early_finish): Output line info unconditionally into
13126 early DWARF and add reference to it.
13128 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13131 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
13132 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
13133 other than V2DFmode using iptr mode attribute.
13134 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
13136 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
13138 PR rtl-optimization/84659
13139 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
13141 2018-04-11 Jakub Jelinek <jakub@redhat.com>
13144 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
13146 (output_loc_list): Pass address of a dummy size variable even in the
13147 locview handling loop.
13148 (index_location_lists): Add comment on why skip_loc_list_entry can't
13151 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
13154 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
13157 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
13160 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
13161 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
13162 and -mstring-compare-inline-limit.
13164 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
13167 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
13168 for stack clash protection in a register whenever we need it to be in
13171 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
13173 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
13174 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
13176 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
13179 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
13181 (mlong-double-): Ditto.
13182 * config/rs6000/sysv4.opt (msdata=): Ditto.
13183 (mtls-size=): Ditto.
13185 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
13187 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
13188 erroneous entries for
13189 "vector int vec_ldl (int, long int *)", and
13190 "vector unsigned int vec_ldl (int, unsigned long int *)".
13191 Add comments and entries for
13192 "vector bool char vec_ldl (int, bool char *)",
13193 "vector bool short vec_ldl (int, bool short *)",
13194 "vector bool int vec_ldl (int, bool int *)",
13195 "vector bool long long vec_ldl (int, bool long long *)",
13196 "vector pixel vec_ldl (int, pixel *)",
13197 "vector long long vec_ldl (int, long long *)",
13198 "vector unsigned long long vec_ldl (int, unsigned long long *)".
13199 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
13200 type tree bool_long_long_type_node and correct definition of
13201 bool_V2DI_type_node to make reference to this new type tree.
13202 (rs6000_mangle_type): Replace erroneous reference to
13203 bool_long_type_node with bool_long_long_type_node.
13204 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
13205 comments to emphasize sign distinctions for char and int types and
13206 replace RS6000_BTI_bool_long constant with
13207 RS6000_BTI_bool_long_long constant. Also add comment to restrict
13208 use of RS6000_BTI_pixel.
13209 (bool_long_type_node): Remove this macro definition.
13210 (bool_long_long_type_node): New macro definition
13212 2018-04-10 Jakub Jelinek <jakub@redhat.com>
13214 PR rtl-optimization/85300
13215 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
13216 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
13217 simplify_unary_operation fails.
13219 2018-04-10 Martin Liska <mliska@suse.cz>
13221 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
13222 cgraph_edge and ipa_ref.
13224 2018-04-10 Jakub Jelinek <jakub@redhat.com>
13228 * config/i386/sse.md
13229 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
13230 computation of the VEC_MERGE selector from mask.
13231 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
13232 Fix decoding of the VEC_MERGE selector into mask.
13234 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
13236 PR tree-optimization/85286
13237 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
13239 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
13241 * final.c (final_1): Set insn_last_address as well as
13242 insn_current_address.
13244 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13247 * explow.c (emit_stack_probe): Call validize_mem on memory location
13248 before passing it to gen_probe_stack. Create address operand and
13249 legitimize it for the probe_stack_address case.
13251 2018-04-09 Jan Hubicka <jh@suse.cz>
13254 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
13255 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
13256 * tree.c (free_lang_data_in_type): Fix handling of binfos;
13258 (free_lang_data): Rebuild type inheritance graph.
13260 2018-04-09 Martin Sebor <msebor@redhat.com>
13262 * invoke.texi (-finline-small-functions): Mention other optimization
13264 (-findirect-inlining, -fpartial-inlining): Same.
13265 (-finline-functions-called-once): Same.
13266 (-freorder-blocks-and-partition): Same.
13268 2018-04-09 Jan Hubicka <jh@suse.cz>
13271 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
13272 jumps; choose last target that matches the criteria (i.e.
13273 no partition changes for non-crossing jumps).
13274 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
13275 support for redirecting crossing jumps to non-crossing.
13277 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
13279 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
13280 also for naked functions.
13282 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
13284 * config/arc/arc.md (add_shift): New pattern.
13285 (add_shift2): Likewise.
13286 (sub_shift): Likewise.
13287 (sub_shift_cmp0_noout): Likewise.
13288 (compare_si_ashiftsi): Likewise.
13289 (xbfu_cmp0_noout): New combine pattern.
13290 (xbfu_cmp0"): Likewise.
13291 (movsi_set_cc_insn): Place the predicable variant first.
13292 (commutative_binary_cmp0_noout): Remove clobber.
13293 (commutative_binary_cmp0): New pattern.
13294 (noncommutative_binary_cmp0): Likewise.
13295 (noncommutative_binary_cmp0_noout): Likewise.
13296 (noncommutative_binary_comparison_result_used): Removed.
13297 (rsub_cmp0): New pattern.
13298 (rsub_cmp0_noout): Likewise.
13299 (extzvsi): Changed, keep only meaningful variants.
13300 (SQH, SEZ): New iterators.
13301 (SQH_postfix): New mode attribute.
13302 (SEZ_prefix): New code attribute.
13303 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
13304 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
13305 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
13306 of numerical value.
13307 (noncommutative_operator): Check the availability of barrel
13310 2018-04-09 Richard Biener <rguenther@suse.de>
13312 PR tree-optimization/85284
13313 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
13314 Only use the niter constraining form of simple_iv when the exit
13315 is always executed.
13317 2018-04-09 Tom de Vries <tom@codesourcery.com>
13320 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13321 (define_expand "*memory_barrier"): New define_expand.
13322 (define_insn "memory_barrier"): New insn.
13324 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13326 PR rtl-optimization/80463
13327 PR rtl-optimization/83972
13328 PR rtl-optimization/83480
13330 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
13331 correct producer for the insn.
13332 (tidy_control_flow): Fixup seqnos in case of debug insns.
13334 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13336 PR rtl-optimization/83913
13338 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
13339 different sched-times when merging exprs.
13341 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13343 PR rtl-optimization/83962
13345 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
13346 tidy_fallthru_edge and tidy_control_flow.
13348 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
13350 PR rtl-optimization/83530
13352 * sel-sched.c (force_next_insn): New global variable.
13353 (remove_insn_for_debug): When force_next_insn is true, also leave only
13354 next insn in the ready list.
13355 (sel_sched_region): When the region wasn't scheduled, make another pass
13356 over it with force_next_insn set to 1.
13358 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
13360 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
13362 * config/nds32/constants.md (unspec_volatile_element): Add enum values
13363 for interrupt control.
13364 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13365 functions for interrupt control.
13366 * config/nds32/nds32-intrinsic.md: Likewise.
13367 * config/nds32/nds32_intrinsic.h: Likewise.
13368 * config/nds32/nds32.h (nds32_builtins): Likewise.
13370 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
13372 * config/nds32/nds32.c (nds32_init_machine_status,
13373 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
13374 strict_aligned_p field.
13375 (nds32_expand_to_rtl_hook): New function.
13376 (TARGET_EXPAND_TO_RTL_HOOK): Define.
13377 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
13379 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13380 Chung-Ju Wu <jasonwucj@gmail.com>
13382 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
13383 * config/nds32/nds32-n7.md: New file.
13384 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
13385 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
13387 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
13388 * config/nds32/nds32.md (pipeline_model): Add n7.
13389 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
13390 * config/nds32/pipelines.md: Include n7 settings.
13392 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13393 Chung-Ju Wu <jasonwucj@gmail.com>
13395 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
13396 * config/nds32/nds32-e8.md: New file.
13397 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
13398 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
13400 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
13401 * config/nds32/nds32.md (pipeline_model): Add e8.
13402 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
13403 * config/nds32/pipelines.md: Include e8 settings.
13405 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13406 Chung-Ju Wu <jasonwucj@gmail.com>
13408 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
13409 * config/nds32/nds32-n8.md: New file.
13410 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
13411 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
13413 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
13414 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
13415 * config/nds32/nds32.md (pipeline_model): Add n8.
13416 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
13417 * config/nds32/pipelines.md: Include n8 settings.
13419 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
13420 Chung-Ju Wu <jasonwucj@gmail.com>
13422 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
13423 * config/nds32/nds32-n9-2r1w.md: New file.
13424 * config/nds32/nds32-n9-3r2w.md: New file.
13425 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
13426 nds32_register_ports): New or modify for cpu n9.
13427 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
13429 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
13430 * config/nds32/nds32-utils.c: New file.
13431 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
13432 TARGET_MUL_SLOW): Define.
13433 * config/nds32/nds32.md (pipeline_model): New attribute.
13434 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
13435 New options that support cpu n9.
13436 * config/nds32/pipelines.md: Include n9 settings.
13437 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
13439 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
13441 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
13442 information if necessary.
13443 (output_cond_branch_compare_zero): Likewise.
13444 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
13445 (nds32_target_alignment): Refine for alignment.
13446 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
13447 (FUNCTION_BOUNDARY): Modify.
13448 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
13450 * config/nds32/nds32.opt (malways-align, malign-functions): New.
13452 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
13454 * config/nds32/constants.md (unspec_volatile_element): Add values for
13455 TLB operation and data prefetch.
13456 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
13457 functions for TLB operation and data prefetch.
13458 * config/nds32/nds32-intrinsic.md: Likewise.
13459 * config/nds32/nds32_intrinsic.h: Likewise.
13460 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
13461 (nds32_print_operand): Likewise.
13462 * config/nds32/nds32.h (nds32_builtins): Likewise.
13464 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
13465 Andrew Pinski <pinsika@gcc.gnu.org>
13467 PR middle-end/82976
13468 * match.pd: Use constant_boolean_node of correct type instead of
13469 boolean_true_node or boolean_false_node for simplifying
13470 pointer comparisons to zero.
13472 2018-04-07 Jakub Jelinek <jakub@redhat.com>
13474 PR tree-optimization/80021
13475 * tree.c (verify_type_variant): Make error call in verify_variant_match
13476 translatable and remove final full stop.
13478 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13480 * config/nds32/constants.md (unspec_volatile_element): Add
13481 UNSPEC_VOLATILE_EH_RETURN.
13482 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
13483 nds32_output_stack_pop): Support dwarf exception handling process.
13484 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
13485 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
13486 exception handling process.
13487 (nds32_compute_stack_frame): Likewise.
13488 (nds32_return_addr_rtx): Likewise.
13489 (nds32_initial_elimination_offset): Likewise.
13490 (nds32_expand_prologue): Likewise.
13491 (nds32_expand_epilogue): Likewise.
13492 (nds32_dynamic_chain_address): New function.
13493 * config/nds32/nds32.h (machine_function): Add fields for dwarf
13494 exception handling.
13495 (DYNAMIC_CHAIN_ADDRESS): Define.
13496 (EH_RETURN_DATA_REGNO): Define.
13497 (EH_RETURN_STACKADJ_RTX): Define.
13498 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
13499 patterns for dwarf exception handling.
13501 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13503 * config/nds32/nds32.h: Clean up obsolete macros.
13505 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13507 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13508 Add enum values for particular instructions.
13509 * config/nds32/nds32-intrinsic.c: Implementation of expanding
13510 particular intrinsic functions.
13511 * config/nds32/nds32-intrinsic.md: Likewise.
13512 * config/nds32/nds32_intrinsic.h: Likewise.
13513 * config/nds32/nds32.h (nds32_builtins): Likewise.
13514 * config/nds32/nds32.md (type): Add pbsad and pbsada.
13515 (btst, ave): New patterns for particular instructions.
13517 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13519 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13520 Add enum values for atomic load/store and memory sync.
13521 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
13523 * config/nds32/nds32-intrinsic.md: Likewise.
13524 * config/nds32/nds32_intrinsic.h: Likewise.
13525 * config/nds32/nds32.h (nds32_builtins): Likewise.
13527 2018-04-07 Jakub Jelinek <jakub@redhat.com>
13529 PR tree-optimization/85257
13530 * fold-const.c (native_encode_vector): If not all elts could fit
13531 and off is -1, return 0 rather than offset.
13532 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
13533 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
13534 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
13535 adjust buffer in native_interpret_expr call.
13537 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13539 * config/nds32/constants.md (unspec_volatile_element): Add cache
13540 control enum values.
13541 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
13542 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
13543 * config/nds32/nds32.c (nds32_cctl_names): New.
13544 (nds32_print_operand): Handle cache control register names.
13545 * config/nds32/nds32.h (nds32_builtins): New enum values.
13546 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
13548 * config/nds32/nds32.md (type): Add mmu.
13549 * config/nds32/pipelines.md (simple_insn): Add mmu.
13551 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13553 * config/nds32/nds32.md (type): Remove call.
13554 * config/nds32/pipelines.md (simple_insn): Likewise.
13556 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13558 * config/nds32/constants.md (unspec_volatile_element): Add
13559 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
13560 UNSPEC_VOLATILE_FMFCFG.
13561 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
13562 description for fmfcfg and fmfcsr.
13563 (bdesc_1arg): Add fmtcsr.
13564 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
13565 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
13566 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
13567 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
13568 unspec_fmfcfg): New patterns.
13569 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
13570 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
13571 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
13572 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
13573 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
13574 __nds32__fmfcfg): Define.
13576 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
13578 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
13579 intrinsic register names.
13580 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
13581 intrinsic register enum values and macros.
13583 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
13585 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
13586 for load/store addressing form.
13587 (nds32_print_operand_address): Likewise.
13589 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
13592 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
13593 based on LABEL_REF. Remove useless assertion.
13594 (pic_address_needs_scratch): Fix formatting.
13595 (sparc_legitimize_pic_address): Minor tweaks.
13596 (sparc_delegitimize_address): Adjust assertion accordingly.
13597 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
13598 into symbolic_operand.
13599 (movsi_high_pic_label_ref): Likewise.
13600 (movsi_lo_sum_pic_label_ref): Likewise.
13601 (movdi_pic_label_ref): Likewise.
13602 (movdi_high_pic_label_ref): Likewise.
13603 (movdi_lo_sum_pic_label_ref): Likewise.
13605 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
13607 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
13608 custom LIB_SPEC setup.
13610 2018-04-06 Ruslan Bukin <br@bsdpad.com>
13611 Kito Cheng <kito.cheng@gmail.com>
13613 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
13614 * config/riscv/freebsd.h: New.
13616 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13618 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
13619 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
13622 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13623 Kito Cheng <kito.cheng@gmail.com>
13625 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
13626 nds32_output_call, nds32_symbol_binds_local_p): New functions.
13627 * config/nds32/nds32-protos.h (nds32_output_call,
13628 nds32_output_return): Declare.
13629 * config/nds32/nds32.md: Refine all the call and return patterns.
13631 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13634 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
13635 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
13637 PR rtl-optimization/84872
13638 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
13639 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
13640 EDGE_CROSSING edge.
13642 2018-04-06 Tamar Christina <tamar.christina@arm.com>
13644 * expr.c (copy_blkmode_to_reg): Revert 254862.
13645 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
13647 2018-04-06 Richard Biener <rguenther@suse.de>
13649 PR middle-end/85244
13650 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
13651 after seeing a component reference with an adjacent field. Treat
13652 refs to arrays at struct end of external decls similar to
13653 refs to unconstrained commons.
13655 2018-04-06 Jakub Jelinek <jakub@redhat.com>
13658 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
13659 look through SAVE_EXPRs with non-side-effects argument. Adjust
13661 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
13664 2018-04-06 Richard Biener <rguenther@suse.de>
13666 PR middle-end/85180
13667 * alias.c (find_base_term): New wrapper around find_base_term
13668 unwinding CSELIB_VAL_PTR changes.
13669 (find_base_term): Do not restore CSELIB_VAL_PTR during the
13672 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13674 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
13676 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
13677 constant definitions.
13678 ("nop"): lr 0,0 -> nopr r0
13679 ("nop_lr0", "nop_lr1"): New insn definitions.
13681 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
13683 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
13684 NDS32_V3PUSH_AVAILABLE_P macro.
13686 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
13687 Chung-Ju Wu <jasonwucj@gmail.com>
13689 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
13690 (nds32*-*-*): Add float and fpu_config into supported_defaults.
13691 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
13692 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
13693 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
13694 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
13695 * config/nds32/constraints.md: New constraints and checking for hard
13696 float configuration.
13697 * config/nds32/iterators.md: New mode iterator and attribute for hard
13698 float configuration.
13699 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
13701 * config/nds32/nds32-fpu.md: New file.
13702 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
13703 deal with hard float code generation.
13704 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
13706 (abi_type, float_reg_number): New enum type.
13707 * config/nds32/nds32-predicates.c: New predicates for hard float.
13708 * config/nds32/nds32-protos.h: Declare functions for hard float.
13709 * config/nds32/nds32.c: Implementation for hard float configuration.
13710 * config/nds32/nds32.h: Definitions for hard float configuration.
13711 * config/nds32/nds32.md: Include hard float machine description and
13712 modify patterns for hard float configuration.
13713 * config/nds32/nds32.opt: New options for hard float configuration.
13714 * config/nds32/predicates.md: New predicates for hard float
13717 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13719 * common/config/nds32/nds32-common.c
13720 (nds32_option_optimization_table): Enable -mreleax-hint by default.
13722 2018-04-05 Jakub Jelinek <jakub@redhat.com>
13724 PR middle-end/85195
13725 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
13726 CONSTRUCTOR_ELT (ctor, ...)->value.
13728 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
13731 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
13733 2018-04-05 Tom de Vries <tom@codesourcery.com>
13736 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
13739 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13740 Kito Cheng <kito.cheng@gmail.com>
13742 * config/nds32/constraints.md (U33): Fine-tune checking condition.
13743 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
13744 * config/nds32/nds32.h (nds32_16bit_address_type): Add
13745 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
13747 2018-04-05 Shiva Chen <shiva0217@gmail.com>
13748 Kito Cheng <kito.cheng@gmail.com>
13750 * config/nds32/constraints.md (Ufe): New memory constraint.
13751 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
13752 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
13753 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
13755 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
13756 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
13758 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13760 * config/nds32/nds32.md: Use optimize_size in the condition for
13761 alu-shift instructions.
13763 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13765 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
13767 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13769 * config/nds32/nds32.md (negsi2): Refine pattern.
13771 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13772 Chung-Ju Wu <jasonwucj@gmail.com>
13774 * config/nds32/iterators.md (shift_rotate): New code iterator.
13775 (shift): New code attribute.
13776 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
13777 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
13778 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
13779 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
13780 bit-wise operations.
13781 (andsi3, *andsi3): Ditto.
13782 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
13783 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
13784 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
13785 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
13786 nds32_ior_operand, nds32_xor_operand): New predicates.
13788 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13790 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
13791 (addsi3, subsi3): ... this.
13793 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13795 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
13797 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13799 * config/nds32/nds32.md: Adjust indention.
13801 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
13803 * config/nds32/nds32.md (feature): New attribute.
13805 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
13807 * config/nds32/nds32.md (subtype): New attribute.
13809 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
13812 * config/arm/arm-builtins.c (arm_expand_builtin): Change
13813 expansion to perform a bitwise AND of the argument followed by a
13814 boolean negation of the result.
13816 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
13818 PR rtl-optimization/84878
13819 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
13820 the basic block. Assert the use reference is not artificial and that
13821 it has an associated insn.
13823 2018-04-04 Michael Matz <matz@suse.de>
13825 * builtins.c (compute_objsize): Pass correct operand
13826 to array_at_struct_end_p.
13828 2018-04-04 Richard Biener <rguenther@suse.de>
13831 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
13832 from contexts for DINFO_LEVEL_TERSE and below.
13834 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13836 * config/nds32/nds32-doubleword.md (move_<mode>): Require
13837 resiter_operand condition.
13838 * config/nds32/nds32.md (*move<mode>): Ditto.
13840 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13841 Monk Chiang <sh.chiang04@gmail.com>
13843 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
13845 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13847 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
13849 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13850 Kito Cheng <kito.cheng@gmail.com>
13852 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
13853 nds32_cond_code_str, output_cond_branch,
13854 output_cond_branch_compare_zero, nds32_expand_cbranch,
13855 nds32_expand_cstore, nds32_expand_movcc,
13856 nds32_output_cbranchsi4_equality_zero,
13857 nds32_output_cbranchsi4_equality_reg,
13858 nds32_output_cbranchsi4_equality_reg_or_const_int,
13859 nds32_output_cbranchsi4_greater_less_zero: New functions.
13860 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
13861 nds32_expand_cstore, nds32_expand_movcc,
13862 nds32_output_cbranchsi4_equality_zero,
13863 nds32_output_cbranchsi4_equality_reg,
13864 nds32_output_cbranchsi4_equality_reg_or_const_int,
13865 nds32_output_cbranchsi4_greater_less_zero): Declare.
13866 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
13867 nds32_rimm11s_operand): New predicates.
13868 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
13869 * config/nds32/nds32.md: Rewrite all the branch and conditional move
13872 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
13874 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
13875 * config/nds32/nds32.md: Ditto.
13876 * config/nds32/pipelines.md: Ditto.
13878 2018-04-04 Richard Biener <rguenther@suse.de>
13880 PR tree-optimization/85168
13881 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
13882 propagating abnormals.
13884 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13886 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
13888 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
13889 Kito Cheng <kito.cheng@gmail.com>
13891 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
13892 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
13893 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
13894 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
13895 * config/nds32/nds32.md (sibcall_internal): New.
13896 (sibcall_register): Remove.
13897 (sibcall_immediate): Remove.
13898 (sibcall_value_internal): New.
13899 (sibcall_value_register): Remove.
13900 (sibcall_value_immediate): Remove.
13901 * config/nds32/predicates.md (nds32_general_register_operand): New.
13902 (nds32_call_address_operand): New.
13904 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13906 PR rtl-optimization/85167
13907 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
13908 bb_defs if *split_p, instead preinitialize it to NULL.
13910 PR tree-optimization/85156
13911 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
13912 evaluating the argument multiple times.
13914 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
13916 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
13918 (_mm_cvtpd_ps): Likewise.
13919 (_mm_cvttpd_epi32): Likewise.
13920 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
13921 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
13922 vector, pixel, and bool following altivec.h include.
13924 2018-04-03 Martin Sebor <msebor@redhat.com>
13926 * doc/extend.texi (Common Function Attributes): Clarify.
13927 (const attribute): Likewise.
13928 (pure attribute): Likewise.
13930 2018-04-03 Jakub Jelinek <jakub@redhat.com>
13933 * config/i386/i386.c (ix86_expand_vector_set): Use
13934 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
13936 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
13938 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
13939 instructions when changing rounding bits to preserve precision bits
13940 in the x87 control word.
13942 2018-04-03 Martin Liska <mliska@suse.cz>
13944 PR tree-optimization/82491
13945 * rtl.h (strip_offset_and_add): Replace += suboffset with
13946 poly_uint64 () + suboffset.
13948 2018-03-29 Martin Liska <mliska@suse.cz>
13949 Martin Jambor <mjambor@suse.cz>
13952 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
13953 param_type is not an integral or pointer type.
13955 2018-04-03 Richard Biener <rguenther@suse.de>
13957 * sese.h (recompute_all_dominators): Remove.
13959 2018-04-02 Martin Sebor <msebor@redhat.com>
13961 * doc/invoke.texi (-Wrestrict): Fix typos.
13963 2018-04-02 Jim Wilson <jimw@sifive.com>
13965 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
13966 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
13967 (<optab>di3, <optab>si3_extend): Likewise.
13968 (<optab>si3_mask, <optab>si3_mask_1): New.
13969 (<optab>di3_mask, <optab>di3_mask_1): New.
13970 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
13971 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
13972 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
13974 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
13976 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
13979 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
13981 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
13982 (nds32_canonicalize_comparison): New function.
13984 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
13985 Kito Cheng <kito.cheng@gmail.com>
13986 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13988 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
13989 * config/nds32/constants.md (unspec_volatile_element): Add
13990 UNSPEC_VOLATILE_RELAX_GROUP.
13991 * config/nds32/nds32-relax-opt.c: New file.
13992 * config/nds32/nds32-predicates.c
13993 (nds32_symbol_load_store_p): New function.
13994 * config/nds32/nds32-protos.h
13995 (nds32_symbol_load_store_p): Declare function.
13996 (make_pass_nds32_relax_opt): Declare new rtl pass function.
13997 * config/nds32/nds32.c
13998 (nds32_register_pass): New function to register pass.
13999 (nds32_register_passes): New function to register passes.
14000 * config/nds32/nds32.md (relax_group): New pattern.
14001 * config/nds32/nds32.opt (mrelax-hint): New option.
14002 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
14004 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
14006 * config/nds32/t-nds32: Modify files dependency.
14008 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
14010 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
14011 (PROFILE_HOOK): Define its implementation.
14013 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
14015 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
14016 type and 32-bit size.
14018 2018-04-01 Jakub Jelinek <jakub@redhat.com>
14020 PR middle-end/85090
14021 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
14022 (V_128_256): New mode iterator.
14023 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
14024 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
14025 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
14027 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
14028 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
14030 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
14033 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
14034 NaN inputs correctly.
14036 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
14039 * config/rs6000/vsx.md (??r): New mode attribute.
14040 (*vsx_mov<mode>_64bit): Use it.
14041 (*vsx_mov<mode>_32bit): Likewise.
14043 2018-03-30 Martin Sebor <msebor@redhat.com>
14045 PR tree-optimization/84818
14046 * builtins.c (check_access): Use warning_n.
14048 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14051 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
14053 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
14056 2018-03-30 Julia Koval <julia.koval@intel.com>
14059 * x86-tune.def (movx, partial_reg_dependency): Enable for
14062 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
14064 PR inline-asm/84985
14065 * lra-constraints.c (process_alt_operands): Move setting
14066 this_alternative_matches below.
14068 2018-03-29 Martin Liska <mliska@suse.cz>
14071 * doc/invoke.texi: Document how LTO works with debug info.
14072 Describe auto-load support of binutils. Mention 'x86-64'
14073 as valid option value of -march option.
14075 2018-03-29 Jakub Jelinek <jakub@redhat.com>
14077 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
14080 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
14081 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
14082 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
14085 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
14088 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
14089 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
14090 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
14091 for RS6000_BTM_POWERPC64.
14092 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
14093 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
14094 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
14097 (DIVDEU): Likewise.
14099 2018-03-28 Carl Love <cel@us.ibm.com>
14102 2017-09-27 Carl Love <cel@us.ibm.com>
14104 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14105 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14106 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14109 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14111 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
14112 instead of __vector bool.
14113 (_mm_max_pu8): Likewise.
14114 (_mm_min_pi16): Likewise.
14116 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
14119 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
14120 (DIVWEUO): Likewise.
14121 (DIVDEO): Likewise.
14122 (DIVDEUO): Likewise.
14123 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
14124 DIVWEUO and DIVDEUO.
14125 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
14126 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
14127 (div_extend): Likewise.
14128 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
14130 (__builtin_divweuo): Likewise.
14131 (__builtin_divdeo): Likewise.
14132 (__builtin_divdeuo): Likewise.
14134 2018-03-28 Jakub Jelinek <jakub@redhat.com>
14137 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
14138 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
14140 PR tree-optimization/82004
14141 * gimple-match-head.c (optimize_pow_to_exp): New function.
14142 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
14143 Don't fold to exp if optimize_pow_to_exp is false.
14145 2018-03-28 Martin Liska <mliska@suse.cz>
14148 * calls.c (initialize_argument_information): Fix trailing space.
14149 * common.opt: Fix typo and provide better explanation for
14150 -fsanitize-coverage option.
14151 * config/i386/i386.opt: Fix typo.
14153 2018-03-28 Jakub Jelinek <jakub@redhat.com>
14154 Martin Liska <mliska@suse.cz>
14157 * gimplify.c (asan_poison_variable): Don't do the check for
14158 gimplify_omp_ctxp here.
14159 (gimplify_decl_expr): Do it here.
14160 (gimplify_target_expr): Likewise.
14162 2018-03-28 Martin Liska <mliska@suse.cz>
14165 * config/i386/i386.c (ix86_function_arg_advance): Do not call
14166 chkp_type_bounds_count if MPX is not enabled.
14168 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
14170 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
14172 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
14175 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
14176 function to create the function decl for complex long double
14177 multiply and divide for -mabi=ieeelongdouble.
14178 (init_float128_ieee): Call it.
14180 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
14183 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
14184 -fcf-protection=branch -mibt.
14185 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
14187 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14190 * config/arm/arm.c (arm_valid_symbolic_address): Handle
14191 arm_word_relocations.
14193 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
14196 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
14197 extern array declarations.
14199 2018-03-27 Richard Biener <rguenther@suse.de>
14201 PR middle-end/84067
14202 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
14203 explicit single_use checks.
14205 2018-03-27 Richard Biener <rguenther@suse.de>
14207 PR tree-optimization/85082
14208 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
14211 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14213 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
14214 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
14215 Turn on fasynchronous-unwind-tables and funwind-tables.
14217 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
14220 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
14221 (*bmi_blsr_<mode>_ccz): Ditto.
14223 2018-03-26 Tom de Vries <tom@codesourcery.com>
14225 PR tree-optimization/85063
14226 * omp-general.c (offloading_function_p): New function. Factor out
14228 * omp-offload.c (pass_omp_target_link::gate): ... here.
14229 * omp-general.h (offloading_function_p): Declare.
14230 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
14231 with attribute omp declare target for offloading functions.
14233 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
14235 PR tree-optimization/84005
14236 * tree-data-ref.h (get_base_for_alignment): Declare.
14237 * tree-data-ref.c (get_base_for_alignment_1): New function.
14238 (get_base_for_alignment): Likewise.
14239 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
14240 get_base_for_alignment to find a suitable base object, instead
14241 of always using drb->base_address.
14243 2018-03-23 Jakub Jelinek <jakub@redhat.com>
14245 PR inline-asm/85022
14246 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
14247 known size by default.
14249 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
14251 PR inline-asm/85030
14252 * lra-constraints.c (process_alt_operands): Don't match BLKmode
14253 and non BLKmode operands.
14255 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14258 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
14259 Clean up attributes.
14261 2018-03-23 Richard Biener <rguenther@suse.de>
14264 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
14265 we are going to emit early debug for LTO.
14267 2018-03-23 Jakub Jelinek <jakub@redhat.com>
14269 PR inline-asm/85034
14270 * function.c (match_asm_constraints_1): Don't optimize if input
14271 doesn't satisfy general_operand predicate for output's mode.
14273 PR inline-asm/85022
14274 * alias.c (write_dependence_p): Don't require for x_canonicalized
14275 non-VOIDmode if x has VOIDmode.
14278 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
14279 just don't try to optimize it rather than assert it never happens.
14281 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
14283 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
14284 macro expansions for definition of ST_INTERNAL_<mode> and
14285 LD_INTERNAL_<mode> builtins.
14286 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
14288 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
14290 (altivec_expand_st_builtin): Likewise.
14291 (altivec_expand_builtin): Remove calls to deleted functions.
14292 (rs6000_address_for_altivec): Delete this function.
14293 * config/rs6000/vector.md: Remove expands for
14294 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
14296 2018-03-22 Sudakshina Das <sudi.das@arm.com>
14299 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
14300 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
14301 re-computing once computed.
14302 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
14303 (arm_init_machine_status): Initialize
14304 machine->static_chain_stack_bytes.
14306 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
14309 * doc/extend.texi: Add four new prototypes for vec_ld.
14310 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
14311 definitions for more logical presentation.
14312 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
14313 entries for V1TI variants of __builtin_altivec_ld builtin.
14314 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
14315 handling of V1TI variant of LVX icode pattern.
14316 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
14317 (rs6000_gimple_fold_builtin): Likewise.
14318 (altivec_init_builtins): Add code to define
14319 __builtin_altivec_lvx_v1ti function.
14321 2018-03-22 Jakub Jelinek <jakub@redhat.com>
14323 PR inline-asm/84941
14324 * function.c (match_asm_constraints_1): Don't do the optimization
14325 if input isn't a REG, SUBREG, MEM or constant.
14327 2018-03-22 Tom de Vries <tom@codesourcery.com>
14329 PR tree-optimization/84956
14330 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
14331 bb_has_abnormal_pred.
14333 2018-03-22 Jakub Jelinek <jakub@redhat.com>
14336 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
14337 DECL_INITIAL (decl) to decl at the end.
14338 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
14339 adjust the comment.
14341 2018-03-21 Joseph Myers <joseph@codesourcery.com>
14343 * doc/extend.texi (__builtin_tgmath): Document when complex
14344 integer types are treated as _Complex _Float64.
14346 2018-03-21 Tom de Vries <tom@codesourcery.com>
14348 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
14350 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14352 PR tree-optimization/84960
14353 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
14354 if it is ENTRY block, move them into single succ of ENTRY in that case.
14356 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
14358 PR tree-optimization/84811
14359 * poly-int.h (poly_span_traits): Remove the T3 parameter and
14360 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
14361 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
14362 (known_subrange_p): Update accordingly. Cast each value involved
14363 in the size comparison, rather than casting the result of the
14366 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14368 PR tree-optimization/84982
14369 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
14370 by flipping the least significant bit rather than all bits from
14371 bitpos to bitpos + bitsize - 1.
14373 2018-03-21 Nathan Sidwell <nathan@acm.org>
14375 * doc/extend.texi (Deprecated Features): Remove mention of
14376 long-deleted deprecations.
14378 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14381 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
14382 * configure: Regenerate.
14384 2018-03-21 Tom de Vries <tom@codesourcery.com>
14386 PR tree-optimization/83126
14387 * tree-parloops.c (num_phis): New function.
14388 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
14390 2018-03-21 Nathan Sidwell <nathan@acm.org>
14392 * doc/extend.texi (Deprecated Features): Update deprecated flags,
14393 mention anon-struct/union members and trailing attributes.
14395 2018-03-21 Bin Cheng <bin.cheng@arm.com>
14397 PR tree-optimization/84969
14398 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
14399 builtin memset partitions if they set different rhs values.
14401 2018-03-21 Jakub Jelinek <jakub@redhat.com>
14403 PR rtl-optimization/84989
14404 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
14405 VEC_DUPLICATE with scalar result mode.
14407 2018-03-21 Martin Liska <mliska@suse.cz>
14410 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
14411 not intended return statement.
14413 2018-03-21 Martin Liska <mliska@suse.cz>
14416 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
14417 (chkp_find_bound_slots_1): Limit number of iterations.
14419 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
14422 * Minor grammar fixes for x86 options.
14424 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14427 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
14428 holding REG_CFA_RESTORE notes, instead turn them into a USE.
14430 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
14433 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
14434 (altivec_lvx_<mode>_1op): Likewise.
14435 (altivec_stvx_<mode>_2op): Likewise.
14436 (altivec_stvx_<mode>_1op): Likewise.
14437 (altivec_lvx_<VM2:mode>): New define_expand.
14438 (altivec_stvx_<VM2:mode>): Likewise.
14439 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
14440 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14441 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
14442 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
14443 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
14444 (rs6000_gen_lvx): Likewise.
14445 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
14446 (altivec_expand_stv_builtin): Likewise.
14447 (altivec_expand_builtin): Likewise.
14448 * config/rs6000/vector.md: Likewise.
14450 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14453 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
14456 2018-03-20 Richard Biener <rguenther@suse.de>
14459 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
14460 sign-conversions as zero, fall back to standard scalar_stmt
14463 2018-03-20 Martin Liska <mliska@suse.cz>
14466 * predict.c (rebuild_frequencies): Handle case when we have
14467 PROFILE_ABSENT, but flag_guess_branch_prob is false.
14469 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14472 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
14473 flag_section_anchors.
14474 * varasm.c (use_blocks_for_decl_p): Remove hack for
14475 dw2_force_const_mem.
14478 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
14480 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
14481 be created, use lowpart_subreg of operands[0] rather than operands[0]
14483 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
14484 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
14485 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
14486 and n constraint instead of aarch64_shift_imm_di and Usd.
14487 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
14488 (*aarch64_<optab>_reg_minus<mode>3): ... this.
14490 2018-03-20 Sudakshina Das <sudi.das@arm.com>
14493 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
14494 to favor GPR over NEON registers.
14495 (<shift>di3_neon): Likewise.
14497 2018-03-20 Tom de Vries <tom@codesourcery.com>
14500 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
14501 (nvptx_process_pars): Emit bar.sync asap and alap.
14503 2018-03-20 Tom de Vries <tom@codesourcery.com>
14506 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
14507 seen_label if seen_label is already set.
14509 2018-03-20 Jakub Jelinek <jakub@redhat.com>
14512 * config/i386/i386.c (fold_builtin_cpu): For features above 31
14513 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
14514 Use 1U instead of 1. Formatting fixes.
14517 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
14518 instead of TREE_TYPE (s1) for the return value.
14520 2018-03-19 Jakub Jelinek <jakub@redhat.com>
14522 PR tree-optimization/84946
14523 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
14524 bitsize + bitsize in poly_uint64 rather than poly_int64.
14527 * dwarf2asm.c: Include fold-const.c.
14528 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
14529 of decl rather than decl itself.
14531 PR rtl-optimization/84643
14532 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
14534 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
14537 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
14538 calling assemble_variable.
14540 2018-03-19 Sudakshina Das <sudi.das@arm.com>
14543 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
14544 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
14546 2018-03-19 Jim Wilson <jimw@sifive.com>
14549 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
14550 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
14551 (riscv_first_stack_step): Likewise.
14552 (riscv_option_override): Use STACK_BOUNDARY instead of
14553 MIN_STACK_BOUNDARY.
14554 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
14555 MIN_STACK_BOUNDARY.
14556 (BIGGEST_ALIGNMENT): Set to 128.
14557 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
14558 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
14561 2018-03-19 Richard Biener <rguenther@suse.de>
14563 PR tree-optimization/84933
14564 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
14565 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
14567 2018-03-19 Richard Biener <rguenther@suse.de>
14569 PR tree-optimization/84859
14570 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
14571 (cond_if_else_store_replacement): Perform sinking operation on
14572 single-store BBs regardless of MAX_STORES_TO_SINK setting.
14573 Generalize what a BB with a single eligible store is.
14575 2018-03-19 Richard Biener <rguenther@suse.de>
14577 PR tree-optimization/84929
14578 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
14579 chrec_is_positive against non-chrec arg.
14581 2018-03-19 Tamar Christina <tamar.christina@arm.com>
14584 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
14586 2018-03-18 Martin Liska <mliska@suse.cz>
14588 PR rtl-optimization/84635
14589 * regrename.c (build_def_use): Use matches_mode only when
14592 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
14594 PR tree-optimization/84913
14595 * tree-vect-loop.c (vectorizable_reduction): Don't try to
14596 vectorize chains of COND_EXPRs.
14598 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14600 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
14602 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14604 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
14606 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
14608 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
14610 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14611 Kito Cheng <kito.cheng@gmail.com>
14613 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
14614 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
14615 (nds32_adjust_reg_alloc_order): New function.
14616 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
14618 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
14620 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
14621 nds32_print_operand, nds32_print_operand_address): Use
14622 HOST_WIDE_INT_PRINT_DEC instead.
14624 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
14626 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
14628 2018-03-17 Jakub Jelinek <jakub@redhat.com>
14631 * config/i386/i386.c (initial_ix86_tune_features,
14632 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
14633 unsigned long long.
14634 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
14635 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
14636 rather than 1u << ix86_tune. Formatting fix.
14637 (ix86_option_override_internal): Change ix86_arch_mask from
14638 unsigned int to unsigned HOST_WIDE_INT, initialize to
14639 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
14640 (ix86_function_specific_restore): Likewise.
14642 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14645 * postreload.c (reload_combine_recognize_pattern): Perform
14646 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
14647 truncate_int_for_mode the result for the destination's mode.
14650 * hsa-gen.c (mem_type_for_type): Fix comment typo.
14651 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
14653 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
14656 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
14659 * lra-assigns.c (lra_split_hard_reg_for): Don't use
14660 regno_allocno_class_array and sorted_pseudos.
14661 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
14662 insns where regno is used.
14664 2018-03-16 Martin Liska <mliska@suse.cz>
14667 * multiple_target.c (create_dispatcher_calls): Redirect
14668 reference in the symbol table.
14670 2018-03-16 Martin Liska <mliska@suse.cz>
14673 * multiple_target.c (create_dispatcher_calls): Redirect also
14676 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14680 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
14681 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
14682 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
14684 2018-03-16 Julia Koval <julia.koval@intel.com>
14686 * doc/invoke.texi (Skylake Server): Add CLWB.
14687 Cannonlake): Remove CLWB.
14689 2018-03-16 Jakub Jelinek <jakub@redhat.com>
14691 PR tree-optimization/84841
14692 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
14694 (FLOAT_ONE_CONST_TYPE): Define.
14695 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
14696 (sort_by_operand_rank): Put entries with higher constant_type last
14697 rather than first to match comments.
14699 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
14701 * config/nios2/nios2.md (movsi_internal): Fix thinko in
14704 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14707 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
14708 check and use address of target always.
14710 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
14713 * config/i386/i386.c (indirect_thunk_needed): Update comments.
14714 (indirect_thunk_bnd_needed): Likewise.
14715 (indirect_thunks_used): Likewise.
14716 (indirect_thunks_bnd_used): Likewise.
14717 (indirect_return_needed): New.
14718 (indirect_return_bnd_needed): Likewise.
14719 (output_indirect_thunk_function): Add a bool argument for
14721 (output_indirect_thunk_function): Don't generate alias for
14722 function return thunk.
14723 (ix86_code_end): Call output_indirect_thunk_function to generate
14724 function return thunks.
14725 (ix86_output_function_return): Set indirect_return_bnd_needed
14726 and indirect_return_needed instead of indirect_thunk_bnd_needed
14727 and indirect_thunk_needed.
14729 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
14731 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
14732 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
14733 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
14735 2018-03-15 David Malcolm <dmalcolm@redhat.com>
14736 Paul Hua <paul.hua.gm@gmail.com>
14739 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
14741 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
14743 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
14744 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
14745 resp. SFmode cases.
14747 2018-03-15 Tamar Christina <tamar.christina@arm.com>
14750 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
14751 instead of GET_MODE_SIZE when comparing Units.
14753 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14756 * varasm.c (hash_section): Return an unchangeble hash value
14757 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
14758 Return !aarch64_can_use_per_function_literal_pools_p ().
14760 2018-03-15 Jakub Jelinek <jakub@redhat.com>
14763 * optabs.c (emit_conditional_move): Pass address of cmode's copy
14764 rather than address of cmode as last argument to prepare_cmp_insn.
14766 2018-03-15 Julia Koval <julia.koval@intel.com>
14768 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
14769 F_AVX512VNNI, F_AVX512BITALG): New.
14771 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
14774 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
14775 insn for floating-point loads and stores.
14777 2018-03-14 Carl Love <cel@us.ibm.com>
14779 * config/rs6000/rs6000-c.c: Add macro definitions for
14780 ALTIVEC_BUILTIN_VEC_PERMXOR.
14781 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
14782 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
14783 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
14784 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
14786 * config/doc/extend.texi: Add prototypes for vec_permxor.
14788 2018-03-14 David Malcolm <dmalcolm@redhat.com>
14791 * diagnostic-show-locus.c (class layout_point): Convert m_line
14792 from int to linenum_type.
14793 (line_span::comparator): Use linenum "compare" function when
14794 comparing line numbers.
14795 (test_line_span): New function.
14796 (layout_range::contains_point): Convert param "row" from int to
14798 (layout_range::intersects_line_p): Likewise.
14799 (layout::will_show_line_p): Likewise.
14800 (layout::print_source_line): Likewise.
14801 (layout::should_print_annotation_line_p): Likewise.
14802 (layout::print_annotation_line): Likewise.
14803 (layout::print_leading_fixits): Likewise.
14804 (layout::annotation_line_showed_range_p): Likewise.
14805 (struct line_corrections): Likewise for field m_row.
14806 (line_corrections::line_corrections): Likewise for param "row".
14807 (layout::print_trailing_fixits): Likewise.
14808 (layout::get_state_at_point): Likewise.
14809 (layout::get_x_bound_for_row): Likewise.
14810 (layout::print_line): Likewise.
14811 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
14812 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
14813 * input.c (selftest::test_linenum_comparisons): New function.
14814 (selftest::input_c_tests): Call it.
14815 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
14816 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
14817 * selftest.h (ASSERT_GT): New macro.
14818 (ASSERT_GT_AT): New macro.
14819 (ASSERT_LT): New macro.
14820 (ASSERT_LT_AT): New macro.
14822 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
14824 PR rtl-optimization/84780
14825 * combine.c (distribute_links): Don't make a link based on pc_rtx.
14827 2018-03-14 Martin Liska <mliska@suse.cz>
14829 * tree.c (record_node_allocation_statistics): Use
14830 get_stats_node_kind.
14831 (get_stats_node_kind): New function extracted from
14832 record_node_allocation_statistics.
14833 (free_node): Use get_stats_node_kind.
14835 2018-03-14 Richard Biener <rguenther@suse.de>
14837 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
14838 that the value-set of ANTIC_IN doesn't grow.
14841 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
14843 (BB_VISITED_WITH_VISITED_SUCCS): New define.
14844 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
14846 2018-03-14 Julia Koval <julia.koval@intel.com>
14848 * config.gcc (icelake-client, icelake-server): New.
14850 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
14851 (initial_ix86_arch_features): Ditto.
14852 (PTA_SKYLAKE): Add SGX.
14853 (PTA_ICELAKE): Remove.
14854 (PTA_ICELAKE_CLIENT): New.
14855 (PTA_ICELAKE_SERVER): New.
14856 (ix86_option_override_internal): Split up icelake on icelake client and
14858 (get_builtin_code_for_version): Ditto.
14859 (fold_builtin_cpu): Ditto.
14860 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
14861 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
14862 * config/i386/i386.h (processor_type): Ditto.
14863 * doc/invoke.texi: Ditto.
14865 2018-03-14 Jakub Jelinek <jakub@redhat.com>
14868 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
14869 INTEGER_CST offset, add it together with bitpos / 8 and
14870 sign extend based on POINTER_SIZE.
14874 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
14877 * config/i386/constraints.md (Yc): New register constraint.
14878 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
14879 Use Yc constraint for alternative 2 of operand 0. Remove
14880 preferred_for_speed attribute.
14882 2018-03-14 Richard Biener <rguenther@suse.de>
14884 PR tree-optimization/84830
14885 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
14886 with the old one to avoid oscillations.
14888 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
14891 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
14893 (assign_by_spills): Return a flag of reload assignment failure.
14894 Do not process the reload assignment failures. Do not spill other
14895 reload pseudos if they has the same reg class. Update n if
14897 (lra_assign): Add a return arg. Set up from the result of
14898 assign_by_spills call.
14899 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
14900 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
14901 usage_insns if it is not NULL.
14902 (spill_hard_reg_in_range): New function.
14903 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
14904 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
14905 function prototypes.
14906 (lra_assign): Change prototype.
14907 * lra.c (lra): Add code to deal with fails by splitting hard reg
14910 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
14912 * config/riscv/riscv.opt (mrelax): New option.
14913 * config/riscv/riscv.c (riscv_file_start): Emit ".option
14914 "norelax" when riscv_mrelax is disabled.
14915 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
14917 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14920 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
14921 reassociation for int modes.
14923 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14925 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
14926 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
14928 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
14929 * config/aarch64/aarch64-sve.md
14930 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
14931 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
14932 (*extend<mode><Vwide>2): Rename to...
14933 (aarch64_sve_extend<mode><Vwide>2): ...this.
14934 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
14935 renaming the old pattern to...
14936 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
14938 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
14939 define_expand, renaming the old pattern to...
14940 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
14941 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
14942 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
14943 account when deciding which SVE instruction the optab should use.
14944 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
14946 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14948 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
14949 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
14950 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
14951 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
14952 (tlsdesc_small_<mode>): Turn a define_expand and use
14953 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
14954 (tlsdesc_small_advsimd_<mode>): ...this.
14955 (tlsdesc_small_sve_<mode>): New pattern.
14957 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
14959 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
14960 (UNSPEC_UMUL_HIGHPART): New constants.
14961 (MUL_HIGHPART): New int iteraor.
14962 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
14963 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
14965 (*<su>mul<mode>3_highpart): New define_insn.
14967 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14970 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14973 2018-03-13 Martin Liska <mliska@suse.cz>
14976 * (sem_item_optimizer::sem_item_optimizer): Initialize new
14978 (sem_item_optimizer::~sem_item_optimizer): Release it.
14979 (sem_item_optimizer::merge_classes): Register variable aliases.
14980 (sem_item_optimizer::fixup_pt_set): New function.
14981 (sem_item_optimizer::fixup_points_to_sets): Likewise.
14982 * ipa-icf.h: Declare new variables and functions.
14984 2018-03-13 Jakub Jelinek <jakub@redhat.com>
14986 PR middle-end/84834
14987 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
14988 integer_pow2p@2 and test integer_pow2p in condition.
14989 (A < 0 ? C : 0): Similarly for @1.
14991 PR middle-end/84831
14992 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
14993 characters starting at p contain '\0' character, don't look beyond
14997 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
14998 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
15001 * reg-stack.c (change_stack): Change update_end var from int to
15002 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
15003 also call set_block_for_insn on the newly added insns and rescan.
15006 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
15007 on the last operand.
15010 * tree.c (stabilize_reference_1): Return save_expr (e) for
15011 STATEMENT_LIST even if it doesn't have side-effects.
15013 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
15015 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
15017 2018-03-12 Renlin Li <renlin.li@arm.com>
15019 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
15020 aarch64_output_scalar_simd_mov_immediate.
15022 2018-03-12 Martin Sebor <msebor@redhat.com>
15024 PR tree-optimization/83456
15025 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
15026 for perfectly overlapping calls to memcpy.
15027 (gimple_fold_builtin_memory_chk): Same.
15028 (gimple_fold_builtin_strcpy): Handle no-warning.
15029 (gimple_fold_builtin_stxcpy_chk): Same.
15030 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
15032 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
15034 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
15035 parameter. Use it for SFmode.
15036 (rs6000_function_arg_advance_1): Adjust.
15037 (rs6000_function_arg): Adjust.
15038 (rs6000_gimplify_va_arg): Pass false for that new parameter.
15040 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
15042 PR rtl-optimization/84169
15043 PR rtl-optimization/84780
15044 * combine.c (can_combine_p): Check for a 2-insn combination whether
15045 the destination register is used between the two insns, too.
15047 2018-03-12 Richard Biener <rguenther@suse.de>
15049 PR tree-optimization/84803
15050 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
15051 for refs DR analysis didn't process.
15053 2018-03-12 Richard Biener <rguenther@suse.de>
15055 PR tree-optimization/84777
15056 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
15057 force-vectorize loops ignore whether we are optimizing for size.
15059 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
15061 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
15062 (TARGET_MD_ASM_ADJUST): Define.
15064 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
15065 Kito Cheng <kito.cheng@gmail.com>
15066 Chung-Ju Wu <jasonwucj@gmail.com>
15068 * config/nds32/nds32.c (nds32_compute_stack_frame,
15069 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
15070 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
15071 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
15072 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
15073 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
15074 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
15075 * config/nds32/nds32.md (prologue, epilogue): Use macro
15076 NDS32_V3PUSH_AVAILABLE_P to do checking.
15078 2018-03-11 Jakub Jelinek <jakub@redhat.com>
15081 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
15082 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
15083 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
15084 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
15085 addition of most attributes on !orig_type_die or the attribute not
15086 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
15088 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15089 Chung-Ju Wu <jasonwucj@gmail.com>
15091 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
15092 __NDS32_VH__ macro.
15093 * config/nds32/nds32.opt (mvh): New option.
15095 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15096 Chung-Ju Wu <jasonwucj@gmail.com>
15098 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
15100 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
15101 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
15104 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
15105 Chung-Ju Wu <jasonwucj@gmail.com>
15107 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
15109 * config/nds32/nds32-multiple.md (strlensi): New pattern.
15110 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
15112 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
15113 Kito Cheng <kito.cheng@gmail.com>
15114 Chung-Ju Wu <jasonwucj@gmail.com>
15116 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
15117 UNSPEC_FFMISM and UNSPEC_FLMISM.
15118 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
15119 for ffb, ffmism and flmism.
15120 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
15121 (unspec_ffmism): Ditto.
15122 (unspec_flmism): Ditto.
15123 (nds32_expand_builtin_impl): Check if string extension is available.
15124 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
15125 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
15127 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
15130 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
15133 * lra-assigns.c (assign_by_spills): Return a flag of reload
15134 assignment failure. Do not process the reload assignment
15135 failures. Do not spill other reload pseudos if they has the same
15137 (lra_assign): Add a return arg. Set up from the result of
15138 assign_by_spills call.
15139 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15140 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15141 usage_insns if it is not NULL.
15142 (spill_hard_reg_in_range): New function.
15143 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15144 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15145 function prototypes.
15146 (lra_assign): Change prototype.
15147 * lra.c (lra): Add code to deal with fails by splitting hard reg
15150 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
15153 * config/i386/i386.opt: Replace Enforcment with Enforcement.
15155 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
15158 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
15159 (dw_val_node): Add val_symbolic_view.
15160 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
15161 (symview_upper_bound): New.
15162 (new_line_info_table): Initialize symviews_since_reset.
15163 (dwarf2out_source_line): Count symviews_since_reset and set
15164 symview_upper_bound.
15165 (dw_val_equal_p): Handle symview.
15166 (add_AT_symview): New.
15167 (print_dw_val): Handle symview.
15168 (attr_checksum, attr_checksum_ordered): Likewise.
15169 (same_dw_val_p, size_of_die): Likewise.
15170 (value_format, output_die): Likewise.
15171 (add_high_low_attributes): Use add_AT_symview for entry_view.
15172 (dwarf2out_finish): Reset symview_upper_bound, clear
15175 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
15178 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
15179 Add strict argument and use it.
15180 (rs6000_split_multireg_move): Update for new strict argument.
15181 (mem_operand_gpr): Disallow all non-offsettable addresses.
15182 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
15184 2018-03-09 Jakub Jelinek <jakub@redhat.com>
15187 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
15188 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
15189 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
15192 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
15193 decl, use remap_type if we want to use the type.
15195 2018-03-09 Martin Sebor <msebor@redhat.com>
15197 PR tree-optimization/84526
15198 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
15200 (builtin_access::generic_overlap): Be prepared to handle non-array
15203 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
15205 PR rtl-optimization/84682
15206 * lra-constraints.c (process_address_1): Check is_address flag
15207 for address constraints.
15208 (process_alt_operands): Likewise.
15209 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
15210 preprocess_constraints.
15211 * recog.h (preprocess_constraints): Add oploc parameter.
15213 * recog.c (preprocess_constraints): Test address_operand for
15214 CT_ADDRESS constraints.
15216 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
15219 * lra-assigns.c (assign_by_spills): Return a flag of reload
15220 assignment failure. Do not process the reload assignment
15221 failures. Do not spill other reload pseudos if they has the same
15223 (lra_assign): Add a return arg. Set up from the result of
15224 assign_by_spills call.
15225 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
15226 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
15227 usage_insns if it is not NULL.
15228 (spill_hard_reg_in_range): New function.
15229 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
15230 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
15231 function prototypes.
15232 (lra_assign): Change prototype.
15233 * lra.c (lra): Add code to deal with fails by splitting hard reg
15236 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15239 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
15240 Accept complain bool parameter. Only emit errors if it is true.
15241 (arm_parse_cpu_option_name): Likewise.
15242 (arm_target_thumb_only): Adjust callers of the above.
15243 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
15244 prototype to take a default true bool parameter.
15245 (arm_parse_arch_option_name): Likewise.
15247 2018-03-09 David Malcolm <dmalcolm@redhat.com>
15248 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
15252 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
15253 * configure: Regenerate.
15254 * configure.ac ("linker --version-script option"): New.
15255 ("linker soname option"): New.
15257 2018-03-09 Richard Biener <rguenther@suse.de>
15259 PR tree-optimization/84775
15260 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
15261 immediate uses of predicate stmts and mark them modified.
15264 PR tree-optimization/84178
15265 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
15267 (version_loop_for_if_conversion): Delay update_ssa call.
15268 (tree_if_conversion): Delay update_ssa until after predicate
15271 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
15274 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
15275 when the function accesses prior frames.
15277 2018-03-08 Jakub Jelinek <jakub@redhat.com>
15280 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
15281 gen_llsym, otherwise call maybe_gen_llsym.
15283 PR inline-asm/84742
15284 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
15285 has ',' character inside of it.
15287 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15290 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
15291 as clobbering CC_REGNUM.
15293 2018-03-08 Richard Biener <rguenther@suse.de>
15295 PR middle-end/84552
15296 * tree-scalar-evolution.c: Include tree-into-ssa.h.
15297 (follow_copies_to_constant): Do not follow SSA names registered
15300 2018-03-08 Richard Biener <rguenther@suse.de>
15302 PR tree-optimization/84178
15303 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
15305 (version_loop_for_if_conversion): Delay update_ssa call.
15306 (tree_if_conversion): Delay update_ssa until after predicate
15309 2018-03-08 David Malcolm <dmalcolm@redhat.com>
15311 PR tree-optimization/84178
15312 * tree-if-conv.c (release_bb_predicate): Remove the
15313 the assertion that the stmts have NULL use_ops.
15314 Discard the statements, asserting that they haven't
15315 yet been added to a BB.
15317 2018-03-08 Richard Biener <rguenther@suse.de>
15319 PR tree-optimization/84746
15320 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
15321 (phi_translate): Pass in destination ANTIC_OUT set.
15322 (phi_translate_1): Likewise. For a simplified result lookup
15323 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
15324 (phi_translate_set): Adjust.
15325 (do_pre_regular_insertion): Likewise.
15326 (do_pre_partial_partial_insertion): Likewise.
15328 2018-03-08 Martin Liska <mliska@suse.cz>
15330 PR gcov-profile/84735
15331 * doc/gcov.texi: Document usage of profile files.
15332 * gcov-io.h: Document changes in the format.
15334 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
15338 * dwarf2out.c (struct dw_line_info_table): Update comments for
15340 (FORCE_RESET_NEXT_VIEW): New.
15341 (FORCE_RESETTING_VIEW_P): New.
15342 (RESETTING_VIEW_P): Check for -1 too.
15343 (ZERO_VIEW_P): Likewise.
15344 (new_line_info_table): Force-reset next view.
15345 (dwarf2out_begin_function): Likewise.
15346 (dwarf2out_source_line): Simplify zero_view_p initialization.
15347 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
15348 view directly. Omit view when omitting .loc at line 0.
15350 2018-03-08 Jakub Jelinek <jakub@redhat.com>
15352 PR tree-optimization/84740
15353 * tree-switch-conversion.c (process_switch): Call build_constructors
15354 only if info.phi_count is non-zero.
15356 PR tree-optimization/84739
15357 * tree-tailcall.c (find_tail_calls): Check call arguments against
15358 DECL_ARGUMENTS (current_function_decl) rather than
15359 DECL_ARGUMENTS (func) when checking for tail recursion.
15361 2018-03-07 Jakub Jelinek <jakub@redhat.com>
15363 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
15364 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
15365 Volker Reichelt's entry and add entries for people that perform
15366 GCC fuzzy testing and report numerous bugs.
15368 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
15371 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
15372 readonly data in sdata, if that is disabled.
15373 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
15374 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
15375 -mreadonly-in-sdata option.
15377 2018-03-07 Martin Sebor <msebor@redhat.com>
15379 PR tree-optimization/84468
15380 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
15381 basic block when looking for nul assignment.
15383 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
15386 * except.h (output_function_exception_table): Adjust prototype.
15387 * except.c (output_function_exception_table): Remove FNNAME parameter
15388 and add SECTION parameter. Ouput one part of the table at a time.
15389 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
15390 the first part of the exception table and emit unwind directives.
15391 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
15392 (i386_pe_seh_cold_init): Likewise.
15393 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
15394 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
15395 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
15396 (ix86_output_call_insn): Emit a nop in one more case for SEH.
15397 * config/i386/winnt.c: Include except.h.
15398 (struct seh_frame_state): Add reg_offset, after_prologue and
15399 in_cold_section fields.
15400 (i386_pe_seh_end_prologue): Set seh->after_prologue.
15401 (i386_pe_seh_cold_init): New function.
15402 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
15403 to seh->in_cold_section.
15404 (seh_emit_push): Record the offset of the push.
15405 (seh_emit_save): Record the offet of the save.
15406 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
15407 Test seh->after_prologue to disregard the epilogue.
15408 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
15409 (i386_pe_end_cold_function): New function.
15411 2018-03-07 Jakub Jelinek <jakub@redhat.com>
15414 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
15415 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
15418 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
15420 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
15421 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
15423 PR middle-end/84723
15424 * multiple_target.c: Include tree-inline.h and intl.h.
15425 (expand_target_clones): Diagnose and fail if node->definition and
15426 !tree_versionable_function_p (node->decl).
15428 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
15430 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
15432 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
15433 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
15434 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
15436 2018-03-06 Jakub Jelinek <jakub@redhat.com>
15439 * combine.c (try_combine): Use reg_or_subregno instead of handling
15440 just paradoxical SUBREGs and REGs.
15442 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
15444 * config/arc/arc.c (arc_finalize_pic): Remove function.
15445 (arc_must_save_register): We use single base PIC register, remove
15446 checks to save/restore the PIC register.
15447 (arc_expand_prologue): Likewise.
15448 * config/arc/arc-protos.h (arc_set_default_type_attributes):
15450 (arc_verify_short): Likewise.
15451 (arc_attr_type): Likewise.
15452 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
15453 (walk_stores): Likewise.
15454 (arc_address_cost): Make it static.
15455 (arc_verify_short): Likewise.
15456 (branch_dest): Likewise.
15457 (arc_attr_type): Likewise.
15458 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
15459 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
15460 (arc_final_prescan_insn): Remove inserting the nops due to
15461 hardware hazards. It is done in reorg step.
15462 (insn_length_variant_t): Remove.
15463 (insn_length_parameters_t): Likewise.
15464 (arc_insn_length_parameters): Likewise.
15465 (arc_get_insn_variants): Likewise.
15466 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
15468 2018-03-06 Jakub Jelinek <jakub@redhat.com>
15470 PR inline-asm/84683
15471 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
15474 PR tree-optimization/84687
15475 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
15477 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
15479 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15481 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
15482 Rename to ppc_speculation_barrier.
15483 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
15484 __builtin_ppc_speculation_barrier.
15486 2018-03-05 Jakub Jelinek <jakub@redhat.com>
15489 * combine.c (combine_simplify_rtx): Don't try to simplify if
15490 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
15493 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
15495 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
15496 to 32 bytes when compiling for POWER9.
15498 2018-03-05 Jakub Jelinek <jakub@redhat.com>
15501 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
15502 regparm >= 3 with no arg reg available also for calls with
15503 flag_force_indirect_call. Pass decl to ix86_function_regparm.
15506 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
15508 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
15510 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
15513 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
15515 2018-03-05 Richard Biener <rguenther@suse.de>
15517 PR tree-optimization/84486
15518 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
15519 When inserting a __builtin_assume_aligned call set the LHS
15520 SSA name alignment info accordingly.
15522 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
15524 PR tree-optimization/84114
15525 * config/aarch64/aarch64.c (aarch64_reassociation_width)
15526 Avoid reassociation of FLOAT_MODE addition.
15528 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
15530 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
15531 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
15532 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
15533 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
15534 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
15535 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
15536 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
15538 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
15539 __builtin_ia32_wbinvd): New builtins.
15540 (SPECIAL_ARGS2): New.
15541 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
15542 (SPECIAL_ARGS2): New.
15543 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
15544 (ix86_valid_target_attribute_inner_p): Ditto.
15545 (ix86_init_mmx_sse_builtins): Add special_args2.
15546 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
15547 TARGET_WBNOINVD_P): New.
15548 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
15549 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
15550 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
15551 * config/i386/immintrin.h (_wbinvd): New intrinsic.
15552 * config/i386/pconfigintrin.h: New file.
15553 * config/i386/wbnoinvdintrin.h: Ditto.
15554 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
15556 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
15558 2018-03-05 Richard Biener <rguenther@suse.de>
15560 PR tree-optimization/84670
15561 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
15563 (BB_VISITED_WITH_VISITED_SUCCS): New define.
15564 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
15565 (compute_antic_aux): Only assert the number of values in ANTIC_IN
15566 doesn't grow if all successors (recursively) were visited at least
15569 2018-03-05 Richard Biener <rguenther@suse.de>
15571 PR tree-optimization/84650
15572 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
15573 if executed in the loop pipeline.
15575 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
15577 * doc/configfiles.texi (Configuration Files): Move info about
15578 conditionalizing $target-protos.h to...
15579 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
15580 differs from $target-protos.h.
15582 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
15583 Chung-Ju Wu <jasonwucj@gmail.com>
15585 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
15586 * config/nds32/nds32-multiple.md (setmemsi): Define.
15587 * config/nds32/nds32-memory-manipulation.c
15588 (nds32_gen_dup_4_byte_to_word_value): New.
15589 (emit_setmem_word_loop): New.
15590 (emit_setmem_byte_loop): New.
15591 (nds32_expand_setmem_loop): New.
15592 (nds32_expand_setmem_loop_v3m): New.
15593 (nds32_expand_setmem_unroll): New.
15594 (nds32_expand_setmem): New.
15596 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15597 Chung-Ju Wu <jasonwucj@gmail.com>
15599 * config/nds32/nds32-memory-manipulation.c
15600 (nds32_emit_load_store): New.
15601 (nds32_emit_post_inc_load_store): New.
15602 (nds32_emit_mem_move): New.
15603 (nds32_emit_mem_move_block): New.
15604 (nds32_expand_movmemsi_loop_unknown_size): New.
15605 (nds32_expand_movmemsi_loop_known_size): New.
15606 (nds32_expand_movmemsi_loop): New.
15607 (nds32_expand_movmemsi_unroll): New.
15608 (nds32_expand_movmemqi): Rename ...
15609 (nds32_expand_movmemsi): ... to this.
15610 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
15611 (movmemsi): ... to this.
15612 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
15613 (nds32_expand_movmemsi): ... to this.
15615 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15616 Monk Chiang <sh.chiang04@gmail.com>
15617 Chung-Ju Wu <jasonwucj@gmail.com>
15619 * config/nds32/nds32-protos.h
15620 (nds32_expand_load_multiple): New arguments.
15621 (nds32_expand_store_multiple): Ditto.
15622 (nds32_valid_multiple_load_store): Rename ...
15623 (nds32_valid_multiple_load_store_p): ... to this.
15624 * config/nds32/nds32-memory-manipulation.c
15625 (nds32_expand_load_multiple): Refine implementation.
15626 (nds32_expand_store_multiple): Ditto.
15627 * config/nds32/nds32-multiple.md
15628 (load_multiple): Update nds32_expand_load_multiple interface.
15629 (store_multiple): Update nds32_expand_store_multiple interface.
15630 * config/nds32/nds32-predicates.c
15631 (nds32_valid_multiple_load_store): Rename ...
15632 (nds32_valid_multiple_load_store_p): ... to this and refine
15634 * config/nds32/predicates.md
15635 (nds32_load_multiple_and_update_address_operation): New predicate.
15636 (nds32_store_multiple_and_update_address_operation): New predicate.
15638 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
15639 Chung-Ju Wu <jasonwucj@gmail.com>
15641 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
15642 (combo): New attribute.
15643 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
15645 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
15647 * config/nds32/nds32.opt: Change -mcmodel= default value.
15649 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
15650 Monk Chiang <sh.chiang04@gmail.com>
15651 Chung-Ju Wu <jasonwucj@gmail.com>
15653 * config/nds32/constants.md (unspec_element): New enum.
15654 * config/nds32/constraints.md (Umw): New constraint.
15655 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
15656 * config/nds32/nds32-intrinsic.md: Likewise.
15657 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
15658 (nds32_valid_smw_lwm_base_p): New.
15659 (nds32_output_smw_single_word): New.
15660 (nds32_output_lmw_single_word): New.
15661 (nds32_expand_unaligned_load): New.
15662 (nds32_expand_unaligned_store): New.
15663 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
15664 (nds32_output_smw_single_word): Declare.
15665 (nds32_output_lmw_single_word): Declare.
15666 (nds32_expand_unaligned_load): Declare.
15667 (nds32_expand_unaligned_store): Declare.
15668 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
15669 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
15670 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
15671 NDS32_BUILTIN_UASTORE_DW.
15672 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
15675 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
15676 Kito Cheng <kito.cheng@gmail.com>
15677 Chung-Ju Wu <jasonwucj@gmail.com>
15679 * config/nds32/nds32-intrinsic.c
15680 (nds32_expand_builtin_null_ftype_reg): Delete.
15681 (nds32_expand_builtin_reg_ftype_imm): Ditto.
15682 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
15683 (nds32_read_argument): New.
15684 (nds32_legitimize_target): Ditto.
15685 (nds32_legitimize_argument): Ditto.
15686 (nds32_check_constant_argument): Ditto.
15687 (nds32_expand_unop_builtin): Ditto.
15688 (nds32_expand_unopimm_builtin): Ditto.
15689 (nds32_expand_binop_builtin): Ditto.
15690 (nds32_builtin_decl_impl): Ditto.
15691 (builtin_description): Ditto.
15692 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
15693 (nds32_init_builtins_impl): Ditto.
15694 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
15695 (nds32_builtin_decl): New.
15696 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
15697 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
15699 2018-03-02 Jeff Law <law@redhat.com>
15701 * reorg.c (stop_search_p): Handle DEBUG_INSN.
15702 (redundant_insn, fill_simple_delay_slots): Likewise.
15703 (fill_slots_from_thread): Likewise.
15704 * resource.c (mark_referenced_resources): Likewise.
15705 (mark_set_resources, find_dead_or_set_registers): Likewise.
15707 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15709 * substring-locations.h (format_warning_va): Formatting fix for
15710 ATTRIBUTE_GCC_DIAG.
15711 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
15713 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
15714 * substring-locations.c: Include intl.h.
15715 (format_warning_va): Turned into small wrapper around
15716 format_warning_n_va, renamed to ...
15717 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
15718 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
15720 (format_warning_at_substring_n): New function.
15721 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
15722 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
15723 format_warning_at_substring with just a shorter name instead of
15724 const function pointer.
15725 (fmtwarn_n): New function.
15726 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
15727 appropriate, get rid of all the fmtstr temporaries, move conditionals
15728 with G_() wrapped string literals directly into fmtwarn arguments,
15729 cast dir.len to (int), formatting fixes.
15731 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
15733 * doc/invoke.texi: Remove "Cilk Plus" references.
15735 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15736 Richard Biener <rguenther@suse.de>
15739 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
15740 for error or warning attributes if CALL_FROM_THUNK_P is set.
15743 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15746 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
15747 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
15750 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
15751 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
15753 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
15754 instead of -1U in last predictors element's probability member.
15756 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
15759 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
15760 arguments if they are comparable.
15762 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15764 PR tree-optimization/84634
15765 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
15766 masks and masked_loop_p with a single loop_masks, making sure it's
15767 null for bb vectorization.
15769 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
15771 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
15772 (vect_analyze_data_ref_access): Use loop->safe_len rather than
15773 loop->force_vectorize to check whether there is no alias.
15775 2018-03-02 Jakub Jelinek <jakub@redhat.com>
15778 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
15780 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
15782 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
15783 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
15784 instead of a loop around prev_real_insn.
15785 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
15788 PR inline-asm/84625
15789 * config/i386/i386.c (ix86_print_operand): Use conditional
15790 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
15793 2018-03-02 Richard Biener <rguenther@suse.de>
15795 PR tree-optimization/84427
15796 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
15797 (bitmap_set_subtract_values): Rewrite to handle multiple
15800 (prune_clobbered_mems): Likewise.
15801 (phi_translate): Take edge instead of pred/phiblock.
15802 (phi_translate_1): Likewise.
15803 (phi_translate_set): Likewise. Insert all translated
15804 exprs for a value into the set, keeping possibly multiple
15805 expressions per value.
15806 (compute_antic_aux): Adjust for phi_translate changes.
15807 When intersecting union the expressions and prune those
15808 not in the final value set, keeping possibly multiple
15809 expressions per value. Do not use value-insertion
15810 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
15811 all expressions. Add verification that the value-sets
15812 only shrink during iteration.
15813 (compute_partial_antic_aux): Adjust for the phi_translate changes.
15814 (do_pre_regular_insertion): Likewise.
15815 (do_pre_partial_partial_insertion): Likewise.
15817 2018-03-02 Richard Biener <rguenther@suse.de>
15820 * config/darwin.c (saved_debug_info_level): New static global.
15821 (darwin_asm_lto_start): Disable debug info generation for LTO out.
15822 (darwin_asm_lto_end): Restore debug info generation settings.
15824 2018-03-01 Martin Liska <mliska@suse.cz>
15827 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
15828 volatile arguments.
15830 2018-03-01 Richard Biener <rguenther@suse.de>
15833 * dwarf2out.c (gen_variable_die): Properly handle late VLA
15834 type annotation with LTO when debug was disabled at compile-time.
15836 2018-03-01 Matthew Fortune <mfortune@gmail.com>
15838 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
15840 (mips_final_postscan_insn): Likewise.
15842 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
15844 PR rtl-optimization/84528
15845 * alias.c (init_alias_target): Add commentary.
15846 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
15847 a unique base value if the frame pointer is not eliminated
15848 to the stack pointer.
15850 2018-03-01 Tom de Vries <tom@codesourcery.com>
15852 PR rtl-optimization/83327
15853 * lra-int.h (hard_regs_spilled_into): Declare.
15854 * lra.c (hard_regs_spilled_into): Define.
15855 (init_reg_info): Init hard_regs_spilled_into.
15856 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
15857 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
15858 (process_bb_lives): Handle hard_regs_spilled_into.
15859 (lra_create_live_ranges_1): Before doing liveness propagation, clear
15860 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
15862 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
15864 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
15865 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
15866 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
15867 * config/rs6000/aix72.h: New file.
15869 2018-02-28 Jakub Jelinek <jakub@redhat.com>
15871 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
15872 instead of warning_at with conditional singular and plural messages
15876 * stor-layout.c (update_alignment_for_field): For
15877 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
15878 && !DECL_PACKED (field), do the alignment update, just use
15879 only desired_align instead of MAX (type_align, desired_align)
15881 (place_field): Don't do known_align < desired_align handling
15882 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
15883 is non-NULL, instead do it after rli->prev_field handling and
15884 only if not within a bitfield word. For DECL_PACKED (field)
15885 use type_align of BITS_PER_UNIT.
15887 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
15889 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
15890 superfluous parentheses and trailing spaces.
15892 2018-02-28 Richard Biener <rguenther@suse.de>
15894 PR tree-optimization/84584
15895 * graphite-scop-detection.c (scop_detection::add_scop): Discard
15896 SCoPs with fake exit edge.
15898 2018-02-28 Martin Liska <mliska@suse.cz>
15901 * timevar.c (timer::print): Fix format to properly print 100%
15904 2018-02-28 Richard Biener <rguenther@suse.de>
15906 PR middle-end/84607
15907 * genmatch.c (capture_info::walk_match): Do not mark
15908 captured expressions without operands as expr_p given
15909 they act more like predicates and should be subject to
15910 "lost tail" side-effect preserving.
15912 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
15914 PR rtl-optimization/81611
15915 * auto-inc-dec.c (attempt_change): Move dead note from
15916 mem_insn if it's the next use of regno
15917 (find_address): Take address use of reg holding
15918 non-incremented value. Add parm to limit search to the named
15920 (merge_in_block): Attempt to use a mem insn that is the next
15921 use of the original regno.
15923 2018-02-27 Martin Sebor <msebor@redhat.com>
15926 * doc/invoke.texi (-Wmissing-attributes): New option.
15927 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
15929 2018-02-27 Martin Sebor <msebor@redhat.com>
15931 PR translation/84207
15932 * diagnostic-core.h (warning_n, error_n, inform_n): Change
15933 n argument to unsigned HOST_WIDE_INT.
15934 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
15935 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
15936 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
15937 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
15939 2018-02-27 Richard Biener <rguenther@suse.de>
15941 PR tree-optimization/84512
15942 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15943 Do not use the estimate returned from record_stmt_cost for
15944 the scalar iteration cost but sum properly using add_stmt_cost.
15946 2018-02-27 Richard Biener <rguenther@suse.de>
15948 PR tree-optimization/84466
15949 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
15950 Adjust last change to less strictly validate use operands.
15952 2018-02-27 Martin Liska <mliska@suse.cz>
15954 PR gcov-profile/84548
15955 * gcov.c (process_file): Allow partial overlap and consider it
15956 also as group functions.
15957 (output_lines): Properly calculate range of lines for a group.
15959 2018-02-27 Martin Liska <mliska@suse.cz>
15961 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
15962 'ggc' suffixes. Change first column width.
15963 (timer::print): Fix formatting of the column.
15965 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
15967 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
15968 preserve inline entry blocks for the sake of debug inline
15969 entry point markers alone.
15970 (remove_unused_locals): Suggest in comments a better place to
15971 force the preservation of inline entry blocks that are
15972 otherwise unused, but do not preserve them.
15974 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15976 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
15978 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
15981 * config/i386/constraints.md (Bs): Replace
15982 ix86_indirect_branch_register with
15983 TARGET_INDIRECT_BRANCH_REGISTER.
15985 * config/i386/i386.md (indirect_jump): Likewise.
15986 (tablejump): Likewise.
15987 (*sibcall_memory): Likewise.
15988 (*sibcall_value_memory): Likewise.
15989 Peepholes of indirect call and jump via memory: Likewise.
15990 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
15991 (*sibcall_value_GOT_32): Likewise.
15992 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15993 (GOT_memory_operand): Likewise.
15994 (call_insn_operand): Likewise.
15995 (sibcall_insn_operand): Likewise.
15996 (GOT32_symbol_operand): Likewise.
15997 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
15999 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
16001 PR rtl-optimization/83496
16002 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
16003 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
16004 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
16005 redundant insn, if any.
16006 (relax_delay_slots): Likewise.
16007 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
16009 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
16011 PR tree-optimization/83965
16012 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
16013 that grouped statements are part of a reduction chain. Return
16014 true if the statement is not marked as a reduction itself but
16015 is part of a group.
16016 (vect_recog_dot_prod_pattern): Don't check whether the statement
16017 is part of a group here.
16018 (vect_recog_sad_pattern): Likewise.
16019 (vect_recog_widen_sum_pattern): Likewise.
16021 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
16024 * final.c (rest_of_clean_state): Also look for calls inside sequences.
16026 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
16029 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
16031 (ix86_output_indirect_function_return): New prototype.
16032 (ix86_split_simple_return_pop_internal): Likewise.
16033 * config/i386/i386.c (indirect_return_via_cx): New.
16034 (indirect_return_via_cx_bnd): Likewise.
16035 (indirect_thunk_name): Handle return va CX_REG.
16036 (output_indirect_thunk_function): Create alias for
16037 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
16038 (ix86_output_indirect_jmp): Remove the bool argument.
16039 (ix86_output_indirect_function_return): New function.
16040 (ix86_split_simple_return_pop_internal): Likewise.
16041 * config/i386/i386.md (*indirect_jump): Don't pass false
16042 to ix86_output_indirect_jmp.
16043 (*tablejump_1): Likewise.
16044 (simple_return_pop_internal): Change it to define_insn_and_split.
16045 Call ix86_split_simple_return_pop_internal to split it for
16046 -mfunction-return=.
16047 (simple_return_indirect_internal): Call
16048 ix86_output_indirect_function_return instead of
16049 ix86_output_indirect_jmp.
16051 2018-02-26 Jakub Jelinek <jakub@redhat.com>
16054 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
16055 memset and value initialization afterwards.
16057 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
16059 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
16061 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16064 * common/config/aarch64/aarch64-common.c
16065 (aarch_option_optimization_table[]): Switch
16066 off fomit-frame-pointer
16068 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
16069 Chung-Ju Wu <jasonwucj@gmail.com>
16071 * config/nds32/nds32-multiple.md (load_multiple): Disallow
16073 (store_multiple): Ditto.
16075 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
16077 * config.gcc: Add --with-cpu support for nds32 target.
16078 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
16079 * config/nds32/nds32.opt: Add -mcpu= option.
16081 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
16083 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
16084 isel=yes): Warn for these deprecated options.
16086 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
16088 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
16089 ISA_2_5_MASKS_EMBEDDED.
16091 2018-02-23 Jakub Jelinek <jakub@redhat.com>
16093 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
16094 p->max as pointers rather than using iterative_hash_expr.
16096 2018-02-23 Carl Love <cel@us.ibm.com>
16098 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
16099 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
16101 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
16102 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
16103 P8V_BUILTIN_VEC_VUNSIGNED2.
16105 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
16108 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
16109 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
16111 * lra-constraints.c (curr_insn_transform): Set up
16112 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
16114 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
16115 * lra-eliminations.c (spill_pseudos): Ditto.
16116 (process_insn_for_elimination): Ditto.
16117 * lra-lives.c (reg_early_clobber_p): Use the new macros.
16118 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
16119 LRA_NON_CLOBBERED_ALT.
16121 2018-02-22 Martin Sebor <msebor@redhat.com>
16123 PR tree-optimization/84480
16124 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
16125 to maybe_diag_stxncpy_trunc. Call it.
16126 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
16127 from gimple_fold_builtin_strcpy. Print inlining stack.
16128 (handle_builtin_stxncpy): Print inlining stack.
16129 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
16131 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
16134 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
16135 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
16136 and -fcheck-pointer-bounds are used together.
16137 (indirect_thunk_prefix): New enum.
16138 (indirect_thunk_need_prefix): New function.
16139 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
16140 "_nt" instead of "_bnd" for NOTRACK prefix.
16141 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
16142 (output_indirect_thunk_function): Likewise.
16144 (ix86_code_end): Update output_indirect_thunk_function calls.
16145 (ix86_output_indirect_branch_via_reg): Replace
16146 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
16147 (ix86_output_indirect_branch_via_push): Likewise.
16148 (ix86_output_function_return): Likewise.
16149 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
16150 incompatible with -fcf-protection=branch and
16151 -fcheck-pointer-bounds.
16153 2018-02-22 Steve Ellcey <sellcey@cavium.com>
16156 * config/aarch64/aarch64.c (aarch64_print_address_internal):
16157 Change gcc_assert call to output_operand_lossage.
16159 2018-02-22 Steve Ellcey <sellcey@cavium.com>
16161 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
16163 2018-02-22 DJ Delorie <dj@redhat.com>
16164 Sebastian Perta <sebastian.perta@renesas.com>
16165 Oleg Endo <olegendo@gcc.gnu.org>
16167 * config/rx/rx.c (rx_rtx_costs): New function.
16168 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
16170 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
16172 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
16174 2018-02-22 Martin Liska <mliska@suse.cz>
16177 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
16178 Add "native" as a possible value.
16180 2018-02-22 Martin Liska <mliska@suse.cz>
16183 * config/i386/i386.c (ix86_option_override_internal):
16184 Add "native" as a possible value for -march and -mtune.
16186 2018-02-22 Jakub Jelinek <jakub@redhat.com>
16189 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
16190 to all type variants.
16192 PR tree-optimization/84503
16193 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
16194 width as info->bitpos + info->bitsize - start.
16195 (merged_store_group::merge_overlapping): Simplify width computation.
16196 (check_no_overlap): New function.
16197 (imm_store_chain_info::try_coalesce_bswap): Compute expected
16198 start + width and last_order of the group, fail if check_no_overlap
16200 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
16201 to group if check_no_overlap fails.
16203 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
16205 * config/rs6000/altivec.md: Delete contraint arguments to
16206 define_expand, define_split, and define_peephole2, and in
16207 define_insn_and_split if always unused.
16208 * config/rs6000/darwin.md: Ditto.
16209 * config/rs6000/dfp.md: Ditto.
16210 * config/rs6000/rs6000.md: Ditto.
16211 * config/rs6000/sync.md: Ditto.
16212 * config/rs6000/vector.md: Ditto.
16213 * config/rs6000/vsx.md: Ditto.
16215 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
16217 * config/rs6000/altivec.md: Write output control strings as braced
16218 blocks instead of double-quoted strings.
16219 * config/rs6000/darwin.md: Ditto.
16220 * config/rs6000/rs6000.md: Ditto.
16221 * config/rs6000/vector.md: Ditto.
16222 * config/rs6000/vsx.md: Ditto.
16224 2018-02-21 Jason Merrill <jason@redhat.com>
16226 PR c++/84314 - ICE with templates and fastcall attribute.
16227 * attribs.c (build_type_attribute_qual_variant): Remove assert.
16229 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
16231 * ipa-cp.c (determine_versionability): Fix comment typos.
16233 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
16236 * ipa-cp.c (determine_versionability): Do not version functions caling
16239 2018-02-21 Martin Liska <mliska@suse.cz>
16242 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
16243 Add "native" as a possible value.
16244 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
16245 the macro when native cpu detection is available.
16247 2018-02-21 Martin Liska <mliska@suse.cz>
16250 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
16251 Add "native" as a possible value.
16252 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
16253 when native cpu detection is available.
16255 2018-02-21 Jakub Jelinek <jakub@redhat.com>
16256 Martin Sebor <msebor@redhat.com>
16258 PR tree-optimization/84478
16259 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
16261 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
16262 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
16263 support which is conservatively correct, for 2 only stay conservative
16264 for maxlen. Formatting and comment capitalization fixes. Add STRICT
16265 argument to the 2 argument get_range_strlen, adjust 6 arg
16266 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
16268 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
16269 (gimple_fold_builtin_strlen): Pass true as last argument to
16272 2018-02-20 Martin Sebor <msebor@redhat.com>
16274 PR middle-end/84095
16275 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
16276 (builtin_memref::set_base_and_offset): Same. Handle inner references.
16277 (builtin_memref::builtin_memref): Factor out parts into
16278 set_base_and_offset and call it.
16280 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
16282 PR middle-end/84406
16283 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
16284 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
16285 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
16286 search at the associated MODE_INT.
16288 2018-02-20 Jeff Law <law@redhat.com>
16290 PR middle-end/82123
16291 PR tree-optimization/81592
16292 PR middle-end/79257
16293 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
16294 for range data rather than using global data.
16295 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
16296 range data rather than using global data.
16297 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
16298 pass it to children as needed.
16299 (struct directive::fmtresult): Similarly.
16300 (struct directive::set_width): Similarly.
16301 (struct directive::set_precision): Similarly.
16302 (format_integer, format_directive, parse_directive): Similarly.
16303 (format_none): Accept unnamed vr_values parameter.
16304 (format_percent, format_floating, format_character): Similarly.
16305 (format_string, format_plain): Similarly.
16306 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
16307 the EVRP range analyzer for range data rather than using global data.
16308 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
16309 gimple-ssa-evrp-analyze.h
16310 (class sprintf_dom_walker): Add after_dom_children member function.
16311 Add evrp_range_analyzer member.
16312 (sprintf_dom_walker::before_dom_children): Call into the EVRP
16313 range analyzer as needed.
16314 (sprintf_dom_walker::after_dom_children): New member function.
16315 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
16317 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
16318 (evrp_range_analyzer::pop_to_marker): Likewise.
16320 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
16322 PR tree-optimization/84419
16323 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
16324 with the required type if its current type is compatible but
16327 2018-02-20 Jakub Jelinek <jakub@redhat.com>
16329 PR middle-end/82004
16330 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
16331 after vectorization.
16333 2018-02-20 Martin Liska <mliska@suse.cz>
16336 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
16337 possible values if we don't have a hint.
16339 2018-02-20 Martin Liska <mliska@suse.cz>
16343 * final.c (shorten_branches): Build align_tab array with one
16345 * opts.c (finish_options): Add alignment option limit check.
16346 (MAX_CODE_ALIGN): Likewise.
16347 (MAX_CODE_ALIGN_VALUE): Likewise.
16348 * doc/invoke.texi: Document maximum allowed option value for
16349 all -falign-* options.
16351 2018-02-19 Jakub Jelinek <jakub@redhat.com>
16354 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
16355 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
16356 * var-tracking.c (emit_note_insn_var_location): Remove all references
16357 to NOTE_INSN_CALL_ARG_LOCATION.
16358 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
16359 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
16360 Use copy_rtx_if_shared.
16361 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
16362 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
16363 (dwarf2out_var_location): Remove handling of
16364 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
16366 * final.c (final_scan_insn): Remove all references to
16367 NOTE_INSN_CALL_ARG_LOCATION.
16368 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
16369 before dumping final insns.
16370 * except.c (emit_note_eh_region_end): Remove all references to
16371 NOTE_INSN_CALL_ARG_LOCATION.
16372 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
16373 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
16374 * config/arc/arc.c (hwloop_optimize): Likewise.
16375 * config/arm/arm.c (create_fix_barrier): Likewise.
16376 * config/s390/s390.c (s390_chunkify_start): Likewise.
16377 * config/sh/sh.c (find_barrier): Likewise.
16378 * config/i386/i386.c (rest_of_insert_endbranch,
16379 ix86_seh_fixup_eh_fallthru): Likewise.
16380 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
16381 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
16382 * config/frv/frv.c (frv_function_prologue): Likewise.
16383 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
16385 (note_outside_basic_block_p): Remove all references to
16386 NOTE_INSN_CALL_ARG_LOCATION.
16387 * gengtype.c (adjust_field_rtx_def): Likewise.
16388 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
16390 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
16391 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
16394 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
16397 PR tree-optimization/84452
16398 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
16399 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
16402 2018-02-19 Martin Liska <mliska@suse.cz>
16405 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
16407 2018-02-19 Martin Liska <mliska@suse.cz>
16408 Richard Sandiford <richard.sandiford@linaro.org>
16410 PR tree-optimization/82491
16411 * gimple-fold.c (get_base_constructor): Make earlier bail out
16414 2018-02-19 Carl Love <cel@us.ibm.com>
16416 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
16417 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
16419 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
16420 P8V_BUILTIN_VEC_NEG.
16422 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
16424 * config/rl78/rl78.md (movdf): New define expand.
16426 2018-02-19 Martin Liska <mliska@suse.cz>
16429 * doc/invoke.texi: Fix typo.
16430 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
16432 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
16434 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
16435 handle rs6000_single_float and rs6000_double_float specially for
16438 2018-02-16 Jeff Law <law@redhat.com>
16440 * config/rx/rx.c (add_pop_cfi_notes): New function.;
16441 (pop_regs): Use it.
16443 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16446 * ipa-inline.c (inline_small_functions): Fix a typo.
16448 2018-02-16 Nathan Sidwell <nathan@acm.org>
16450 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
16452 2018-02-16 Carl Love <cel@us.ibm.com>
16454 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
16455 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
16456 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
16457 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
16458 expansion to P8V_BUILTIN_VEC_FLOAT2.
16460 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
16462 PR rtl-optimization/70023
16463 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
16464 src_regno into account.
16466 2018-02-16 Carl Love <cel@us.ibm.com>
16468 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
16469 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
16470 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
16471 * config/rs6000/rs6000.c: Remove case statements for
16472 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
16473 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
16474 and P9V_BUILTIN_VEC_VINSERT4B.
16475 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
16476 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
16477 * config/rs6000/vsx.md:
16478 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
16481 2018-02-16 Carl Love <cel@us.ibm.com>
16483 * config/rs6000/altivec.h: Add builtin names vec_extract4b
16485 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
16487 * config/rs6000/rs6000-c.c: Add the definitions for
16488 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
16489 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
16490 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
16491 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
16492 definition for insert4b and define insn *insert3b_internal.
16493 * doc/extend.texi: Add documentation for vec_extract4b.
16495 2018-02-16 Nathan Sidwell <nathan@acm.org>
16497 * doc/extend.texi (Backwards Compatibility): Mention friend
16498 injection. Note for-scope is deprecated.
16499 * doc/invoke.texi (-ffriend-injection): Deprecate.
16501 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
16503 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
16504 that moved to I2, also allow destinations that are a paradoxical
16505 subreg (instead of a normal reg).
16507 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
16510 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
16513 2018-02-16 Richard Biener <rguenther@suse.de>
16515 PR tree-optimization/84037
16516 PR tree-optimization/84016
16518 * config/i386/i386.c (ix86_builtin_vectorization_cost):
16519 Adjust vec_construct for the fact we need additional higher latency
16520 128bit inserts for AVX256 and AVX512 vector builds.
16521 (ix86_add_stmt_cost): Scale vector construction cost for
16524 2018-02-16 Richard Biener <rguenther@suse.de>
16526 PR tree-optimization/84417
16527 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
16528 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
16529 (non_rewritable_lvalue_p): Likewise, use poly-ints.
16531 2018-02-16 Martin Liska <mliska@suse.cz>
16534 * internal-fn.def (ASAN_CHECK): Set proper flags.
16535 (ASAN_MARK): Likewise.
16537 2018-02-16 Julia Koval <julia.koval@intel.com>
16539 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
16540 from PTA_CANNONLAKE.
16542 2018-02-16 Jakub Jelinek <jakub@redhat.com>
16545 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
16546 Use ++iter rather than iter++ for std::list iterators.
16547 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
16548 defer deleting them until all nodes in the forest are processed. Do
16549 free even leaf nodes. Change to_process into auto_vec.
16552 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
16553 * vec.h (vec_default_construct): Use memset instead of placement new
16554 if BROKEN_VALUE_INITIALIZATION is defined.
16555 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
16556 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
16559 PR rtl-optimization/83723
16560 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
16561 * lra.c (lra_substitute_pseudo): Likewise. If true, use
16562 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
16564 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
16566 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
16568 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
16570 PR rtl-optimization/81443
16571 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
16572 from inner REGs to paradoxical SUBREGs.
16574 2018-02-16 Richard Biener <rguenther@suse.de>
16576 PR tree-optimization/84399
16577 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
16578 For operands we can analyze at their definition make sure we can
16579 analyze them at each use as well.
16581 2018-02-16 Richard Biener <rguenther@suse.de>
16583 PR tree-optimization/84190
16584 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
16585 volatile accesses if the decl isn't volatile.
16587 2018-02-15 Jason Merrill <jason@redhat.com>
16589 PR c++/84314 - ICE with templates and fastcall attribute.
16590 * attribs.c (build_type_attribute_qual_variant): Don't clobber
16591 TYPE_CANONICAL on an existing type.
16593 2018-02-15 Jakub Jelinek <jakub@redhat.com>
16595 PR tree-optimization/84383
16596 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
16597 dstoff nor call operand_equal_p if dstbase is NULL.
16599 PR tree-optimization/84334
16600 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
16601 also a CONSTANT_CLASS_P, punt.
16603 2018-02-14 Jim Wilson <jimw@sifive.com>
16605 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
16606 first SMALL_OPERAND check. New local min_second_step. Move assert
16607 to where locals are set. Add TARGET_RVC support.
16608 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
16610 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
16612 * doc/invoke.texi: Correct -Wformat-overflow code sample.
16614 2018-02-14 Martin Sebor <msebor@redhat.com>
16616 PR tree-optimization/83698
16617 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
16618 arrays constrain the offset range to their bounds.
16619 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
16620 (builtin_access::overlap): Avoid setting the size of overlap if it's
16622 (maybe_diag_overlap): Also consider arrays when deciding what values
16623 of offsets to include in diagnostics.
16625 2018-02-14 Martin Sebor <msebor@redhat.com>
16628 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
16629 that correspond to the kind of a declaration.
16631 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
16634 * config/pa/pa.md: Load address of PIC label using the linkage table
16635 if the label is nonlocal.
16637 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
16639 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
16640 warning message if user requests -maltivec=be.
16641 * doc/invoke.texi: Document deprecation of -maltivec=be.
16643 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
16646 * config/rs6000/rs6000-c.c: Update definitions for
16647 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
16648 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
16650 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
16653 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
16654 add _get_ssp intrinsics. Remove argument from
16655 __builtin_ia32_rdssp[d|q].
16656 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
16657 * config/i386/i386-builtin.def: Remove argument from
16658 __builtin_ia32_rdssp[d|q].
16659 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
16660 ix86_expand_special_args_builtin for _rdssp[d|q].
16661 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
16662 Clear register before usage.
16663 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
16664 Add documentation for new _get_ssp and _inc_ssp intrinsics.
16666 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
16668 PR tree-optimization/84357
16669 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
16670 operand 1 of an ARRAY_REF too.
16672 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
16675 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
16676 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16678 (set_of_reg): New struct.
16679 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
16680 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
16681 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
16683 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
16684 Split into bitclr, bitset, bitinvert patterns if appropriate.
16685 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
16686 use rx_fuse_in_memory_bitop.
16687 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
16688 to named insn, correct maximum insn length.
16690 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
16693 * machmode.def: Define a complex mode for PARTIAL_INT.
16694 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
16696 * doc/rtl.texi: Document CSPImode.
16697 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
16699 (msp430_hard_regno_nregs_with_padding): Likewise.
16701 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
16704 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
16706 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
16708 PR rtl-optimization/84169
16709 * combine.c (try_combine): New variable split_i2i3. Set it to true if
16710 we generated a parallel as new i3 and we split that to new i2 and i3
16711 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
16712 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
16713 those to i2, not i1. Partially rewrite this scan code.
16715 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16718 * stor-layout.c (place_field): For variable length fields, adjust
16719 offset_align afterwards not just based on the field's alignment,
16720 but also on the size.
16722 PR middle-end/84309
16723 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
16724 of exps and logs in the use_exp2 case.
16726 2018-02-13 Jeff Law <law@redhat.com>
16728 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
16729 entry for "vector".
16731 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
16734 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
16738 * common.opt (gas-loc-support, gas-locview-support): New.
16739 (ginline-points, ginternal-reset-location-views): New.
16740 * doc/invoke.texi: Document them. Use @itemx where intended.
16741 (gvariable-location-views): Adjust.
16742 * target.def (reset_location_view): New.
16743 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
16744 (TARGET_RESET_LOCATION_VIEW): New.
16745 * doc/tm.texi: Rebuilt.
16746 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
16747 (dwarf2out_default_as_locview_support): New.
16748 (output_asm_line_debug_info): Use option variables.
16749 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
16750 (output_loc_list): Likewise.
16751 (add_high_low_attributes): Check option variables.
16752 Don't output entry view attribute in strict mode.
16753 (gen_inlined_subroutine_die): Check option variables.
16754 (dwarf2out_inline_entry): Likewise.
16755 (init_sections_and_labels): Likewise.
16756 (dwarf2out_early_finish): Likewise.
16757 (maybe_reset_location_view): New, from...
16758 (dwarf2out_var_location): ... here. Call it.
16759 * debug.h (dwarf2out_default_as_loc_support): Declare.
16760 (dwarf2out_default_as_locview_support): Declare.
16761 * hooks.c (hook_int_rtx_insn_0): New.
16762 * hooks.h (hook_int_rtx_insn_0): Declare.
16763 * toplev.c (process_options): Take -gas-loc-support and
16764 -gas-locview-support from dwarf2out. Enable
16765 -gvariable-location-views by default only with locview
16766 assembler support. Enable -ginternal-reset-location-views by
16767 default only if the target defines the corresponding hook.
16768 Enable -ginline-points by default if location views are
16769 enabled; force it disabled if statement frontiers are
16771 * tree-inline.c (expand_call_inline): Check option variables.
16772 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
16774 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
16776 PR tree-optimization/84321
16777 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
16778 handling. Also check whether the anti-range contains any values
16779 that satisfy the mask; switch to a VR_RANGE if not.
16781 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
16784 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
16786 2018-02-13 Martin Jambor <mjambor@suse.cz>
16789 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
16790 of call statements, also set location of a load to a temporary.
16792 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
16794 * config/rl78/rl78.c (add_vector_labels): New function.
16795 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
16796 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
16797 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
16798 which checks that no arguments are passed.
16799 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
16800 * doc/extend.texi: Documentation for the new attribute.
16802 2018-02-13 Andreas Schwab <schwab@suse.de>
16804 * config/riscv/linux.h (CPP_SPEC): Define.
16806 2018-02-13 Jakub Jelinek <jakub@redhat.com>
16809 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
16810 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
16811 OPTION_MASK_ISA_AES as first argument to def_builtin_const
16812 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
16813 instead of OPTION_MASK_ISA_PCLMUL as first argument to
16814 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
16815 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
16816 temporarily for AES and PCLMUL builtins.
16818 PR tree-optimization/84339
16819 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
16820 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
16823 PR middle-end/84309
16824 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
16825 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
16826 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
16828 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
16830 * omp-simd-clone.h: New file.
16831 * omp-simd-clone.c: Include omp-simd-clone.h.
16832 (expand_simd_clones): No longer static.
16833 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
16834 cgraph.h and omp-simd-clone.h.
16835 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
16836 (vect_recog_widen_shift_pattern): Formatting fix.
16837 (vect_pattern_recog_1): Don't check optab for calls.
16840 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
16841 operands[2] into a REG before using gen_lowpart on it.
16843 2018-02-12 Jeff Law <law@redhat.com>
16846 * config/sh/sh.c (find_barrier): Consider a sibling call
16849 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
16850 successfully back substituting a reg.
16852 2018-02-12 Richard Biener <rguenther@suse.de>
16854 PR tree-optimization/84037
16855 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
16856 parameter, move visited init to caller.
16857 (vect_slp_analyze_operations): Separate cost from validity
16858 check, initialize visited once for all instances.
16859 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
16861 * tree-vect-stmts.c (vect_model_simple_cost): Make early
16863 (vect_model_promotion_demotion_cost): Likewise.
16864 (vectorizable_bswap): Guard cost modeling with !slp_node
16865 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
16867 (vectorizable_call): Likewise.
16868 (vectorizable_conversion): Likewise.
16869 (vectorizable_assignment): Likewise.
16870 (vectorizable_shift): Likewise.
16871 (vectorizable_operation): Likewise.
16872 (vectorizable_store): Likewise.
16873 (vectorizable_load): Likewise.
16874 (vectorizable_condition): Likewise.
16875 (vectorizable_comparison): Likewise.
16877 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
16880 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
16881 (ASAN_MARK): Fix fnspec to account for return value, change pointer
16882 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
16884 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
16886 PR middle-end/83665
16887 * params.def (inline-min-speedup): Increase from 8 to 15.
16888 (max-inline-insns-auto): Decrease from 40 to 30.
16889 * ipa-split.c (consider_split): Add some buffer for function to
16890 be considered inlining candidate.
16891 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
16894 2018-02-12 Richard Biener <rguenther@suse.de>
16896 PR tree-optimization/84037
16897 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
16898 matched stmts if we cannot swap the non-matched ones.
16900 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16902 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
16903 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
16904 _mm_maskz_scalef_round_ss): New intrinsics.
16905 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
16906 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
16907 __builtin_ia32_scalefss_round): Remove.
16908 (__builtin_ia32_scalefsd_mask_round,
16909 __builtin_ia32_scalefss_mask_round): New intrinsics.
16910 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
16911 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
16912 ((match_operand:VF_128 2 "<round_nimm_predicate>"
16913 "<round_constraint>")): Changed to ...
16914 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
16915 "<round_scalar_constraint>")): ... this.
16916 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
16917 %0, %1, %2<round_op3>}"): Changed to ...
16918 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
16919 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
16920 %2<round_scalar_mask_op3>}"): ... this.
16921 * config/i386/subst.md (round_scalar_nimm_predicate): New.
16923 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
16925 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
16926 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
16927 (_mm_maskz_sqrt_round_ss): New intrinsics.
16928 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
16929 (__builtin_ia32_sqrtsd_mask_round)
16930 (__builtin_ia32_sqrtss_mask_round): New builtins.
16931 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
16932 (__builtin_ia32_sqrtss_round): Remove.
16933 (__builtin_ia32_sqrtsd_mask_round)
16934 (__builtin_ia32_sqrtss_mask_round): New builtins.
16935 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
16936 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
16937 ((match_operand:VF_128 1 "vector_operand"
16938 "xBm,<round_constraint>")): Changed to ...
16939 ((match_operand:VF_128 1 "vector_operand"
16940 "xBm,<round_scalar_constraint>")): ... this.
16941 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
16942 %0, %2, %<iptr>1<round_op3>}): Changed to ...
16943 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
16944 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
16945 %<iptr>1<round_scalar_mask_op3>}): ... this.
16946 ((set_attr "prefix" "<round_prefix>")): Changed to ...
16947 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
16949 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
16952 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
16953 Cast vec_cmpeq result to correct type.
16954 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
16955 Cast vec_cmpgt result to correct type.
16957 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
16959 * final.c (final_scan_insn_1): Renamed from...
16960 (final_scan_insn): ... this. New wrapper, to recover
16961 seen from the outermost call in recursive ones.
16962 * config/sparc/sparc.c (output_return): Drop seen from call.
16963 (output_sibcall): Likewise.
16964 * config/visium/visium.c (output_branch): Likewise.
16966 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
16968 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
16971 2018-02-10 Alan Modra <amodra@gmail.com>
16974 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
16975 Specify LR as an input.
16977 2018-02-10 Jakub Jelinek <jakub@redhat.com>
16980 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
16981 remove_member_access_dummy_vars): New functions.
16982 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
16983 lower_omp_1, execute_lower_omp): Use them.
16985 PR rtl-optimization/84308
16986 * shrink-wrap.c (spread_components): Release todo vector.
16988 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
16990 PR rtl-optimization/57193
16991 * ira-color.c (struct allocno_color_data): Add member
16992 conflict_allocno_hard_prefs.
16993 (update_conflict_allocno_hard_prefs): New.
16994 (bucket_allocno_compare_func): Add a preference based on
16995 conflict_allocno_hard_prefs.
16996 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
16997 (color_allocnos): Remove a dead code. Initiate
16998 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
17000 2018-02-09 Jakub Jelinek <jakub@redhat.com>
17003 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
17004 constraint from =wa to wa. Avoid a subreg on the output operand,
17005 instead use a pseudo and subreg it in a move.
17006 (p9_xxbrd_<mode>): Changed to ...
17007 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
17008 (p9_xxbrd_v2df): New expander.
17009 (p9_xxbrw_<mode>): Changed to ...
17010 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
17011 (p9_xxbrw_v4sf): New expander.
17013 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
17015 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
17017 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
17020 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
17021 multiply in 32-bit mode.
17022 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
17023 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
17026 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
17028 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
17029 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
17030 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
17031 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
17033 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
17036 * dwarf2out.c (is_trivial_indirect_ref): New function.
17037 (dwarf2out_late_global_decl): Do not generate a location
17038 attribute for variables that have a non-trivial DECL_VALUE_EXPR
17039 and that are not defined in the current unit.
17041 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
17043 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
17044 instead of a libcall for UNORDERED.
17046 2018-02-09 Tamar Christina <tamar.christina@arm.com>
17049 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
17050 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
17052 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17055 * config/s390/s390.c (s390_set_current_function): Invoke
17056 s390_indirect_branch_settings also if fndecl didn't change.
17058 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
17060 * config/rs6000/rs6000.md (blockage): Set length to zero.
17062 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
17064 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
17066 2018-02-09 Jakub Jelinek <jakub@redhat.com>
17069 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
17070 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
17074 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
17075 PARALLEL incoming that failed vt_get_decl_and_offset check.
17077 PR middle-end/84237
17078 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
17079 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
17081 (get_variable_section): For decls in named .bss* sections pass true as
17082 second argument to bss_initializer_p.
17084 2018-02-09 Marek Polacek <polacek@redhat.com>
17085 Jakub Jelinek <jakub@redhat.com>
17088 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
17089 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
17090 Sync some changes from cxx_fold_indirect_ref.
17092 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
17094 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
17096 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
17097 (BLOCK_INLINE_ENTRY_LABEL): New.
17098 (dwarf2out_var_location): Disregard inline entry markers.
17099 (inline_entry_data): New struct.
17100 (inline_entry_data_hasher): New hashtable type.
17101 (inline_entry_data_hasher::hash): New.
17102 (inline_entry_data_hasher::equal): New.
17103 (inline_entry_data_table): New variable.
17104 (add_high_low_attributes): Add DW_AT_entry_pc and
17105 DW_AT_GNU_entry_view attributes if a pending entry is found
17106 in inline_entry_data_table. Add old entry_pc attribute only
17107 if debug nonbinding markers are disabled.
17108 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
17109 markers are enabled.
17110 (block_within_block_p, dwarf2out_inline_entry): New.
17111 (dwarf2out_finish): Check that no entries remained in
17112 inline_entry_data_table.
17113 * final.c (reemit_insn_block_notes): Handle inline entry notes.
17114 (final_scan_insn, notice_source_line): Likewise.
17115 (rest_of_clean_state): Skip inline entry markers.
17116 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
17118 * gimple.c (gimple_build_debug_inline_entry): New.
17119 * gimple.h (enum gimple_debug_subcode): Add
17120 GIMPLE_DEBUG_INLINE_ENTRY.
17121 (gimple_build_debug_inline_entry): Declare.
17122 (gimple_debug_inline_entry_p): New.
17123 (gimple_debug_nonbind_marker_p): Adjust.
17124 * insn-notes.def (INLINE_ENTRY): New.
17125 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
17126 inline entry marker notes.
17127 (print_insn): Likewise.
17128 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
17129 (INSN_DEBUG_MARKER_KIND): Likewise.
17130 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
17131 * tree-inline.c (expand_call_inline): Build and insert
17132 debug_inline_entry stmt.
17133 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
17134 inline entry blocks early, if nonbind markers are enabled.
17135 (dump_scope_block): Dump fragment info.
17136 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
17137 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
17138 (gimple_build_debug_inline_entry): New.
17139 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
17140 Enable/disable inline entry points too.
17141 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
17142 (DEBUG_INSN): Describe inline entry markers.
17144 * common.opt (gvariable-location-views): New.
17145 (gvariable-location-views=incompat5): New.
17146 * config.in: Rebuilt.
17147 * configure: Rebuilt.
17148 * configure.ac: Test assembler for view support.
17149 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
17150 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
17151 * dwarf2out.c (var_loc_view): New typedef.
17152 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
17153 (dwarf2out_locviews_in_attribute): New.
17154 (dwarf2out_locviews_in_loclist): New.
17155 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
17156 (enum dw_line_info_opcode): Add LI_adv_address.
17157 (struct dw_line_info_table): Add view.
17158 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
17159 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
17160 (zero_view_p): New variable.
17161 (ZERO_VIEW_P): New macro.
17162 (output_asm_line_debug_info): New.
17163 (struct var_loc_node): Add view.
17164 (add_AT_view_list, AT_loc_list): New.
17165 (add_var_loc_to_decl): Add view param. Test it against last.
17166 (new_loc_list): Add view params. Record them.
17167 (AT_loc_list_ptr): Handle loc and view lists.
17168 (view_list_to_loc_list_val_node): New.
17169 (print_dw_val): Handle dw_val_class_view_list.
17170 (size_of_die): Likewise.
17171 (value_format): Likewise.
17172 (loc_list_has_views): New.
17173 (gen_llsym): Set vl_symbol too.
17174 (maybe_gen_llsym, skip_loc_list_entry): New.
17175 (dwarf2out_maybe_output_loclist_view_pair): New.
17176 (output_loc_list): Output view list or entries too.
17177 (output_view_list_offset): New.
17178 (output_die): Handle dw_val_class_view_list.
17179 (output_dwarf_version): New.
17180 (output_compilation_unit_header): Use it.
17181 (output_skeleton_debug_sections): Likewise.
17182 (output_rnglists, output_line_info): Likewise.
17183 (output_pubnames, output_aranges): Update version comments.
17184 (output_one_line_info_table): Output view numbers in asm comments.
17185 (dw_loc_list): Determine current endview, pass it to new_loc_list.
17186 Call maybe_gen_llsym.
17187 (loc_list_from_tree_1): Adjust.
17188 (add_AT_location_description): Create view list attribute if
17189 needed, check it's absent otherwise.
17190 (convert_cfa_to_fb_loc_list): Adjust.
17191 (maybe_emit_file): Call output_asm_line_debug_info for test.
17192 (dwarf2out_var_location): Reset views as needed. Precompute
17193 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
17194 attribute. Set view.
17195 (new_line_info_table): Reset next view.
17196 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
17197 (dwarf2out_source_line): Likewise. Output view resets and labels to
17198 the assembler, or select appropriate line info opcodes.
17199 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
17200 (optimize_string_length): Catch it. Adjust.
17201 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
17202 dw_val_class_view_list, and remove it if no longer needed.
17203 (hash_loc_list): Hash view numbers.
17204 (loc_list_hasher::equal): Compare them.
17205 (optimize_location_lists): Check whether a view list symbol is
17206 needed, and whether the locview attribute is present, and
17207 whether they match. Remove the locview attribute if no longer
17209 (index_location_lists): Call skip_loc_list_entry for test.
17210 (dwarf2out_finish): Call output_asm_line_debug_info for test.
17211 Use output_dwarf_version.
17212 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
17213 (struct dw_val_node): Add val_view_list.
17214 * final.c (SEEN_NEXT_VIEW): New.
17215 (set_next_view_needed): New.
17216 (clear_next_view_needed): New.
17217 (maybe_output_next_view): New.
17218 (final_start_function): Rename to...
17219 (final_start_function_1): ... this. Take pointer to FIRST,
17220 add SEEN parameter. Emit param bindings in the initial view.
17221 (final_start_function): Reintroduce SEEN-less interface.
17222 (final): Rename to...
17223 (final_1): ... this. Take SEEN parameter. Output final pending
17224 next view at the end.
17225 (final): Reintroduce seen-less interface.
17226 (final_scan_insn): Output pending next view before switching
17227 sections or ending a block. Mark the next view as needed when
17228 outputting variable locations. Notify debug backend of section
17229 changes, and of location view changes.
17230 (rest_of_handle_final): Adjust.
17231 * toplev.c (process_options): Autodetect value for debug variable
17232 location views option. Warn on incompat5 without -gdwarf-5.
17233 * doc/invoke.texi (gvariable-location-views): New.
17234 (gvariable-location-views=incompat5): New.
17235 (gno-variable-location-views): New.
17237 2018-02-08 David Malcolm <dmalcolm@redhat.com>
17239 PR tree-optimization/84136
17240 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
17241 that the result of find_edge is non-NULL.
17243 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
17246 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
17247 storing integer register in SImode. Fix cost of 256 and 512
17248 byte aligned SSE register store.
17250 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
17252 * config/i386/i386.c (ix86_multiplication_cost): Fix
17253 multiplication cost for TARGET_AVX512DQ.
17255 2018-02-08 Marek Polacek <polacek@redhat.com>
17257 PR tree-optimization/84238
17258 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
17261 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17263 PR tree-optimization/84265
17264 * tree-vect-stmts.c (vectorizable_store): Don't treat
17265 VMAT_CONTIGUOUS accesses as grouped.
17266 (vectorizable_load): Likewise.
17268 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17270 PR tree-optimization/81635
17271 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
17272 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
17273 (test_round_for_mask): New functions.
17274 (wide_int_cc_tests): Call test_round_for_mask.
17275 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
17276 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
17277 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
17278 range returned by get_range_info.
17280 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
17283 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
17284 * symtab.c: Include builtins.h
17285 (symtab_node::output_to_lto_symbol_table_p): Move here
17286 from lto-streamer-out.c:output_symbol_p.
17287 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
17288 (output_symbol_p): Move all logic to symtab.c
17289 (produce_symtab): Update.
17291 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17293 * config/s390/s390-opts.h (enum indirect_branch): Define.
17294 * config/s390/s390-protos.h (s390_return_addr_from_memory)
17295 (s390_indirect_branch_via_thunk)
17296 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
17297 (enum s390_indirect_branch_type): Define.
17298 * config/s390/s390.c (struct s390_frame_layout, struct
17299 machine_function): Remove.
17300 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
17301 (indirect_branch_table_label_no, indirect_branch_table_name):
17303 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
17304 (enum s390_indirect_branch_option): Define.
17305 (s390_return_addr_from_memory): New function.
17306 (s390_handle_string_attribute): New function.
17307 (s390_attribute_table): Add new attribute handler.
17308 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
17309 (s390_indirect_branch_via_thunk): New function.
17310 (s390_indirect_branch_via_inline_thunk): New function.
17311 (s390_function_ok_for_sibcall): When jumping via thunk disallow
17312 sibling call optimization for non z10 compiles.
17313 (s390_emit_call): Force indirect branch target to be a single
17314 register. Add r1 clobber for non-z10 compiles.
17315 (s390_emit_epilogue): Emit return jump via return_use expander.
17316 (s390_reorg): Handle JUMP_INSNs as execute targets.
17317 (s390_option_override_internal): Perform validity checks for the
17318 new command line options.
17319 (s390_indirect_branch_attrvalue): New function.
17320 (s390_indirect_branch_settings): New function.
17321 (s390_set_current_function): Invoke s390_indirect_branch_settings.
17322 (s390_output_indirect_thunk_function): New function.
17323 (s390_code_end): Implement target hook.
17324 (s390_case_values_threshold): Implement target hook.
17325 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
17327 * config/s390/s390.h (struct s390_frame_layout)
17328 (struct machine_function): Move here from s390.c.
17329 (TARGET_INDIRECT_BRANCH_NOBP_RET)
17330 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
17331 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
17332 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
17333 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
17334 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
17335 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
17336 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
17337 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
17338 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
17339 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
17340 (mnemonic attribute): Add values which aren't recognized
17342 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
17343 pattern for branch conversion. Fix mnemonic attribute.
17344 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
17345 indirect branch via thunk if requested.
17346 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
17347 ("*indirect_jump"): Disable for branch conversion using out of
17349 ("indirect_jump_via_thunk<mode>_z10")
17350 ("indirect_jump_via_thunk<mode>")
17351 ("indirect_jump_via_inlinethunk<mode>_z10")
17352 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
17353 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
17354 ("casesi_jump_via_inlinethunk<mode>_z10")
17355 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
17356 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
17357 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
17358 ("*indirect2_jump"): Disable for branch conversion.
17359 ("casesi_jump"): Turn into expander and expand patterns for branch
17361 ("return_use"): New expander.
17362 ("*return"): Emit return via thunk and rename it to ...
17363 ("*return<mode>"): ... this one.
17364 * config/s390/s390.opt: Add new options and and enum for the
17367 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
17369 * lra-constraints.c (match_reload): Unconditionally use
17370 gen_lowpart_SUBREG, rather than selecting between that
17371 and equivalent gen_rtx_SUBREG code.
17373 2018-02-08 Richard Biener <rguenther@suse.de>
17375 PR tree-optimization/84233
17376 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
17377 changed flag instead of boguously re-using phi_inserted.
17379 2018-02-08 Martin Jambor <mjambor@suse.cz>
17381 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
17382 static local variables.
17384 2018-02-08 Richard Biener <rguenther@suse.de>
17386 PR tree-optimization/84278
17387 * tree-vect-stmts.c (vectorizable_store): When looking for
17388 smaller vector types to perform grouped strided loads/stores
17389 make sure the mode is supported by the target.
17390 (vectorizable_load): Likewise.
17392 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
17394 * config/aarch64/aarch64.c (aarch64_components_for_bb):
17395 Increase LDP/STP opportunities by adding adjacent callee-saves.
17397 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
17399 PR rtl-optimization/84068
17400 PR rtl-optimization/83459
17401 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
17403 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
17405 PR tree-optimization/84224
17406 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
17407 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
17408 non-zero arguments.
17410 2018-02-07 Iain Sandoe <iain@codesourcery.com>
17413 * config/rs6000/altivec.md (*restore_world): Remove LR use.
17414 * config/rs6000/predicates.md (restore_world_operation): Adjust op
17415 count, remove one USE.
17417 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
17419 * doc/install.texi (Configuration): Document the
17420 --with-long-double-format={ibm,ieee} PowerPC configuration
17424 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
17425 Convert from define_expand to be define_insn_and_split. Rework
17426 float/double/_Float128 conversions to QI/HI/SImode to work with
17427 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
17428 conversions to QI/HImode types did a store and then a load to
17429 truncate the value. For conversions to VSX registers, don't split
17430 the insn, instead emit the code directly. Use the code iterator
17431 any_fix to combine signed and unsigned conversions.
17432 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
17433 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
17434 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17435 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
17436 (fix_<mode>di2_hw): Likewise.
17437 (fixuns_<mode>di2_hw): Likewise.
17438 (fix_<mode>si2_hw): Likewise.
17439 (fixuns_<mode>si2_hw): Likewise.
17440 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
17441 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
17442 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
17443 fix<uns>_trunc<SFDF:mode>si2_p8.
17444 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
17445 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
17446 (fix<uns>_<mode>_mem): Likewise.
17447 (fctiw<u>z_<mode>_mem): Likewise.
17448 (fix<uns>_<mode>_mem): Likewise.
17449 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
17450 the register allocator from doing a direct move to the GPRs to do
17451 a store, and instead use the ISA 3.0 store byte/half-word from
17452 vector register instruction. For IEEE 128-bit floating point,
17453 also optimize stores of 32-bit ints.
17454 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
17456 2018-02-07 Alan Hayward <alan.hayward@arm.com>
17458 * genextract.c (push_pathstr_operand): New function to support
17460 (walk_rtx): Call push_pathstr_operand.
17461 (print_path): Support [a-zA-Z].
17463 2018-02-07 Richard Biener <rguenther@suse.de>
17465 PR tree-optimization/84037
17466 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
17467 (cse_and_gimplify_to_preheader): Declare.
17468 (vect_get_place_in_interleaving_chain): Likewise.
17469 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17471 (_loop_vec_info::~_loop_vec_info): Delete it.
17472 (cse_and_gimplify_to_preheader): New function.
17473 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
17474 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
17475 (vectorizable_load): Likewise. For grouped stores always base
17476 the IV on the first element.
17477 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
17478 condition before gimplifying.
17480 2018-02-07 Jakub Jelinek <jakub@redhat.com>
17482 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
17483 *DIV_EXPR and *MOD_EXPR.
17485 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
17488 * config/i386/i386.c (ix86_option_override_internal): Mask out
17489 the CF_SET bit when checking -fcf-protection.
17491 2018-02-07 Tom de Vries <tom@codesourcery.com>
17494 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
17497 2018-02-07 Richard Biener <rguenther@suse.de>
17499 PR tree-optimization/84204
17500 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
17503 PR tree-optimization/84205
17504 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
17505 special-case isl_ast_op_zdiv_r.
17507 PR tree-optimization/84223
17508 * graphite-scop-detection.c (gather_bbs::before_dom_children):
17509 Only add conditions from within the region.
17510 (gather_bbs::after_dom_children): Adjust.
17512 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
17515 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
17516 * config/avr/avr.md: Only post-reload split REG-REG moves if
17517 either register is GENERAL_REG_P.
17519 2018-02-07 Jakub Jelinek <jakub@redhat.com>
17521 PR tree-optimization/84235
17522 * tree-ssa-scopedtables.c
17523 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
17524 if the subtraction is performed in floating point type where NaNs are
17525 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
17526 build 1. Formatting fix.
17528 2018-02-06 Jakub Jelinek <jakub@redhat.com>
17531 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
17532 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
17533 and skip it regardless of bb boundaries. Use CALL_P macro,
17534 don't test INSN_P (insn) together with CALL_P or JUMP_P check
17535 unnecessarily, formatting fix.
17537 2018-02-06 Michael Collison <michael.collison@arm.com>
17539 * config/arm/thumb2.md:
17540 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
17541 (*thumb_mov_notscc): Ditto.
17543 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
17546 * config/rs6000/rs6000.md (su code attribute): Use "u" for
17547 unsigned_fix, not "s".
17549 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17551 * configure.ac (gcc_fn_eh_frame_ro): New function.
17552 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
17553 correct .eh_frame permissions.
17554 * configure: Regenerate.
17556 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
17558 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
17559 irrelevant options.
17561 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17563 * config/rs6000/rs6000.c (rs6000_option_override_internal):
17564 Display warning message for -mno-speculate-indirect-jumps.
17566 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
17568 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
17570 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
17572 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
17574 PR tree-optimization/84225
17575 * tree-eh.c (find_trapping_overflow): Only call
17576 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
17578 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17581 * config/i386/i386.c: Reimplement the check of possible options
17582 -mibt/-mshstk conbination. Change error messages.
17583 * doc/invoke.texi: Fix a typo: remove extra '='.
17585 2018-02-06 Marek Polacek <polacek@redhat.com>
17587 PR tree-optimization/84228
17588 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
17590 2018-02-06 Tamar Christina <tamar.christina@arm.com>
17593 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
17594 emitted arch directives.
17595 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
17596 __ARM_FEATURE_COPROC before changing architectures.
17598 2018-02-06 Richard Biener <rguenther@suse.de>
17600 * config/i386/i386.c (print_reg): Fix typo.
17601 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
17603 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
17605 * configure: Regenerate.
17607 2018-02-05 Martin Sebor <msebor@redhat.com>
17609 PR tree-optimization/83369
17610 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
17613 2018-02-05 Martin Liska <mliska@suse.cz>
17615 * doc/invoke.texi: Cherry-pick upstream r323995.
17617 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
17619 * ira.c (ira_init_register_move_cost): Adjust comment.
17621 2018-02-05 Martin Liska <mliska@suse.cz>
17623 PR gcov-profile/84137
17624 * doc/gcov.texi: Fix typo in documentation.
17626 2018-02-05 Martin Liska <mliska@suse.cz>
17628 PR gcov-profile/83879
17629 * doc/gcov.texi: Document necessity of --dynamic-list-data when
17630 using dlopen functionality.
17632 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
17634 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
17635 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
17636 _mm_maskz_range_ss, _mm_mask_range_round_ss,
17637 _mm_maskz_range_round_ss): New intrinsics.
17638 (__builtin_ia32_rangesd128_round)
17639 (__builtin_ia32_rangess128_round): Remove.
17640 (__builtin_ia32_rangesd128_mask_round,
17641 __builtin_ia32_rangess128_mask_round): New builtins.
17642 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
17643 __builtin_ia32_rangess128_round): Remove.
17644 (__builtin_ia32_rangesd128_mask_round,
17645 __builtin_ia32_rangess128_mask_round): New builtins.
17646 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
17647 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
17648 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
17649 "<round_saeonly_constraint>")): Changed to ...
17650 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
17651 "<round_saeonly_scalar_constraint>")): ... this.
17652 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
17653 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
17654 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
17655 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
17656 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
17658 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
17660 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
17662 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
17663 Remove all values except native, 8540 and 8548.
17665 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
17667 * config/i386/i386.c (ix86_output_function_return): Pass
17668 INVALID_REGNUM, instead of -1, as invalid register number to
17669 indirect_thunk_name and output_indirect_thunk.
17671 2018-02-02 Julia Koval <julia.koval@intel.com>
17673 * config.gcc: Add -march=icelake.
17674 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
17675 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
17676 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
17677 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
17678 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
17679 (processor_target_table): Add icelake.
17680 (ix86_option_override_internal): Handle new PTAs.
17681 (get_builtin_code_for_version): Handle icelake.
17682 (M_INTEL_COREI7_ICELAKE): New.
17683 (fold_builtin_cpu): Handle icelake.
17684 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
17685 * doc/invoke.texi: Add -march=icelake.
17687 2018-02-02 Julia Koval <julia.koval@intel.com>
17689 * config/i386/i386.c (ix86_option_override_internal): Change flags type
17690 to wide_int_bitmask.
17691 * wide-int-bitmask.h: New.
17693 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
17696 * config/i386/i386.md: Replace Pmode with word_mode in
17697 builtin_setjmp_setup and builtin_longjmp to support x32.
17699 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
17703 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
17705 (rs6000_supported_cpu_names): New static variable.
17706 (linux_cpu_translation_table): Likewise.
17707 (elf_platform) <cpu>: Define new static variable and use it.
17708 Translate kernel AT_PLATFORM name to canonical name if needed.
17709 Error if platform name is unknown.
17711 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
17714 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
17716 2018-02-01 Jeff Law <law@redhat.com>
17719 * config/i386/i386.c (release_scratch_register_on_entry): Add new
17720 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
17721 the scratch if RELEASE_VIA_POP is false.
17722 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
17723 If we have to save a temporary register, decrement SIZE appropriately.
17724 Pass new arguments to release_scratch_register_on_entry.
17725 (ix86_adjust_stack_and_probe): Likewise.
17726 (ix86_emit_probe_stack_range): Pass new arguments to
17727 release_scratch_register_on_entry.
17729 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
17731 PR rtl-optimization/84157
17732 * combine.c (change_zero_ext): Use REG_P predicate in
17733 front of HARD_REGISTER_P predicate.
17735 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
17737 * config/avr/avr.c (avr_option_override): Move disabling of
17738 -fdelete-null-pointer-checks to...
17739 * common/config/avr/avr-common.c (avr_option_optimization_table):
17742 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17744 PR tree-optimization/81635
17745 * tree-data-ref.c (split_constant_offset_1): For types that
17746 wrap on overflow, try to use range info to prove that wrapping
17749 2018-02-01 Renlin Li <renlin.li@arm.com>
17752 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
17753 TAILCALL_ADDR_REGS.
17754 (aarch64_register_move_cost): Likewise.
17755 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
17756 TAILCALL_ADDR_REGS.
17757 (REG_CLASS_NAMES): Likewise.
17758 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
17759 TAILCALL_ADDR_REGS. Remove IP registers.
17760 * config/aarch64/aarch64.md (Ucs): Update register constraint.
17762 2018-02-01 Richard Biener <rguenther@suse.de>
17764 * domwalk.h (dom_walker::dom_walker): Add additional constructor
17765 for specifying RPO order and allow NULL for that.
17766 * domwalk.c (dom_walker::dom_walker): Likewise.
17767 (dom_walker::walk): Handle NULL RPO order.
17768 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
17770 (rewrite_update_dom_walker): Likewise.
17771 (mark_def_dom_walker): Likewise.
17773 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17775 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
17776 (aarch64_maybe_expand_sve_subreg_move): Declare.
17777 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
17778 * config/aarch64/predicates.md (aarch64_any_register_operand): New
17780 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
17781 that are semantically a reverse operation.
17782 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
17783 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
17784 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
17786 (aarch64_can_change_mode_class): For big-endian, forbid changes
17787 between two SVE modes if they have different element sizes.
17789 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17791 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
17792 the TImode handling for big-endian targets.
17794 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17796 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
17797 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
17799 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
17800 Remove BSWAP handing for big-endian targets and use the form of
17801 LD1RQ appropariate for the mode.
17803 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17805 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
17806 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
17807 duplicated element.
17809 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
17812 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
17813 check for operands that need to go through aarch64_sve_reload_be.
17815 2018-02-01 Jakub Jelinek <jakub@redhat.com>
17817 PR tree-optimization/81661
17818 PR tree-optimization/84117
17819 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
17820 * tree-eh.c: Include gimplify.h.
17821 (find_trapping_overflow, replace_trapping_overflow,
17822 rewrite_to_non_trapping_overflow): New functions.
17823 * tree-vect-loop.c: Include tree-eh.h.
17824 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
17825 * tree-data-ref.c: Include tree-eh.h.
17826 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
17828 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
17830 PR rtl-optimization/84123
17831 * combine.c (change_zero_ext): Check if hard register satisfies
17832 can_change_dest_mode before calling gen_lowpart_SUBREG.
17834 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
17837 * ira.c (ira_init_register_move_cost): Remove assert.
17839 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17841 PR rtl-optimization/84071
17842 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
17843 * doc/tm.texi: Regenerate.
17845 2018-01-31 Richard Biener <rguenther@suse.de>
17847 PR tree-optimization/84132
17848 * tree-data-ref.c (analyze_miv_subscript): Properly
17849 check whether evolution_function_is_affine_multivariate_p
17850 before calling gcd_of_steps_may_divide_p.
17852 2018-01-31 Julia Koval <julia.koval@intel.com>
17855 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
17856 * config/i386/i386.md (rdpid_rex64) New.
17857 (rdpid): Make 32bit only.
17859 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
17862 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
17863 an IDENTIFIER_NODE for FUNCTION_TYPE's.
17865 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17868 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17870 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17872 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
17874 PR rtl-optimization/84071
17875 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
17876 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
17878 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17880 * config/arc/arc.c (arc_handle_aux_attribute): New function.
17881 (arc_attribute_table): Add 'aux' attribute.
17882 (arc_in_small_data_p): Consider aux like variables.
17883 (arc_is_aux_reg_p): New function.
17884 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
17885 (arc_get_aux_arg): New function.
17886 (prepare_move_operands): Handle aux-register access.
17887 (arc_handle_aux_attribute): New function.
17888 * doc/extend.texi (ARC Variable attributes): Add subsection.
17890 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
17892 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
17893 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
17894 (arc_attribute_table): Add 'uncached' attribute.
17895 (arc_print_operand): Print '.di' flag for uncached memory
17897 (arc_in_small_data_p): Do not consider for small data the uncached
17899 (arc_is_uncached_mem_p): New function.
17900 * config/arc/predicates.md (compact_store_memory_operand): Check
17901 for uncached memory accesses.
17902 (nonvol_nonimm_operand): Likewise.
17903 * doc/extend.texi (ARC Type Attribute): New subsection.
17905 2018-01-31 Jakub Jelinek <jakub@redhat.com>
17908 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
17909 falign-loops=): Add Optimization flag.
17911 2018-01-30 Jeff Law <law@redhat.com>
17914 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
17915 INT_REGISTERS_SAVED. Check it prior to calling
17916 get_scratch_register_on_entry.
17917 (ix86_adjust_stack_and_probe): Similarly.
17918 (ix86_emit_probe_stack_range): Similarly.
17919 (ix86_expand_prologue): Corresponding changes.
17921 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17924 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
17925 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
17927 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
17930 * lra-constraints.c (curr_insn_transform): Process AND in the
17933 2018-01-30 Jakub Jelinek <jakub@redhat.com>
17935 PR rtl-optimization/83986
17936 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
17937 dependence against last_pending_memory_flush in addition to
17938 pending_jump_insns.
17940 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
17942 PR tree-optimization/81611
17943 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
17946 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17949 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
17952 2018-01-30 Richard Biener <rguenther@suse.de>
17953 Jakub Jelinek <jakub@redhat.com>
17955 PR tree-optimization/84111
17956 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
17957 inner loops added during recursion, as they don't have up-to-date
17960 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17963 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
17964 (can_inline_edge_by_limits_p): ... here.
17965 (can_early_inline_edge_p, check_callers,
17966 update_caller_keys, update_callee_keys, recursive_inlining,
17967 add_new_edges_to_heap, speculation_useful_p,
17968 inline_small_functions,
17969 inline_small_functions, flatten_function,
17970 inline_to_all_callers_1): Update.
17972 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
17974 * profile-count.c (profile_count::combine_with_ipa_count): Handle
17977 2018-01-30 Richard Biener <rguenther@suse.de>
17979 PR tree-optimization/83008
17980 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
17981 invariant and constant vector uses in stmts when they need
17982 more than one stmt.
17984 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17987 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
17988 * configure: Regenerate.
17990 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17992 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
17994 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
17995 Use gen_rtx_REG rather than gen_lowpart.
17997 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
17999 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
18000 rather than 0 when creating partial subregs.
18002 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
18004 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
18007 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
18010 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
18011 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
18012 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
18013 flags. This restores the settings used before the 2017-07-24.
18014 Turning off pre increment/decrement/modify allows IVOPTS to
18015 optimize DF/SF loops where the index is an int.
18017 2018-01-29 Richard Biener <rguenther@suse.de>
18018 Kelvin Nilsen <kelvin@gcc.gnu.org>
18021 * tree-vect-stmts.c (vectorizable_call): Don't call
18022 targetm.vectorize_builtin_md_vectorized_function if callee is
18025 2018-01-22 Carl Love <cel@us.ibm.com>
18027 * doc/extend.tex: Fix typo in second arg in
18028 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
18030 2018-01-29 Richard Biener <rguenther@suse.de>
18032 PR tree-optimization/84086
18033 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
18034 (flush_ssaname_freelist): When SSA names were released reset
18035 the SCEV hash table.
18037 2018-01-29 Richard Biener <rguenther@suse.de>
18039 PR tree-optimization/84057
18040 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
18041 removed paths when removing edges.
18043 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
18045 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
18046 -mfunction-return=@var{choice}.
18048 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
18050 PR diagnostic/84034
18051 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
18052 Handle CR like TAB.
18053 (layout::print_source_line): Likewise.
18054 (test_get_line_width_without_trailing_whitespace): Add test cases.
18056 2018-01-27 Jakub Jelinek <jakub@redhat.com>
18058 PR middle-end/84040
18059 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
18062 2018-01-26 Jim Wilson <jimw@sifive.com>
18064 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
18066 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
18069 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18071 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
18072 and CMP + SUB-immediate -> SUBS.
18074 2018-01-26 Martin Sebor <msebor@redhat.com>
18076 PR tree-optimization/83896
18077 * tree-ssa-strlen.c (get_string_len): Rename...
18078 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
18079 Avoid assuming length is constant.
18080 (handle_char_store): Use HOST_WIDE_INT for string length.
18082 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
18085 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
18086 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
18088 2018-01-26 Richard Biener <rguenther@suse.de>
18090 PR rtl-optimization/84003
18091 * dse.c (record_store): Only record redundant stores when
18092 the earlier store aliases at least all accesses the later one does.
18094 2018-01-26 Jakub Jelinek <jakub@redhat.com>
18096 PR rtl-optimization/83985
18097 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
18098 REG_CFA_RESTORE insns.
18099 (delete_unmarked_insns): Don't ignore separate shrink wrapping
18100 REG_CFA_RESTORE insns here.
18103 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
18104 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
18106 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18108 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
18109 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
18110 (arc_init): Likewise.
18111 (arc_override_options): Likewise.
18112 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
18114 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
18116 * config/arc/arc.h (TARGET_DBNZ): Define.
18117 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
18118 properly set the tune attribute.
18119 (dbnz): Use TARGET_DBNZ guard.
18120 * config/arc/arc.opt (mtune): Add core3 option.
18122 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18124 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
18125 recognize new pic like addresses.
18126 (arc_delegitimize_address): Clean up.
18128 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18130 * config/arc/arc-arches.def: Option mrf16 valid for all
18132 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
18133 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
18134 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
18135 * config/arc/arc-tables.opt: Regenerate.
18136 * config/arc/arc.c (arc_conditional_register_usage): Handle
18137 reduced register file case.
18138 (arc_file_start): Set must have build attributes.
18139 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
18140 mrf16 option value.
18141 * config/arc/arc.opt (mrf16): Add new option.
18142 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
18143 * config/arc/genmultilib.awk: Handle new mrf16 option.
18144 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
18145 * config/arc/t-multilib: Regenerate.
18146 * doc/invoke.texi (ARC Options): Document mrf16 option.
18148 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18150 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
18151 * config/arc/arc.c (arc_handle_secure_attribute): New function.
18152 (arc_attribute_table): Add 'secure_call' attribute.
18153 (arc_print_operand): Print secure call operand.
18154 (arc_function_ok_for_sibcall): Don't optimize tail calls when
18156 (arc_is_secure_call_p): New function. * config/arc/arc.md
18157 (call_i): Add support for sjli instruction.
18158 (call_value_i): Likewise.
18159 * config/arc/constraints.md (Csc): New constraint.
18161 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
18162 John Eric Martin <John.Martin@emmicro-us.com>
18164 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
18165 * config/arc/arc.c (_arc_jli_section): New struct.
18166 (arc_jli_section): New type.
18167 (rc_jli_sections): New static variable.
18168 (arc_handle_jli_attribute): New function.
18169 (arc_attribute_table): Add jli_always and jli_fixed attribute.
18170 (arc_file_end): New function.
18171 (TARGET_ASM_FILE_END): Define.
18172 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
18173 (arc_add_jli_section): New function.
18174 (jli_call_scan): Likewise.
18175 (arc_reorg): Call jli_call_scan.
18176 (arc_output_addsi): Remove 'S' from printing asm operand.
18177 (arc_is_jli_call_p): New function.
18178 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
18180 (movhi_insn): Likewise.
18181 (movsi_insn): Likewise.
18182 (movsi_set_cc_insn): Likewise.
18183 (loadqi_update): Likewise.
18184 (load_zeroextendqisi_update): Likewise.
18185 (load_signextendqisi_update): Likewise.
18186 (loadhi_update): Likewise.
18187 (load_zeroextendhisi_update): Likewise.
18188 (load_signextendhisi_update): Likewise.
18189 (loadsi_update): Likewise.
18190 (loadsf_update): Likewise.
18191 (movsicc_insn): Likewise.
18192 (bset_insn): Likewise.
18193 (bxor_insn): Likewise.
18194 (bclr_insn): Likewise.
18195 (bmsk_insn): Likewise.
18196 (bicsi3_insn): Likewise.
18197 (cmpsi_cc_c_insn): Likewise.
18198 (movsi_ne): Likewise.
18199 (movsi_cond_exec): Likewise.
18200 (clrsbsi2): Likewise.
18201 (norm_f): Likewise.
18211 (call_i): Remove 'S' asm letter, add jli instruction.
18212 (call_value_i): Likewise.
18213 * config/arc/arc.op (mjli-always): New option.
18214 * config/arc/constraints.md (Cji): New constraint.
18215 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
18217 (subsf3_fpx): Likewise.
18218 (mulsf3_fpx): Likewise.
18219 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
18221 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
18222 function attrbutes.
18223 * doc/invoke.texi (ARC): Document mjli-always option.
18225 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
18227 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
18228 avoid addition with 0 and use incw and decw where possible.
18230 2018-01-26 Richard Biener <rguenther@suse.de>
18232 PR tree-optimization/81082
18233 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
18234 association if it requires casting to unsigned.
18235 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
18236 from fold_plusminus_mult_expr to catch important cases late when
18237 range info is available.
18239 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18241 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
18242 * configure.ac (hidden_linkonce): New test.
18243 * configure: Regenerate.
18244 * config.in: Regenerate.
18246 2018-01-26 Julia Koval <julia.koval@intel.com>
18248 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
18249 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
18250 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
18251 _mm_mask_bitshuffle_epi64_mask): Fix type.
18252 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
18253 USI_FTYPE_V4DI_V4DI_USI): Remove.
18254 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
18255 __builtin_ia32_vpshufbitqmb256_mask,
18256 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
18257 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
18258 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
18260 2018-01-26 Alan Modra <amodra@gmail.com>
18263 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
18264 UNSPEC_VBPERMQ. Sort other unspecs.
18266 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
18268 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
18270 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
18272 PR middle-end/83055
18273 * predict.c (drop_profile): Do not push/pop cfun; update also
18275 (handle_missing_profiles): Fix logic looking for zero profiles.
18277 2018-01-25 Jakub Jelinek <jakub@redhat.com>
18279 PR middle-end/83977
18280 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
18281 on functions with #pragma omp declare simd or functions with simd
18283 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
18284 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
18285 Remove trailing \n from warning_at calls.
18287 2018-01-25 Tom de Vries <tom@codesourcery.com>
18290 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
18291 for neutered workers.
18293 2018-01-24 Joseph Myers <joseph@codesourcery.com>
18296 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
18297 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
18299 2018-01-24 Jeff Law <law@redhat.com>
18302 * i386.c (get_probe_interval): Move to earlier point.
18303 (ix86_compute_frame_layout): If -fstack-clash-protection and
18304 the frame is larger than the probe interval, then use pushes
18305 to save registers rather than reg->mem moves.
18306 (ix86_expand_prologue): Remove conditional for int_registers_saved
18309 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
18312 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
18313 min/max for never referenced object.
18315 2018-01-24 Jakub Jelinek <jakub@redhat.com>
18317 PR middle-end/83977
18318 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
18320 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
18321 attributes from DECL_ATTRIBUTES (decl) without affecting
18322 DECL_ATTRIBUTES (current_function_decl).
18323 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
18324 functions with non-NULL DECL_ABSTRACT_ORIGIN.
18326 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
18328 PR tree-optimization/83979
18329 * fold-const.c (fold_comparison): Use constant_boolean_node
18330 instead of boolean_{true,false}_node.
18332 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
18334 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
18337 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18339 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18340 Simplify the clause that sets the length attribute.
18341 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18342 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
18343 clause that sets the length attribute.
18344 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18346 2018-01-24 Tom de Vries <tom@codesourcery.com>
18349 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
18350 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
18351 Add strict parameter.
18352 (prevent_branch_around_nothing): Insert dummy insn between branch to
18353 label and label with no ptx insn inbetween.
18354 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
18356 2018-01-24 Tom de Vries <tom@codesourcery.com>
18359 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
18360 for neutered threads in warp.
18361 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
18363 2018-01-24 Richard Biener <rguenther@suse.de>
18365 PR tree-optimization/83176
18366 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
18369 2018-01-24 Richard Biener <rguenther@suse.de>
18371 PR tree-optimization/82819
18372 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
18373 code generating pluses that are no-ops in the target precision.
18375 2018-01-24 Richard Biener <rguenther@suse.de>
18377 PR middle-end/84000
18378 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
18380 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18382 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
18383 to merge probabilities.
18384 * predict.c (probably_never_executed): Also mark as cold functions
18385 with global 0 profile and guessed local profile.
18386 * profile-count.c (profile_probability::combine_with_count): New
18388 * profile-count.h (profile_probability::operator*,
18389 profile_probability::operator*=, profile_probability::operator/,
18390 profile_probability::operator/=): Reduce precision to adjusted
18391 and set value to guessed on contradictory divisions.
18392 (profile_probability::combine_with_freq): Remove.
18393 (profile_probability::combine_wiht_count): Declare.
18394 (profile_count::force_nonzero):: Set to adjusted.
18395 (profile_count::probability_in):: Set quality to adjusted.
18396 * tree-ssa-tail-merge.c (replace_block_by): Use
18397 combine_with_count.
18399 2018-01-23 Andrew Waterman <andrew@sifive.com>
18400 Jim Wilson <jimw@sifive.com>
18402 * config/riscv/riscv.c (riscv_stack_boundary): New.
18403 (riscv_option_override): Set riscv_stack_boundary. Handle
18404 riscv_preferred_stack_boundary_arg.
18405 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
18406 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
18407 (STACK_BOUNDARY): Set to riscv_stack_boundary.
18408 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
18409 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
18410 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
18412 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
18415 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
18416 of struct ix86_frame.
18417 (ix86_expand_epilogue): Likewise. Add a local variable for
18418 the reg_save_offset field in struct ix86_frame.
18420 2018-01-23 Bin Cheng <bin.cheng@arm.com>
18422 PR tree-optimization/82604
18423 * tree-loop-distribution.c (enum partition_kind): New enum item
18424 PKIND_PARTIAL_MEMSET.
18425 (partition_builtin_p): Support above new enum item.
18426 (generate_code_for_partition): Ditto.
18427 (compute_access_range): Differentiate cases that equality can be
18428 proven at all loops, the innermost loops or no loops.
18429 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
18430 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
18431 (finalize_partitions, distribute_loop): Don't fuse partition of
18432 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
18433 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
18434 parloop is enabled.
18436 2018-01-23 Martin Liska <mliska@suse.cz>
18438 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
18439 order to ignore the predictor.
18440 (PRED_POLYMORPHIC_CALL): Likewise.
18441 (PRED_RECURSIVE_CALL): Likewise.
18443 2018-01-23 Martin Liska <mliska@suse.cz>
18445 * tree-profile.c (tree_profiling): Print function header to
18446 aware reader which function we are working on.
18447 * value-prof.c (gimple_find_values_to_profile): Do not print
18448 not interesting value histograms.
18450 2018-01-23 Martin Liska <mliska@suse.cz>
18452 * profile-count.h (enum profile_quality): Add
18453 profile_uninitialized as the first value. Do not number values
18454 as they are zero based.
18455 (profile_count::verify): Update sanity check.
18456 (profile_probability::verify): Likewise.
18458 2018-01-23 Nathan Sidwell <nathan@acm.org>
18460 * doc/invoke.texi (ffor-scope): Deprecate.
18462 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18464 PR tree-optimization/83510
18465 * domwalk.c (set_all_edges_as_executable): New function.
18466 (dom_walker::dom_walker): Convert bool param
18467 "skip_unreachable_blocks" to enum reachability. Move setup of
18468 edge flags to set_all_edges_as_executable and only do it when
18469 reachability is REACHABLE_BLOCKS.
18470 * domwalk.h (enum dom_walker::reachability): New enum.
18471 (dom_walker::dom_walker): Convert bool param
18472 "skip_unreachable_blocks" to enum reachability.
18473 (set_all_edges_as_executable): New decl.
18474 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
18475 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
18477 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
18478 but converting true to REACHABLE_BLOCKS.
18479 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
18481 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
18482 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
18483 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
18485 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
18486 if check_all_array_refs will be called.
18488 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18490 * tree.c (selftest::test_location_wrappers): Add more test
18493 2018-01-23 David Malcolm <dmalcolm@redhat.com>
18495 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
18496 (selftest::test_bit_in_range): Likewise.
18498 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
18501 * doc/sourcebuild.texi (vect_float): Say that the selector
18502 only describes the situation when -funsafe-math-optimizations is on.
18503 (vect_float_strict): Document.
18505 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
18507 PR tree-optimization/83965
18508 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
18509 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
18510 instead of checking only for a reduction.
18511 (vect_recog_widen_sum_pattern): Likewise.
18513 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18515 * predict.c (probably_never_executed): Only use precise profile info.
18516 (compute_function_frequency): Skip after inlining hack since we now
18517 have quality checking.
18519 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
18521 * profile-count.h (profile_probability::very_unlikely,
18522 profile_probability::unlikely, profile_probability::even): Set
18523 precision to guessed.
18525 2018-01-23 Richard Biener <rguenther@suse.de>
18527 PR tree-optimization/83963
18528 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
18529 Properly terminate dominator walk when crossing the exit edge not
18530 when visiting its source block.
18532 2018-01-23 Jakub Jelinek <jakub@redhat.com>
18535 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
18536 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
18538 2018-01-22 Jakub Jelinek <jakub@redhat.com>
18540 PR tree-optimization/83957
18541 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
18542 semicolon after for body surrounded by braces.
18544 PR tree-optimization/83081
18545 * profile-count.h (profile_probability::split): New method.
18546 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
18547 Use profile_probability::split.
18548 (do_compare_rtx_and_jump): Fix adjustment of probabilities
18549 when splitting a single conditional jump into 2.
18551 2018-01-22 David Malcolm <dmalcolm@redhat.com>
18553 PR tree-optimization/69452
18554 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
18557 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18559 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
18560 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
18561 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
18563 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18565 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
18567 * config/rl78/rl78.md (movdi): New define_expand.
18568 * config/rl78/rl78.c (rl78_split_movdi): New function.
18570 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
18573 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
18575 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
18576 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
18577 128-bit to produce an UNSPEC move to get the double word with the
18578 signbit and then a shift directly to do signbit.
18579 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
18580 implementation with a new version that just does either a direct
18581 move or a regular move. Move memory interface to separate insns.
18582 Move insns so they are next to the expander.
18583 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
18584 with signbit move. Split big and little endian case.
18585 (signbit<mode>2_dm_mem_le): Likewise.
18586 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
18587 (signbit<mode>2_dm2): Likewise.
18589 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18591 * config/rl78/rl78.md (anddi3): New define_expand.
18593 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18595 * config/rl78/rl78.md (umindi3): New define_expand.
18597 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18599 * config/rl78/rl78.md (smindi3): New define_expand.
18601 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18603 * config/rl78/rl78.md (smaxdi3): New define_expand.
18605 2018-01-22 Carl Love <cel@us.ibm.com>
18607 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
18608 LVX_V1TI): Add macro expansion.
18609 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
18610 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
18611 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
18612 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
18613 Change check to determine if the instruction is a byte reversing
18614 entry. Fix typo in comment.
18615 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
18616 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
18617 Add def_builtin calls for new builtins.
18618 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
18619 Add define_insn expansion.
18621 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18623 * config/rl78/rl78.md (umaxdi3): New define_expand.
18625 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
18627 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
18628 for non-QImode registers.
18630 2018-01-22 Richard Biener <rguenther@suse.de>
18632 PR tree-optimization/83963
18633 * graphite-scop-detection.c (scop_detection::get_sese): Delay
18634 including the loop exit block.
18635 (scop_detection::merge_sese): Likewise.
18636 (scop_detection::add_scop): Do it here instead.
18638 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18640 * doc/sourcebuild.texi (arm_softfloat): Document.
18642 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
18645 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
18646 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
18647 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
18649 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18650 David Edelsohn <dje.gcc@gmail.com>
18653 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18654 Change "crset eq" to "crset 2".
18655 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18656 (*call_indirect_aix<mode>_nospec): Likewise.
18657 (*call_value_indirect_aix<mode>_nospec): Likewise.
18658 (*call_indirect_elfv2<mode>_nospec): Likewise.
18659 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
18660 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
18661 change assembly output from . to $.
18662 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18663 (indirect_jump<mode>_nospec): Change assembly output from . to $.
18664 (*tablejump<mode>_internal1_nospec): Likewise.
18666 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
18669 * config/sh/sh_optimize_sett_clrt.cc:
18670 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
18672 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
18674 PR tree-optimization/83940
18675 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
18676 offset_dt to vect_constant_def rather than vect_unknown_def_type.
18677 (vect_check_load_store_mask): Add a mask_dt_out parameter and
18678 use it to pass back the definition type.
18679 (vect_check_store_rhs): Likewise rhs_dt_out.
18680 (vect_build_gather_load_calls): Add a mask_dt argument and use
18681 it instead of a call to vect_is_simple_use.
18682 (vectorizable_store): Update calls to vect_check_load_store_mask
18683 and vect_check_store_rhs. Use the dt returned by the latter instead
18684 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
18685 instead of calls to vect_is_simple_use. Pass the scalar rather
18686 than the vector operand to vect_is_simple_use when handling
18687 second and subsequent copies of an rhs value.
18688 (vectorizable_load): Update calls to vect_check_load_store_mask
18689 and vect_build_gather_load_calls. Use the cached mask_dt and
18690 gs_info.offset_dt instead of calls to vect_is_simple_use.
18692 2018-01-20 Jakub Jelinek <jakub@redhat.com>
18694 PR middle-end/83945
18695 * tree-emutls.c: Include gimplify.h.
18696 (lower_emutls_2): New function.
18697 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
18698 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
18699 it before further processing.
18702 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
18703 UINTVAL (trueop1) instead of INTVAL (op1).
18705 2018-01-19 Jakub Jelinek <jakub@redhat.com>
18709 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
18710 INCOMING_FRAME_SP_OFFSET if not defined.
18711 (scan_trace): Add ENTRY argument. If true and
18712 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
18713 emit a note to adjust the CFA offset.
18714 (create_cfi_notes): Adjust scan_trace callers.
18715 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
18716 INCOMING_FRAME_SP_OFFSET in the CIE.
18717 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
18718 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
18720 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
18721 * doc/tm.texi: Regenerated.
18723 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18725 PR rtl-optimization/83147
18726 * lra-constraints.c (remove_inheritance_pseudos): Use
18727 lra_substitute_pseudo_within_insn.
18729 2018-01-19 Tom de Vries <tom@codesourcery.com>
18730 Cesar Philippidis <cesar@codesourcery.com>
18733 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
18735 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
18738 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
18739 spaces for function labels.
18741 2018-01-19 Martin Liska <mliska@suse.cz>
18743 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
18744 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
18745 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
18746 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
18747 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
18748 (PRED_CONST_RETURN): Change from 69 to 65.
18749 (PRED_NULL_RETURN): Change from 91 to 71.
18750 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
18751 (PRED_LOOP_GUARD): Change from 66 to 73.
18753 2018-01-19 Martin Liska <mliska@suse.cz>
18755 * predict.c (predict_insn_def): Add new assert.
18756 (struct branch_predictor): Change type to signed integer.
18757 (test_prediction_value_range): Amend test to cover
18758 PROB_UNINITIALIZED.
18759 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
18760 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
18761 (PRED_LOOP_ITERATIONS_MAX): Likewise.
18762 (PRED_LOOP_IV_COMPARE): Likewise.
18763 * predict.h (PROB_UNINITIALIZED): Define new constant.
18765 2018-01-19 Martin Liska <mliska@suse.cz>
18767 * predict.c (dump_prediction): Add new format for
18768 analyze_brprob.py script which is enabled with -details
18770 * profile-count.h (precise_p): New function.
18772 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18774 PR tree-optimization/83922
18775 * tree-vect-loop.c (vect_verify_full_masking): Return false if
18776 there are no statements that need masking.
18777 (vect_active_double_reduction_p): New function.
18778 (vect_analyze_loop_operations): Use it when handling phis that
18779 are not in the loop header.
18781 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
18783 PR tree-optimization/83914
18784 * tree-vect-loop.c (vectorizable_induction): Don't convert
18785 init_expr or apply the peeling adjustment for inductions
18786 that are nested within the vectorized loop.
18788 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18790 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
18793 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18797 * function.h (gimplify_parameters): Add gimple_seq * argument.
18798 * function.c: Include gimple.h and options.h.
18799 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
18800 for the added local temporaries if needed.
18801 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
18802 if there are any parameter cleanups, wrap whole body into a
18803 try/finally with the cleanups.
18805 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
18808 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
18809 Use GET_MODE_CLASS for scalar floating point.
18811 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18815 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
18816 Fix call of call_cgraph_insertion_hooks.
18818 2018-01-18 Martin Sebor <msebor@redhat.com>
18820 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
18822 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
18825 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
18828 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
18831 * common.opt: (-ffile-prefix-map): New option.
18832 * opts.c (common_handle_option): Defer it.
18833 * opts-global.c (handle_common_deferred_options): Handle it.
18834 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
18835 * file-prefix-map.h: New file.
18836 (remap_debug_filename, add_debug_prefix_map): ...here.
18837 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
18838 * final.c (debug_prefix_map, add_debug_prefix_map
18839 remap_debug_filename): Move to...
18840 * file-prefix-map.c: New file.
18841 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
18842 generalize, get rid of alloca(), use strrchr() instead of strchr().
18843 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
18844 Implement in terms of add_prefix_map().
18845 (remap_macro_filename, remap_debug_filename): Implement in term of
18847 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
18848 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
18849 * dbxout.c: Include file-prefix-map.h.
18850 * varasm.c: Likewise.
18851 * vmsdbgout.c: Likewise.
18852 * xcoffout.c: Likewise.
18853 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
18854 * doc/cppopts.texi (-fmacro-prefix-map): Document.
18855 * doc/invoke.texi (-ffile-prefix-map): Document.
18856 (-fdebug-prefix-map): Update description.
18858 2018-01-18 Martin Liska <mliska@suse.cz>
18860 * config/i386/i386.c (indirect_thunk_name): Document that also
18862 (output_indirect_thunk): Document why both instructions
18863 (pause and lfence) are generated.
18865 2018-01-18 Richard Biener <rguenther@suse.de>
18867 PR tree-optimization/83887
18868 * graphite-scop-detection.c
18869 (scop_detection::get_nearest_dom_with_single_entry): Remove.
18870 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
18871 (scop_detection::merge_sese): Re-implement with a flood-fill
18872 algorithm that properly finds a SESE region if it exists.
18874 2018-01-18 Jakub Jelinek <jakub@redhat.com>
18877 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
18878 pointer_diff optimizations use view_convert instead of convert.
18880 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18882 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
18883 Generate different code for -mno-speculate-indirect-jumps.
18884 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
18885 (*call_indirect_aix<mode>): Disable for
18886 -mno-speculate-indirect-jumps.
18887 (*call_indirect_aix<mode>_nospec): New define_insn.
18888 (*call_value_indirect_aix<mode>): Disable for
18889 -mno-speculate-indirect-jumps.
18890 (*call_value_indirect_aix<mode>_nospec): New define_insn.
18891 (*sibcall_nonlocal_sysv<mode>): Generate different code for
18892 -mno-speculate-indirect-jumps.
18893 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
18895 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
18897 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
18898 long double type, set the flags for noting the default long double
18899 type, even if we don't pass or return a long double type.
18901 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
18904 * ipa-inline.c (flatten_function): Do not overwrite final inlining
18907 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
18909 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
18910 support for merge[hl].
18911 (fold_mergehl_helper): New helper function.
18912 (tree-vector-builder.h): New #include for tree_vector_builder usage.
18913 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
18914 (altivec_vmrglw_direct): Add xxmrglw insn.
18916 2018-01-17 Andrew Waterman <andrew@sifive.com>
18918 * config/riscv/riscv.c (riscv_conditional_register_usage): If
18919 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
18921 2018-01-17 David Malcolm <dmalcolm@redhat.com>
18924 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
18925 call the lto_location_cache before reading the
18926 DECL_SOURCE_LOCATION of the types.
18928 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
18929 Richard Sandiford <richard.sandiford@linaro.org>
18931 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
18932 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
18933 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
18934 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
18935 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
18937 * config/aarch64/constraints.md (aarch64_movti_operand):
18939 * config/aarch64/predicates.md (Uti): Add new constraint.
18941 2018-01-17 Carl Love <cel@us.ibm.com>
18943 * config/rs6000/vsx.md (define_expand xl_len_r,
18944 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
18945 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
18947 (define_expand, define_insn): Move the shift left from the
18948 define_insn to the define_expand for lxvl and stxvl instructions.
18949 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
18950 and XL_LEN_R definitions to PURE.
18952 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
18954 * config/i386/i386.c (indirect_thunk_name): Declare regno
18955 as unsigned int. Compare regno with INVALID_REGNUM.
18956 (output_indirect_thunk): Ditto.
18957 (output_indirect_thunk_function): Ditto.
18958 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
18959 in the call to output_indirect_thunk_function.
18961 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
18963 PR middle-end/83884
18964 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
18965 rather than the size of inner_type to determine the stack slot size
18966 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
18968 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
18971 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
18974 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
18976 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
18977 endian Linux systems to optionally enable multilibs for selecting
18978 the long double type if the user configured an explicit type.
18979 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
18980 have no long double multilibs if not defined.
18981 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
18982 warn if the user used -mabi={ieee,ibm}longdouble and we built
18983 multilibs for long double.
18984 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
18985 appropriate multilib option.
18986 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
18988 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
18989 for building long double multilibs.
18990 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
18992 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
18994 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
18997 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
18999 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
19002 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
19005 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
19008 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
19010 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
19011 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
19013 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
19015 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
19016 different rtl trees depending on TARGET_64BIT.
19017 (rs6000_gen_lvx): Likewise.
19019 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
19021 * config/visium/visium.md (nop): Tweak comment.
19022 (hazard_nop): Likewise.
19024 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19026 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
19027 -mspeculate-indirect-jumps.
19028 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
19029 for -mno-speculate-indirect-jumps.
19030 (*call_indirect_elfv2<mode>_nospec): New define_insn.
19031 (*call_value_indirect_elfv2<mode>): Disable for
19032 -mno-speculate-indirect-jumps.
19033 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
19034 (indirect_jump): Emit different RTL for
19035 -mno-speculate-indirect-jumps.
19036 (*indirect_jump<mode>): Disable for
19037 -mno-speculate-indirect-jumps.
19038 (*indirect_jump<mode>_nospec): New define_insn.
19039 (tablejump): Emit different RTL for
19040 -mno-speculate-indirect-jumps.
19041 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
19042 (tablejumpsi_nospec): New define_expand.
19043 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
19044 (tablejumpdi_nospec): New define_expand.
19045 (*tablejump<mode>_internal1): Disable for
19046 -mno-speculate-indirect-jumps.
19047 (*tablejump<mode>_internal1_nospec): New define_insn.
19048 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
19051 2018-01-16 Artyom Skrobov tyomitch@gmail.com
19053 * caller-save.c (insert_save): Drop unnecessary parameter. All
19056 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19057 Richard Biener <rguenth@suse.de>
19060 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
19061 return early, inline manually is_gimple_sizepos. Make sure if we
19062 call gimplify_expr we don't end up with a gimple constant.
19063 * tree.c (variably_modified_type_p): Don't return true for
19064 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
19065 * gimplify.h (is_gimple_sizepos): Remove.
19067 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19069 PR tree-optimization/83857
19070 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
19071 vectorizable_live_operation for pure SLP statements.
19072 (vectorizable_live_operation): Handle PHIs.
19074 2018-01-16 Richard Biener <rguenther@suse.de>
19076 PR tree-optimization/83867
19077 * tree-vect-stmts.c (vect_transform_stmt): Precompute
19078 nested_in_vect_loop_p since the scalar stmt may get invalidated.
19080 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19083 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
19084 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
19085 If off is not INTEGER_CST, issue a may not be aligned warning
19086 rather than isn't aligned. Use isn%'t rather than isn't.
19087 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
19089 <case MULT_EXPR>: Improve the case when bottom and one of the
19090 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
19091 operand, in that case check if the other operand is multiple of
19092 bottom divided by the INTEGER_CST operand.
19094 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19097 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
19098 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
19099 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
19100 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
19101 * config/pa/pa.c (pa_function_arg_advance): Likewise.
19102 (pa_function_arg, pa_arg_partial_bytes): Likewise.
19103 (pa_function_arg_size): New function.
19105 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19107 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
19108 in a separate statement.
19110 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
19112 PR tree-optimization/83847
19113 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
19114 group gathers and scatters.
19116 2018-01-16 Jakub Jelinek <jakub@redhat.com>
19118 PR rtl-optimization/86620
19119 * params.def (max-sched-ready-insns): Bump minimum value to 1.
19121 PR rtl-optimization/83213
19122 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
19123 to last if both are JUMP_INSNs.
19125 PR tree-optimization/83843
19126 * gimple-ssa-store-merging.c
19127 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
19128 store_immediate_info for bswap/nop orig_stores.
19130 2018-01-15 Andrew Waterman <andrew@sifive.com>
19132 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
19134 <UDIV>: Increase cost if !TARGET_DIV.
19136 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
19138 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
19139 (define_attr "cr_logical_3op"): New.
19140 (cceq_ior_compare): Adjust.
19141 (cceq_ior_compare_complement): Adjust.
19142 (*cceq_rev_compare): Adjust.
19143 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
19144 (is_cracked_insn): Adjust.
19145 (insn_must_be_first_in_group): Adjust.
19146 * config/rs6000/40x.md: Adjust.
19147 * config/rs6000/440.md: Adjust.
19148 * config/rs6000/476.md: Adjust.
19149 * config/rs6000/601.md: Adjust.
19150 * config/rs6000/603.md: Adjust.
19151 * config/rs6000/6xx.md: Adjust.
19152 * config/rs6000/7450.md: Adjust.
19153 * config/rs6000/7xx.md: Adjust.
19154 * config/rs6000/8540.md: Adjust.
19155 * config/rs6000/cell.md: Adjust.
19156 * config/rs6000/e300c2c3.md: Adjust.
19157 * config/rs6000/e500mc.md: Adjust.
19158 * config/rs6000/e500mc64.md: Adjust.
19159 * config/rs6000/e5500.md: Adjust.
19160 * config/rs6000/e6500.md: Adjust.
19161 * config/rs6000/mpc.md: Adjust.
19162 * config/rs6000/power4.md: Adjust.
19163 * config/rs6000/power5.md: Adjust.
19164 * config/rs6000/power6.md: Adjust.
19165 * config/rs6000/power7.md: Adjust.
19166 * config/rs6000/power8.md: Adjust.
19167 * config/rs6000/power9.md: Adjust.
19168 * config/rs6000/rs64.md: Adjust.
19169 * config/rs6000/titan.md: Adjust.
19171 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19173 * config/i386/predicates.md (indirect_branch_operand): Rewrite
19174 ix86_indirect_branch_register logic.
19176 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19178 * config/i386/constraints.md (Bs): Update
19179 ix86_indirect_branch_register check. Don't check
19180 ix86_indirect_branch_register with GOT_memory_operand.
19182 * config/i386/predicates.md (GOT_memory_operand): Don't check
19183 ix86_indirect_branch_register here.
19184 (GOT32_symbol_operand): Likewise.
19186 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19188 * config/i386/predicates.md (constant_call_address_operand):
19189 Rewrite ix86_indirect_branch_register logic.
19190 (sibcall_insn_operand): Likewise.
19192 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19194 * config/i386/constraints.md (Bs): Replace
19195 ix86_indirect_branch_thunk_register with
19196 ix86_indirect_branch_register.
19198 * config/i386/i386.md (indirect_jump): Likewise.
19199 (tablejump): Likewise.
19200 (*sibcall_memory): Likewise.
19201 (*sibcall_value_memory): Likewise.
19202 Peepholes of indirect call and jump via memory: Likewise.
19203 * config/i386/i386.opt: Likewise.
19204 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19205 (GOT_memory_operand): Likewise.
19206 (call_insn_operand): Likewise.
19207 (sibcall_insn_operand): Likewise.
19208 (GOT32_symbol_operand): Likewise.
19210 2018-01-15 Jakub Jelinek <jakub@redhat.com>
19212 PR middle-end/83837
19213 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
19214 type rather than type addr's type points to.
19215 (expand_omp_atomic_mutex): Likewise.
19216 (expand_omp_atomic): Likewise.
19218 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
19221 * config/i386/i386.c (output_indirect_thunk_function): Use
19222 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
19223 for __x86_return_thunk.
19225 2018-01-15 Richard Biener <rguenther@suse.de>
19227 PR middle-end/83850
19228 * expmed.c (extract_bit_field_1): Fix typo.
19230 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19233 * config/arm/iterators.md (VF): New mode iterator.
19234 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
19235 Remove integer-related logic from pattern.
19236 (neon_vabd<mode>_3): Likewise.
19238 2018-01-15 Jakub Jelinek <jakub@redhat.com>
19240 PR middle-end/82694
19241 * common.opt (fstrict-overflow): No longer an alias.
19242 (fwrapv-pointer): New option.
19243 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
19244 also for pointer types based on flag_wrapv_pointer.
19245 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
19246 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
19247 opts->x_flag_wrapv got set.
19248 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
19249 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
19250 POINTER_TYPE_OVERFLOW_UNDEFINED.
19251 * match.pd: Likewise in address comparison pattern.
19252 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
19254 2018-01-15 Richard Biener <rguenther@suse.de>
19257 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
19258 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
19259 Reset type names to their identifier if their TYPE_DECL doesn't
19260 have linkage (and thus is used for ODR and devirt).
19261 (save_debug_info_for_decl): Remove.
19262 (save_debug_info_for_type): Likewise.
19263 (add_tree_to_fld_list): Adjust.
19264 * tree-pretty-print.c (dump_generic_node): Make dumping of
19265 type names more robust.
19267 2018-01-15 Richard Biener <rguenther@suse.de>
19269 * BASE-VER: Bump to 8.0.1.
19271 2018-01-14 Martin Sebor <msebor@redhat.com>
19274 * builtins.c (check_access): Avoid warning when the no-warning bit
19277 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
19279 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
19280 * ira-color (allocno_hard_regs_compare): Likewise.
19282 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
19285 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
19286 Use .pushsection/.popsection.
19288 2018-01-14 Martin Sebor <msebor@redhat.com>
19291 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
19293 2018-01-14 Jakub Jelinek <jakub@redhat.com>
19295 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
19296 entry from extra_headers.
19297 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
19298 extra_headers, make the list bitwise identical to the i?86-*-* one.
19300 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19302 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
19303 -mcmodel=large with -mindirect-branch=thunk,
19304 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
19305 -mfunction-return=thunk-extern.
19306 * doc/invoke.texi: Document -mcmodel=large is incompatible with
19307 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
19308 -mfunction-return=thunk and -mfunction-return=thunk-extern.
19310 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19312 * config/i386/i386.c (print_reg): Print the name of the full
19313 integer register without '%'.
19314 (ix86_print_operand): Handle 'V'.
19315 * doc/extend.texi: Document 'V' modifier.
19317 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19319 * config/i386/constraints.md (Bs): Disallow memory operand for
19320 -mindirect-branch-register.
19322 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19323 (GOT_memory_operand): Likewise.
19324 (call_insn_operand): Likewise.
19325 (sibcall_insn_operand): Likewise.
19326 (GOT32_symbol_operand): Likewise.
19327 * config/i386/i386.md (indirect_jump): Call convert_memory_address
19328 for -mindirect-branch-register.
19329 (tablejump): Likewise.
19330 (*sibcall_memory): Likewise.
19331 (*sibcall_value_memory): Likewise.
19332 Disallow peepholes of indirect call and jump via memory for
19333 -mindirect-branch-register.
19334 (*call_pop): Replace m with Bw.
19335 (*call_value_pop): Likewise.
19336 (*sibcall_pop_memory): Replace m with Bs.
19337 * config/i386/i386.opt (mindirect-branch-register): New option.
19338 * doc/invoke.texi: Document -mindirect-branch-register option.
19340 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19342 * config/i386/i386-protos.h (ix86_output_function_return): New.
19343 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
19344 set function_return_type.
19345 (indirect_thunk_name): Add ret_p to indicate thunk for function
19347 (output_indirect_thunk_function): Pass false to
19348 indirect_thunk_name.
19349 (ix86_output_indirect_branch_via_reg): Likewise.
19350 (ix86_output_indirect_branch_via_push): Likewise.
19351 (output_indirect_thunk_function): Create alias for function
19352 return thunk if regno < 0.
19353 (ix86_output_function_return): New function.
19354 (ix86_handle_fndecl_attribute): Handle function_return.
19355 (ix86_attribute_table): Add function_return.
19356 * config/i386/i386.h (machine_function): Add
19357 function_return_type.
19358 * config/i386/i386.md (simple_return_internal): Use
19359 ix86_output_function_return.
19360 (simple_return_internal_long): Likewise.
19361 * config/i386/i386.opt (mfunction-return=): New option.
19362 (indirect_branch): Mention -mfunction-return=.
19363 * doc/extend.texi: Document function_return function attribute.
19364 * doc/invoke.texi: Document -mfunction-return= option.
19366 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
19368 * config/i386/i386-opts.h (indirect_branch): New.
19369 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
19370 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
19371 with local indirect jump when converting indirect call and jump.
19372 (ix86_set_indirect_branch_type): New.
19373 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
19374 (indirectlabelno): New.
19375 (indirect_thunk_needed): Likewise.
19376 (indirect_thunk_bnd_needed): Likewise.
19377 (indirect_thunks_used): Likewise.
19378 (indirect_thunks_bnd_used): Likewise.
19379 (INDIRECT_LABEL): Likewise.
19380 (indirect_thunk_name): Likewise.
19381 (output_indirect_thunk): Likewise.
19382 (output_indirect_thunk_function): Likewise.
19383 (ix86_output_indirect_branch_via_reg): Likewise.
19384 (ix86_output_indirect_branch_via_push): Likewise.
19385 (ix86_output_indirect_branch): Likewise.
19386 (ix86_output_indirect_jmp): Likewise.
19387 (ix86_code_end): Call output_indirect_thunk_function if needed.
19388 (ix86_output_call_insn): Call ix86_output_indirect_branch if
19390 (ix86_handle_fndecl_attribute): Handle indirect_branch.
19391 (ix86_attribute_table): Add indirect_branch.
19392 * config/i386/i386.h (machine_function): Add indirect_branch_type
19393 and has_local_indirect_jump.
19394 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
19396 (tablejump): Likewise.
19397 (*indirect_jump): Use ix86_output_indirect_jmp.
19398 (*tablejump_1): Likewise.
19399 (simple_return_indirect_internal): Likewise.
19400 * config/i386/i386.opt (mindirect-branch=): New option.
19401 (indirect_branch): New.
19404 (thunk-inline): Likewise.
19405 (thunk-extern): Likewise.
19406 * doc/extend.texi: Document indirect_branch function attribute.
19407 * doc/invoke.texi: Document -mindirect-branch= option.
19409 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
19412 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
19414 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
19416 * ipa-inline.c (want_inline_small_function_p): Return false if
19417 inlining has already failed with CIF_FINAL_ERROR.
19418 (update_caller_keys): Call want_inline_small_function_p before
19420 (update_callee_keys): Likewise.
19422 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
19424 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
19426 (rs6000_quadword_masked_address_p): Likewise.
19427 (quad_aligned_load_p): Likewise.
19428 (quad_aligned_store_p): Likewise.
19429 (const_load_sequence_p): Add comment to describe the outer-most loop.
19430 (mimic_memory_attributes_and_flags): New function.
19431 (rs6000_gen_stvx): Likewise.
19432 (replace_swapped_aligned_store): Likewise.
19433 (rs6000_gen_lvx): Likewise.
19434 (replace_swapped_aligned_load): Likewise.
19435 (replace_swapped_load_constant): Capitalize argument name in
19436 comment describing this function.
19437 (rs6000_analyze_swaps): Add a third pass to search for vector loads
19438 and stores that access quad-word aligned addresses and replace
19439 with stvx or lvx instructions when appropriate.
19440 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
19441 New function prototype.
19442 (rs6000_quadword_masked_address_p): Likewise.
19443 (rs6000_gen_lvx): Likewise.
19444 (rs6000_gen_stvx): Likewise.
19445 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
19446 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
19447 when memory address is aligned.
19448 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
19449 this split to select lvx instruction when memory address is aligned.
19450 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
19451 instruction when memory address is aligned.
19452 (*vsx_le_perm_load_v16qi): Likewise.
19453 (four unnamed splitters): Modify to select the stvx instruction
19454 when memory is aligned.
19456 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
19458 * predict.c (determine_unlikely_bbs): Handle correctly BBs
19459 which appears in the queue multiple times.
19461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19462 Alan Hayward <alan.hayward@arm.com>
19463 David Sherwood <david.sherwood@arm.com>
19465 * tree-vectorizer.h (vec_lower_bound): New structure.
19466 (_loop_vec_info): Add check_nonzero and lower_bounds.
19467 (LOOP_VINFO_CHECK_NONZERO): New macro.
19468 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
19469 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
19470 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
19471 fields. Make seg_len the distance travelled, not including the
19473 (dr_direction_indicator): Declare.
19474 (dr_zero_step_indicator): Likewise.
19475 (dr_known_forward_stride_p): Likewise.
19476 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
19478 (runtime_alias_check_p): Allow runtime alias checks with
19480 (operator ==): Compare access_size and align.
19481 (prune_runtime_alias_test_list): Rework for new distinction between
19482 the access_size and seg_len.
19483 (create_intersect_range_checks_index): Likewise. Cope with polynomial
19485 (get_segment_min_max): New function.
19486 (create_intersect_range_checks): Use it.
19487 (dr_step_indicator): New function.
19488 (dr_direction_indicator): Likewise.
19489 (dr_zero_step_indicator): Likewise.
19490 (dr_known_forward_stride_p): Likewise.
19491 * tree-loop-distribution.c (data_ref_segment_size): Return
19492 DR_STEP * (niters - 1).
19493 (compute_alias_check_pairs): Update call to the dr_with_seg_len
19495 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
19496 (vect_preserves_scalar_order_p): New function, split out from...
19497 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
19498 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
19499 (vect_vfa_access_size): New function.
19500 (vect_vfa_align): Likewise.
19501 (vect_compile_time_alias): Take access_size_a and access_b arguments.
19502 (dump_lower_bound): New function.
19503 (vect_check_lower_bound): Likewise.
19504 (vect_small_gap_p): Likewise.
19505 (vectorizable_with_step_bound_p): Likewise.
19506 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
19507 depencies if the vectorization factor is 1. Convert the checks
19508 for nonzero steps into checks on the bounds of DR_STEP. Try using
19509 a bunds check for variable steps if the minimum required step is
19510 relatively small. Update calls to the dr_with_seg_len
19511 constructor and to vect_compile_time_alias.
19512 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
19514 (vect_loop_versioning): Call it.
19515 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
19517 (vect_estimate_min_profitable_iters): Account for any bounds checks.
19519 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19520 Alan Hayward <alan.hayward@arm.com>
19521 David Sherwood <david.sherwood@arm.com>
19523 * doc/sourcebuild.texi (vect_scatter_store): Document.
19524 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
19526 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
19528 * genopinit.c (main): Add supports_vec_scatter_store and
19529 supports_vec_scatter_store_cached to target_optabs.
19530 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
19531 IFN_MASK_SCATTER_STORE.
19532 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
19534 * internal-fn.h (internal_store_fn_p): Declare.
19535 (internal_fn_stored_value_index): Likewise.
19536 * internal-fn.c (scatter_store_direct): New macro.
19537 (expand_scatter_store_optab_fn): New function.
19538 (direct_scatter_store_optab_supported_p): New macro.
19539 (internal_store_fn_p): New function.
19540 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
19541 IFN_MASK_SCATTER_STORE.
19542 (internal_fn_mask_index): Likewise.
19543 (internal_fn_stored_value_index): New function.
19544 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
19545 for scatter stores.
19546 * optabs-query.h (supports_vec_scatter_store_p): Declare.
19547 * optabs-query.c (supports_vec_scatter_store_p): New function.
19548 * tree-vectorizer.h (vect_get_store_rhs): Declare.
19549 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
19550 true for scatter stores.
19551 (vect_gather_scatter_fn_p): Handle scatter stores too.
19552 (vect_check_gather_scatter): Consider using scatter stores if
19553 supports_vec_scatter_store_p.
19554 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
19555 scatter stores too.
19556 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19557 internal_fn_stored_value_index.
19558 (check_load_store_masking): Handle scatter stores too.
19559 (vect_get_store_rhs): Make public.
19560 (vectorizable_call): Use internal_store_fn_p.
19561 (vectorizable_store): Handle scatter store internal functions.
19562 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
19563 when deciding whether the end of the group has been reached.
19564 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
19565 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
19566 (mask_scatter_store<mode>): New insns.
19568 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19569 Alan Hayward <alan.hayward@arm.com>
19570 David Sherwood <david.sherwood@arm.com>
19572 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
19573 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
19574 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
19576 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
19577 Use vect_truncate_gather_scatter_offset if we can't treat the
19578 operation as a normal gather load or scatter store.
19579 (get_group_load_store_type): Take the gather_scatter_info
19580 as argument. Try using a gather load or scatter store for
19581 single-element groups.
19582 (get_load_store_type): Update calls to get_group_load_store_type
19583 and vect_use_strided_gather_scatters_p.
19585 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19586 Alan Hayward <alan.hayward@arm.com>
19587 David Sherwood <david.sherwood@arm.com>
19589 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
19590 optional tree argument.
19591 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
19593 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
19594 but continue to use the current value as a fallback.
19595 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
19596 to compare the updates.
19597 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
19598 (get_load_store_type): Use it when handling a strided access.
19599 (vect_get_strided_load_store_ops): New function.
19600 (vect_get_data_ptr_increment): Likewise.
19601 (vectorizable_load): Handle strided gather loads. Always pass
19602 a step to vect_create_data_ref_ptr and bump_vector_ptr.
19604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19605 Alan Hayward <alan.hayward@arm.com>
19606 David Sherwood <david.sherwood@arm.com>
19608 * doc/md.texi (gather_load@var{m}): Document.
19609 (mask_gather_load@var{m}): Likewise.
19610 * genopinit.c (main): Add supports_vec_gather_load and
19611 supports_vec_gather_load_cached to target_optabs.
19612 * optabs-tree.c (init_tree_optimization_optabs): Use
19613 ggc_cleared_alloc to allocate target_optabs.
19614 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
19615 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
19617 * internal-fn.h (internal_load_fn_p): Declare.
19618 (internal_gather_scatter_fn_p): Likewise.
19619 (internal_fn_mask_index): Likewise.
19620 (internal_gather_scatter_fn_supported_p): Likewise.
19621 * internal-fn.c (gather_load_direct): New macro.
19622 (expand_gather_load_optab_fn): New function.
19623 (direct_gather_load_optab_supported_p): New macro.
19624 (direct_internal_fn_optab): New function.
19625 (internal_load_fn_p): Likewise.
19626 (internal_gather_scatter_fn_p): Likewise.
19627 (internal_fn_mask_index): Likewise.
19628 (internal_gather_scatter_fn_supported_p): Likewise.
19629 * optabs-query.c (supports_at_least_one_mode_p): New function.
19630 (supports_vec_gather_load_p): Likewise.
19631 * optabs-query.h (supports_vec_gather_load_p): Declare.
19632 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
19633 and memory_type field.
19634 (NUM_PATTERNS): Bump to 15.
19635 * tree-vect-data-refs.c: Include internal-fn.h.
19636 (vect_gather_scatter_fn_p): New function.
19637 (vect_describe_gather_scatter_call): Likewise.
19638 (vect_check_gather_scatter): Try using internal functions for
19639 gather loads. Recognize existing calls to a gather load function.
19640 (vect_analyze_data_refs): Consider using gather loads if
19641 supports_vec_gather_load_p.
19642 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
19643 (vect_get_gather_scatter_offset_type): Likewise.
19644 (vect_convert_mask_for_vectype): Likewise.
19645 (vect_add_conversion_to_patterm): Likewise.
19646 (vect_try_gather_scatter_pattern): Likewise.
19647 (vect_recog_gather_scatter_pattern): New pattern recognizer.
19648 (vect_vect_recog_func_ptrs): Add it.
19649 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
19650 internal_fn_mask_index and internal_gather_scatter_fn_p.
19651 (check_load_store_masking): Take the gather_scatter_info as an
19652 argument and handle gather loads.
19653 (vect_get_gather_scatter_ops): New function.
19654 (vectorizable_call): Check internal_load_fn_p.
19655 (vectorizable_load): Likewise. Handle gather load internal
19657 (vectorizable_store): Update call to check_load_store_masking.
19658 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
19659 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
19660 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
19661 (aarch64_gather_scale_operand_d): New predicates.
19662 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
19663 (mask_gather_load<mode>): New insns.
19665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19666 Alan Hayward <alan.hayward@arm.com>
19667 David Sherwood <david.sherwood@arm.com>
19669 * optabs.def (fold_left_plus_optab): New optab.
19670 * doc/md.texi (fold_left_plus_@var{m}): Document.
19671 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
19672 * internal-fn.c (fold_left_direct): Define.
19673 (expand_fold_left_optab_fn): Likewise.
19674 (direct_fold_left_optab_supported_p): Likewise.
19675 * fold-const-call.c (fold_const_fold_left): New function.
19676 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
19677 * tree-parloops.c (valid_reduction_p): New function.
19678 (gather_scalar_reductions): Use it.
19679 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
19680 (vect_finish_replace_stmt): Declare.
19681 * tree-vect-loop.c (fold_left_reduction_fn): New function.
19682 (needs_fold_left_reduction_p): New function, split out from...
19683 (vect_is_simple_reduction): ...here. Accept reductions that
19684 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
19685 (vect_force_simple_reduction): Also store the reduction type in
19686 the assignment's STMT_VINFO_REDUC_TYPE.
19687 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
19688 (merge_with_identity): New function.
19689 (vect_expand_fold_left): Likewise.
19690 (vectorize_fold_left_reduction): Likewise.
19691 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
19692 scalar phi in place for it. Check for target support and reject
19693 cases that would reassociate the operation. Defer the transform
19694 phase to vectorize_fold_left_reduction.
19695 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
19696 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
19697 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
19699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19701 * tree-if-conv.c (predicate_mem_writes): Remove redundant
19702 call to ifc_temp_var.
19704 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19705 Alan Hayward <alan.hayward@arm.com>
19706 David Sherwood <david.sherwood@arm.com>
19708 * target.def (legitimize_address_displacement): Take the original
19709 offset as a poly_int.
19710 * targhooks.h (default_legitimize_address_displacement): Update
19712 * targhooks.c (default_legitimize_address_displacement): Likewise.
19713 * doc/tm.texi: Regenerate.
19714 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
19715 as an argument, moving assert of ad->disp == ad->disp_term to...
19716 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
19717 Try calling targetm.legitimize_address_displacement before expanding
19718 the address rather than afterwards, and adjust for the new interface.
19719 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
19720 Match the new hook interface. Handle SVE addresses.
19721 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
19722 new hook interface.
19724 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19726 * Makefile.in (OBJS): Add early-remat.o.
19727 * target.def (select_early_remat_modes): New hook.
19728 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
19729 * doc/tm.texi: Regenerate.
19730 * targhooks.h (default_select_early_remat_modes): Declare.
19731 * targhooks.c (default_select_early_remat_modes): New function.
19732 * timevar.def (TV_EARLY_REMAT): New timevar.
19733 * passes.def (pass_early_remat): New pass.
19734 * tree-pass.h (make_pass_early_remat): Declare.
19735 * early-remat.c: New file.
19736 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
19738 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
19740 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19741 Alan Hayward <alan.hayward@arm.com>
19742 David Sherwood <david.sherwood@arm.com>
19744 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
19745 vfm1 with a bound_epilog parameter.
19746 (vect_do_peeling): Update calls accordingly, and move the prologue
19747 call earlier in the function. Treat the base bound_epilog as 0 for
19748 fully-masked loops and retain vf - 1 for other loops. Add 1 to
19749 this base when peeling for gaps.
19750 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
19751 with fully-masked loops.
19752 (vect_estimate_min_profitable_iters): Handle the single peeled
19753 iteration in that case.
19755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19756 Alan Hayward <alan.hayward@arm.com>
19757 David Sherwood <david.sherwood@arm.com>
19759 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
19760 single-element interleaving even if the size is not a power of 2.
19761 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
19762 accesses for single-element interleaving if the group size is
19765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19766 Alan Hayward <alan.hayward@arm.com>
19767 David Sherwood <david.sherwood@arm.com>
19769 * doc/md.texi (fold_extract_last_@var{m}): Document.
19770 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
19771 * optabs.def (fold_extract_last_optab): New optab.
19772 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
19773 * internal-fn.c (fold_extract_direct): New macro.
19774 (expand_fold_extract_optab_fn): Likewise.
19775 (direct_fold_extract_optab_supported_p): Likewise.
19776 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
19777 * tree-vect-loop.c (vect_model_reduction_cost): Handle
19778 EXTRACT_LAST_REDUCTION.
19779 (get_initial_def_for_reduction): Do not create an initial vector
19780 for EXTRACT_LAST_REDUCTION reductions.
19781 (vectorizable_reduction): Leave the scalar phi in place for
19782 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
19783 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
19784 epilogue code for EXTRACT_LAST_REDUCTION and defer the
19785 transform phase to vectorizable_condition.
19786 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
19788 (vect_finish_stmt_generation): ...here.
19789 (vect_finish_replace_stmt): New function.
19790 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
19791 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
19793 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
19795 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19796 Alan Hayward <alan.hayward@arm.com>
19797 David Sherwood <david.sherwood@arm.com>
19799 * doc/md.texi (extract_last_@var{m}): Document.
19800 * optabs.def (extract_last_optab): New optab.
19801 * internal-fn.def (EXTRACT_LAST): New internal function.
19802 * internal-fn.c (cond_unary_direct): New macro.
19803 (expand_cond_unary_optab_fn): Likewise.
19804 (direct_cond_unary_optab_supported_p): Likewise.
19805 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
19806 loops using EXTRACT_LAST.
19807 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
19808 (extract_last_<mode>): ...this optab.
19809 (vec_extract<mode><Vel>): Update accordingly.
19811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19812 Alan Hayward <alan.hayward@arm.com>
19813 David Sherwood <david.sherwood@arm.com>
19815 * target.def (empty_mask_is_expensive): New hook.
19816 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
19817 * doc/tm.texi: Regenerate.
19818 * targhooks.h (default_empty_mask_is_expensive): Declare.
19819 * targhooks.c (default_empty_mask_is_expensive): New function.
19820 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
19821 if the target says that empty masks are expensive.
19822 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
19824 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
19826 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19827 Alan Hayward <alan.hayward@arm.com>
19828 David Sherwood <david.sherwood@arm.com>
19830 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
19831 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
19832 (vect_use_loop_mask_for_alignment_p): New function.
19833 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
19834 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
19835 niters_skip argument. Make sure that the first niters_skip elements
19836 of the first iteration are inactive.
19837 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
19838 Update call to vect_set_loop_masks_directly.
19839 (get_misalign_in_elems): New function, split out from...
19840 (vect_gen_prolog_loop_niters): ...here.
19841 (vect_update_init_of_dr): Take a code argument that specifies whether
19842 the adjustment should be added or subtracted.
19843 (vect_update_init_of_drs): Likewise.
19844 (vect_prepare_for_masked_peels): New function.
19845 (vect_do_peeling): Skip prologue peeling if we're using a mask
19846 instead. Update call to vect_update_inits_of_drs.
19847 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19849 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
19850 alignment. Do not include the number of peeled iterations in
19851 the minimum threshold in that case.
19852 (vectorizable_induction): Adjust the start value down by
19853 LOOP_VINFO_MASK_SKIP_NITERS iterations.
19854 (vect_transform_loop): Call vect_prepare_for_masked_peels.
19855 Take the number of skipped iterations into account when calculating
19857 * tree-vect-stmts.c (vect_gen_while_not): New function.
19859 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19860 Alan Hayward <alan.hayward@arm.com>
19861 David Sherwood <david.sherwood@arm.com>
19863 * doc/sourcebuild.texi (vect_fully_masked): Document.
19864 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
19865 default value to 0.
19866 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
19868 (vect_analyze_loop_2): ...here. Don't check the vectorization
19869 factor against the number of loop iterations if the loop is
19872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19873 Alan Hayward <alan.hayward@arm.com>
19874 David Sherwood <david.sherwood@arm.com>
19876 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
19877 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
19878 (dump_groups): Update accordingly.
19879 (iv_use::mem_type): New member variable.
19880 (address_p): New function.
19881 (record_use): Add a mem_type argument and initialize the new
19883 (record_group_use): Add a mem_type argument. Use address_p.
19884 Remove obsolete null checks of base_object. Update call to record_use.
19885 (find_interesting_uses_op): Update call to record_group_use.
19886 (find_interesting_uses_cond): Likewise.
19887 (find_interesting_uses_address): Likewise.
19888 (get_mem_type_for_internal_fn): New function.
19889 (find_address_like_use): Likewise.
19890 (find_interesting_uses_stmt): Try find_address_like_use before
19891 calling find_interesting_uses_op.
19892 (addr_offset_valid_p): Use the iv mem_type field as the type
19893 of the addressed memory.
19894 (add_autoinc_candidates): Likewise.
19895 (get_address_cost): Likewise.
19896 (split_small_address_groups_p): Use address_p.
19897 (split_address_groups): Likewise.
19898 (add_iv_candidate_for_use): Likewise.
19899 (autoinc_possible_for_pair): Likewise.
19900 (rewrite_groups): Likewise.
19901 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
19902 (determine_group_iv_cost): Update after split of USE_ADDRESS.
19903 (get_alias_ptr_type_for_ptr_address): New function.
19904 (rewrite_use_address): Rewrite address uses in calls that were
19905 identified by find_address_like_use.
19907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19908 Alan Hayward <alan.hayward@arm.com>
19909 David Sherwood <david.sherwood@arm.com>
19911 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
19913 * gimple-expr.h (is_gimple_addressable: Likewise.
19914 * gimple-expr.c (is_gimple_address): Likewise.
19915 * internal-fn.c (expand_call_mem_ref): New function.
19916 (expand_mask_load_optab_fn): Use it.
19917 (expand_mask_store_optab_fn): Likewise.
19919 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19920 Alan Hayward <alan.hayward@arm.com>
19921 David Sherwood <david.sherwood@arm.com>
19923 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
19924 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
19925 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
19926 (cond_umax@var{mode}): Document.
19927 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
19928 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
19929 (cond_umin_optab, cond_umax_optab): New optabs.
19930 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
19931 (COND_IOR, COND_XOR): New internal functions.
19932 * internal-fn.h (get_conditional_internal_fn): Declare.
19933 * internal-fn.c (cond_binary_direct): New macro.
19934 (expand_cond_binary_optab_fn): Likewise.
19935 (direct_cond_binary_optab_supported_p): Likewise.
19936 (get_conditional_internal_fn): New function.
19937 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
19938 Cope with reduction statements that are vectorized as calls rather
19940 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
19941 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
19942 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
19943 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
19944 (UNSPEC_COND_EOR): New unspecs.
19945 (optab): Add mappings for them.
19946 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
19947 (sve_int_op, sve_fp_op): New int attributes.
19949 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
19950 Alan Hayward <alan.hayward@arm.com>
19951 David Sherwood <david.sherwood@arm.com>
19953 * optabs.def (while_ult_optab): New optab.
19954 * doc/md.texi (while_ult@var{m}@var{n}): Document.
19955 * internal-fn.def (WHILE_ULT): New internal function.
19956 * internal-fn.h (direct_internal_fn_supported_p): New override
19957 that takes two types as argument.
19958 * internal-fn.c (while_direct): New macro.
19959 (expand_while_optab_fn): New function.
19960 (convert_optab_supported_p): Likewise.
19961 (direct_while_optab_supported_p): New macro.
19962 * wide-int.h (wi::udiv_ceil): New function.
19963 * tree-vectorizer.h (rgroup_masks): New structure.
19964 (vec_loop_masks): New typedef.
19965 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
19966 and fully_masked_p.
19967 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
19968 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
19969 (vect_max_vf): New function.
19970 (slpeel_make_loop_iterate_ntimes): Delete.
19971 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
19972 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
19973 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
19974 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
19975 internal-fn.h, stor-layout.h and optabs-query.h.
19976 (vect_set_loop_mask): New function.
19977 (add_preheader_seq): Likewise.
19978 (add_header_seq): Likewise.
19979 (interleave_supported_p): Likewise.
19980 (vect_maybe_permute_loop_masks): Likewise.
19981 (vect_set_loop_masks_directly): Likewise.
19982 (vect_set_loop_condition_masked): Likewise.
19983 (vect_set_loop_condition_unmasked): New function, split out from
19984 slpeel_make_loop_iterate_ntimes.
19985 (slpeel_make_loop_iterate_ntimes): Rename to..
19986 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
19987 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
19988 (vect_do_peeling): Update call accordingly.
19989 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
19991 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19992 mask_compare_type, can_fully_mask_p and fully_masked_p.
19993 (release_vec_loop_masks): New function.
19994 (_loop_vec_info): Use it to free the loop masks.
19995 (can_produce_all_loop_masks_p): New function.
19996 (vect_get_max_nscalars_per_iter): Likewise.
19997 (vect_verify_full_masking): Likewise.
19998 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
19999 retries, and free the mask rgroups before retrying. Check loop-wide
20000 reasons for disallowing fully-masked loops. Make the final decision
20001 about whether use a fully-masked loop or not.
20002 (vect_estimate_min_profitable_iters): Do not assume that peeling
20003 for the number of iterations will be needed for fully-masked loops.
20004 (vectorizable_reduction): Disable fully-masked loops.
20005 (vectorizable_live_operation): Likewise.
20006 (vect_halve_mask_nunits): New function.
20007 (vect_double_mask_nunits): Likewise.
20008 (vect_record_loop_mask): Likewise.
20009 (vect_get_loop_mask): Likewise.
20010 (vect_transform_loop): Handle the case in which the final loop
20011 iteration might handle a partial vector. Call vect_set_loop_condition
20012 instead of slpeel_make_loop_iterate_ntimes.
20013 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
20014 (check_load_store_masking): New function.
20015 (prepare_load_store_mask): Likewise.
20016 (vectorizable_store): Handle fully-masked loops.
20017 (vectorizable_load): Likewise.
20018 (supportable_widening_operation): Use vect_halve_mask_nunits for
20020 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
20021 (vect_gen_while): New function.
20022 * config/aarch64/aarch64.md (umax<mode>3): New expander.
20023 (aarch64_uqdec<mode>): New insn.
20025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20026 Alan Hayward <alan.hayward@arm.com>
20027 David Sherwood <david.sherwood@arm.com>
20029 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
20030 (reduc_xor_scal_optab): New optabs.
20031 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
20032 (reduc_xor_scal_@var{m}): Document.
20033 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
20034 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
20035 internal functions.
20036 * fold-const-call.c (fold_const_call): Handle them.
20037 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
20038 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
20039 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
20040 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
20041 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
20042 (UNSPEC_XORV): New unspecs.
20043 (optab): Add entries for them.
20044 (BITWISEV): New int iterator.
20045 (bit_reduc_op): New int attributes.
20047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20048 Alan Hayward <alan.hayward@arm.com>
20049 David Sherwood <david.sherwood@arm.com>
20051 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
20052 * internal-fn.def (VEC_SHL_INSERT): New internal function.
20053 * optabs.def (vec_shl_insert_optab): New optab.
20054 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
20055 (duplicate_and_interleave): Likewise.
20056 * tree-vect-loop.c: Include internal-fn.h.
20057 (neutral_op_for_slp_reduction): New function, split out from
20058 get_initial_defs_for_reduction.
20059 (get_initial_def_for_reduction): Handle option 2 for variable-length
20060 vectors by loading the neutral value into a vector and then shifting
20061 the initial value into element 0.
20062 (get_initial_defs_for_reduction): Replace the code argument with
20063 the neutral value calculated by neutral_op_for_slp_reduction.
20064 Use gimple_build_vector for constant-length vectors.
20065 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
20066 but the first group_size elements have a neutral value.
20067 Use duplicate_and_interleave otherwise.
20068 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
20069 Update call to get_initial_defs_for_reduction. Handle SLP
20070 reductions for variable-length vectors by creating one vector
20071 result for each scalar result, with the elements associated
20072 with other scalar results stubbed out with the neutral value.
20073 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
20074 Require IFN_VEC_SHL_INSERT for double reductions on
20075 variable-length vectors, or SLP reductions that have
20076 a neutral value. Require can_duplicate_and_interleave_p
20077 support for variable-length unchained SLP reductions if there
20078 is no neutral value, such as for MIN/MAX reductions. Also require
20079 the number of vector elements to be a multiple of the number of
20080 SLP statements when doing variable-length unchained SLP reductions.
20081 Update call to vect_create_epilog_for_reduction.
20082 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
20083 and remove initial values.
20084 (duplicate_and_interleave): Make public.
20085 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
20086 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
20088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20089 Alan Hayward <alan.hayward@arm.com>
20090 David Sherwood <david.sherwood@arm.com>
20092 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
20093 (can_duplicate_and_interleave_p): New function.
20094 (vect_get_and_check_slp_defs): Take the vector of statements
20095 rather than just the current one. Remove excess parentheses.
20096 Restriction rejectinon of vect_constant_def and vect_external_def
20097 for variable-length vectors to boolean types, or types for which
20098 can_duplicate_and_interleave_p is false.
20099 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
20100 (duplicate_and_interleave): New function.
20101 (vect_get_constant_vectors): Use gimple_build_vector for
20102 constant-length vectors and suitable variable-length constant
20103 vectors. Use duplicate_and_interleave for other variable-length
20104 vectors. Don't defer the update when inserting new statements.
20106 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20107 Alan Hayward <alan.hayward@arm.com>
20108 David Sherwood <david.sherwood@arm.com>
20110 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
20111 min_profitable_iters doesn't go negative.
20113 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20114 Alan Hayward <alan.hayward@arm.com>
20115 David Sherwood <david.sherwood@arm.com>
20117 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
20118 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
20119 * optabs.def (vec_mask_load_lanes_optab): New optab.
20120 (vec_mask_store_lanes_optab): Likewise.
20121 * internal-fn.def (MASK_LOAD_LANES): New internal function.
20122 (MASK_STORE_LANES): Likewise.
20123 * internal-fn.c (mask_load_lanes_direct): New macro.
20124 (mask_store_lanes_direct): Likewise.
20125 (expand_mask_load_optab_fn): Handle masked operations.
20126 (expand_mask_load_lanes_optab_fn): New macro.
20127 (expand_mask_store_optab_fn): Handle masked operations.
20128 (expand_mask_store_lanes_optab_fn): New macro.
20129 (direct_mask_load_lanes_optab_supported_p): Likewise.
20130 (direct_mask_store_lanes_optab_supported_p): Likewise.
20131 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
20133 (vect_load_lanes_supported): Likewise.
20134 * tree-vect-data-refs.c (strip_conversion): New function.
20135 (can_group_stmts_p): Likewise.
20136 (vect_analyze_data_ref_accesses): Use it instead of checking
20137 for a pair of assignments.
20138 (vect_store_lanes_supported): Take a masked_p parameter.
20139 (vect_load_lanes_supported): Likewise.
20140 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
20141 vect_store_lanes_supported and vect_load_lanes_supported.
20142 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
20143 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
20144 parameter. Don't allow gaps for masked accesses.
20145 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
20146 and vect_load_lanes_supported.
20147 (get_load_store_type): Take a masked_p parameter and update
20148 call to get_group_load_store_type.
20149 (vectorizable_store): Update call to get_load_store_type.
20150 Handle IFN_MASK_STORE_LANES.
20151 (vectorizable_load): Update call to get_load_store_type.
20152 Handle IFN_MASK_LOAD_LANES.
20154 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20155 Alan Hayward <alan.hayward@arm.com>
20156 David Sherwood <david.sherwood@arm.com>
20158 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
20160 * config/aarch64/aarch64-protos.h
20161 (aarch64_sve_struct_memory_operand_p): Declare.
20162 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
20163 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
20164 (VPRED, vpred): Handle SVE structure modes.
20165 * config/aarch64/constraints.md (Utx): New constraint.
20166 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
20167 (aarch64_sve_struct_nonimmediate_operand): New predicates.
20168 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
20169 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
20170 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
20171 structure modes. Split into pieces after RA.
20172 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
20173 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
20175 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
20176 SVE structure modes.
20177 (aarch64_classify_address): Likewise.
20178 (sizetochar): Move earlier in file.
20179 (aarch64_print_operand): Handle SVE register lists.
20180 (aarch64_array_mode): New function.
20181 (aarch64_sve_struct_memory_operand_p): Likewise.
20182 (TARGET_ARRAY_MODE): Redefine.
20184 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20185 Alan Hayward <alan.hayward@arm.com>
20186 David Sherwood <david.sherwood@arm.com>
20188 * target.def (array_mode): New target hook.
20189 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
20190 * doc/tm.texi: Regenerate.
20191 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
20192 * hooks.c (hook_optmode_mode_uhwi_none): New function.
20193 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
20194 targetm.array_mode.
20195 * stor-layout.c (mode_for_array): Likewise. Support polynomial
20198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20199 Alan Hayward <alan.hayward@arm.com>
20200 David Sherwood <david.sherwood@arm.com>
20202 * fold-const.c (fold_binary_loc): Check the argument types
20203 rather than the result type when testing for a vector operation.
20205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20207 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
20208 * doc/tm.texi: Regenerate.
20210 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20211 Alan Hayward <alan.hayward@arm.com>
20212 David Sherwood <david.sherwood@arm.com>
20214 * doc/invoke.texi (-msve-vector-bits=): Document new option.
20215 (sve): Document new AArch64 extension.
20216 * doc/md.texi (w): Extend the description of the AArch64
20217 constraint to include SVE vectors.
20218 (Upl, Upa): Document new AArch64 predicate constraints.
20219 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
20221 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
20222 (msve-vector-bits=): New option.
20223 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
20224 SVE when these are disabled.
20225 (sve): New extension.
20226 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
20227 modes. Adjust their number of units based on aarch64_sve_vg.
20228 (MAX_BITSIZE_MODE_ANY_MODE): Define.
20229 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
20230 aarch64_addr_query_type.
20231 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
20232 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
20233 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
20234 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
20235 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
20236 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
20237 (aarch64_simd_imm_zero_p): Delete.
20238 (aarch64_check_zero_based_sve_index_immediate): Declare.
20239 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
20240 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
20241 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
20242 (aarch64_sve_float_mul_immediate_p): Likewise.
20243 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
20244 rather than an rtx.
20245 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
20246 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
20247 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
20248 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
20249 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
20250 (aarch64_regmode_natural_size): Likewise.
20251 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
20252 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
20254 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
20255 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
20256 for VG and the SVE predicate registers.
20257 (V_ALIASES): Add a "z"-prefixed alias.
20258 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
20259 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
20260 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
20261 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
20262 (REG_CLASS_NAMES): Add entries for them.
20263 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
20264 and the predicate registers.
20265 (aarch64_sve_vg): Declare.
20266 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
20267 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
20268 (REGMODE_NATURAL_SIZE): Define.
20269 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
20271 * config/aarch64/aarch64.c: Include cfgrtl.h.
20272 (simd_immediate_info): Add a constructor for series vectors,
20273 and an associated step field.
20274 (aarch64_sve_vg): New variable.
20275 (aarch64_dbx_register_number): Handle VG and the predicate registers.
20276 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
20277 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
20278 (VEC_ANY_DATA, VEC_STRUCT): New constants.
20279 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
20280 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
20281 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
20282 (aarch64_get_mask_mode): New functions.
20283 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
20284 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
20285 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
20286 predicate modes and predicate registers. Explicitly restrict
20287 GPRs to modes of 16 bytes or smaller. Only allow FP registers
20288 to store a vector mode if it is recognized by
20289 aarch64_classify_vector_mode.
20290 (aarch64_regmode_natural_size): New function.
20291 (aarch64_hard_regno_caller_save_mode): Return the original mode
20293 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
20294 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
20295 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
20296 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
20298 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
20299 does not overlap dest if the function is frame-related. Handle
20301 (aarch64_split_add_offset): New function.
20302 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
20303 them aarch64_add_offset.
20304 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
20305 and update call to aarch64_sub_sp.
20306 (aarch64_add_cfa_expression): New function.
20307 (aarch64_expand_prologue): Pass extra temporary registers to the
20308 functions above. Handle the case in which we need to emit new
20309 DW_CFA_expressions for registers that were originally saved
20310 relative to the stack pointer, but now have to be expressed
20311 relative to the frame pointer.
20312 (aarch64_output_mi_thunk): Pass extra temporary registers to the
20314 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
20315 IP0 and IP1 values for SVE frames.
20316 (aarch64_expand_vec_series): New function.
20317 (aarch64_expand_sve_widened_duplicate): Likewise.
20318 (aarch64_expand_sve_const_vector): Likewise.
20319 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
20320 Handle SVE constants. Use emit_move_insn to move a force_const_mem
20321 into the register, rather than emitting a SET directly.
20322 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
20323 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
20324 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
20325 (offset_9bit_signed_scaled_p): New functions.
20326 (aarch64_replicate_bitmask_imm): New function.
20327 (aarch64_bitmask_imm): Use it.
20328 (aarch64_cannot_force_const_mem): Reject expressions involving
20329 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
20330 (aarch64_classify_index): Handle SVE indices, by requiring
20331 a plain register index with a scale that matches the element size.
20332 (aarch64_classify_address): Handle SVE addresses. Assert that
20333 the mode of the address is VOIDmode or an integer mode.
20334 Update call to aarch64_classify_symbol.
20335 (aarch64_classify_symbolic_expression): Update call to
20336 aarch64_classify_symbol.
20337 (aarch64_const_vec_all_in_range_p): New function.
20338 (aarch64_print_vector_float_operand): Likewise.
20339 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
20340 "vN" for FP registers with SVE modes. Handle (const ...) vectors
20341 and the FP immediates 1.0 and 0.5.
20342 (aarch64_print_address_internal): Handle SVE addresses.
20343 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
20344 (aarch64_regno_regclass): Handle predicate registers.
20345 (aarch64_secondary_reload): Handle big-endian reloads of SVE
20347 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
20348 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
20349 (aarch64_convert_sve_vector_bits): New function.
20350 (aarch64_override_options): Use it to handle -msve-vector-bits=.
20351 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
20352 rather than an rtx.
20353 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
20354 Handle SVE vector and predicate modes. Accept VL-based constants
20355 that need only one temporary register, and VL offsets that require
20356 no temporary registers.
20357 (aarch64_conditional_register_usage): Mark the predicate registers
20358 as fixed if SVE isn't available.
20359 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
20360 Return true for SVE vector and predicate modes.
20361 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
20362 rather than an unsigned int. Handle SVE modes.
20363 (aarch64_preferred_simd_mode): Update call accordingly. Handle
20365 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
20367 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
20368 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
20369 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
20370 (aarch64_sve_float_mul_immediate_p): New functions.
20371 (aarch64_sve_valid_immediate): New function.
20372 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
20373 Explicitly reject structure modes. Check for INDEX constants.
20374 Handle PTRUE and PFALSE constants.
20375 (aarch64_check_zero_based_sve_index_immediate): New function.
20376 (aarch64_simd_imm_zero_p): Delete.
20377 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
20378 vector modes. Accept constants in the range of CNT[BHWD].
20379 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
20380 ask for an Advanced SIMD mode.
20381 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
20382 (aarch64_simd_vector_alignment): Handle SVE predicates.
20383 (aarch64_vectorize_preferred_vector_alignment): New function.
20384 (aarch64_simd_vector_alignment_reachable): Use it instead of
20386 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
20387 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
20389 (MAX_VECT_LEN): Delete.
20390 (expand_vec_perm_d): Add a vec_flags field.
20391 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
20392 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
20393 (aarch64_evpc_ext): Don't apply a big-endian lane correction
20395 (aarch64_evpc_rev): Rename to...
20396 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
20397 (aarch64_evpc_rev_global): New function.
20398 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
20399 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
20401 (aarch64_evpc_sve_tbl): New function.
20402 (aarch64_expand_vec_perm_const_1): Update after rename of
20403 aarch64_evpc_rev. Handle SVE permutes too, trying
20404 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
20405 than aarch64_evpc_tbl.
20406 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
20407 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
20408 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
20409 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
20410 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
20411 (aarch64_expand_sve_vcond): New functions.
20412 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
20413 of aarch64_vector_mode_p.
20414 (aarch64_dwarf_poly_indeterminate_value): New function.
20415 (aarch64_compute_pressure_classes): Likewise.
20416 (aarch64_can_change_mode_class): Likewise.
20417 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
20418 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
20419 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
20420 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
20421 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
20422 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
20423 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
20424 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
20426 (Dn, Dl, Dr): Accept const as well as const_vector.
20427 (Dz): Likewise. Compare against CONST0_RTX.
20428 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
20429 of "vector" where appropriate.
20430 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
20431 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
20432 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
20433 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
20434 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
20435 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
20436 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
20437 (v_int_equiv): Extend to SVE modes.
20438 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
20440 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
20441 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
20442 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
20443 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
20444 (SVE_COND_FP_CMP): New int iterators.
20445 (perm_hilo): Handle the new unpack unspecs.
20446 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
20448 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
20449 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
20450 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
20451 (aarch64_equality_operator, aarch64_constant_vector_operand)
20452 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
20453 (aarch64_sve_nonimmediate_operand): Likewise.
20454 (aarch64_sve_general_operand): Likewise.
20455 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
20456 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
20457 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
20458 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
20459 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
20460 (aarch64_sve_float_arith_immediate): Likewise.
20461 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
20462 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
20463 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
20464 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
20465 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
20466 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
20467 (aarch64_sve_float_arith_operand): Likewise.
20468 (aarch64_sve_float_arith_with_sub_operand): Likewise.
20469 (aarch64_sve_float_mul_operand): Likewise.
20470 (aarch64_sve_vec_perm_operand): Likewise.
20471 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
20472 (aarch64_mov_operand): Accept const_poly_int and const_vector.
20473 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
20474 as well as const_vector.
20475 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
20476 in file. Use CONST0_RTX and CONSTM1_RTX.
20477 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
20478 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
20479 Use aarch64_simd_imm_zero.
20480 * config/aarch64/aarch64-sve.md: New file.
20481 * config/aarch64/aarch64.md: Include it.
20482 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
20483 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
20484 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
20485 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
20486 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
20487 (sve): New attribute.
20488 (enabled): Disable instructions with the sve attribute unless
20490 (movqi, movhi): Pass CONST_POLY_INT operaneds through
20491 aarch64_expand_mov_immediate.
20492 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
20493 CNT[BHSD] immediates.
20494 (movti): Split CONST_POLY_INT moves into two halves.
20495 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
20496 Split additions that need a temporary here if the destination
20497 is the stack pointer.
20498 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
20499 (*add<mode>3_poly_1): New instruction.
20500 (set_clobber_cc): New expander.
20502 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20504 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
20505 parameter and use it instead of GET_MODE_SIZE (innermode). Use
20506 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
20507 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
20508 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
20509 Change innermode from fixed_mode_size to machine_mode.
20510 (simplify_subreg): Update call accordingly. Handle a constant-sized
20511 subreg of a variable-length CONST_VECTOR.
20513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
20514 Alan Hayward <alan.hayward@arm.com>
20515 David Sherwood <david.sherwood@arm.com>
20517 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
20518 (add_offset_to_base): New function, split out from...
20519 (create_mem_ref): ...here. When handling a scale other than 1,
20520 check first whether the address is valid without the offset.
20521 Add it into the base if so, leaving the index and scale as-is.
20523 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20526 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
20527 fold_for_warn before checking if arg2 is INTEGER_CST.
20529 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
20531 * config/rs6000/predicates.md (load_multiple_operation): Delete.
20532 (store_multiple_operation): Delete.
20533 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
20534 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
20535 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
20536 guarded by TARGET_STRING.
20537 (rs6000_output_load_multiple): Delete.
20538 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
20539 OPTION_MASK_STRING / TARGET_STRING handling.
20540 (print_operand) <'N', 'O'>: Add comment that these are unused now.
20541 (const rs6000_opt_masks) <"string">: Change mask to 0.
20542 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
20543 (MASK_STRING): Delete.
20544 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
20546 (load_multiple): Delete.
20553 (store_multiple): Delete.
20560 (movmemsi_8reg): Delete.
20561 (corresponding unnamed define_insn): Delete.
20562 (movmemsi_6reg): Delete.
20563 (corresponding unnamed define_insn): Delete.
20564 (movmemsi_4reg): Delete.
20565 (corresponding unnamed define_insn): Delete.
20566 (movmemsi_2reg): Delete.
20567 (corresponding unnamed define_insn): Delete.
20568 (movmemsi_1reg): Delete.
20569 (corresponding unnamed define_insn): Delete.
20570 * config/rs6000/rs6000.opt (mno-string): New.
20571 (mstring): Replace by deprecation warning stub.
20572 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
20574 2018-01-12 Jakub Jelinek <jakub@redhat.com>
20576 * regrename.c (regrename_do_replace): If replacing the same
20577 reg multiple times, try to reuse last created gen_raw_REG.
20580 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
20581 main to workaround a bug in GDB.
20583 2018-01-12 Tom de Vries <tom@codesourcery.com>
20586 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
20588 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
20590 PR rtl-optimization/80481
20591 * ira-color.c (get_cap_member): New function.
20592 (allocnos_conflict_by_live_ranges_p): Use it.
20593 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
20594 (setup_slot_coalesced_allocno_live_ranges): Ditto.
20596 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
20599 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
20600 (*saddl_se_1): Ditto.
20601 (*ssubsi_1): Ditto.
20602 (*ssubl_se_1): Ditto.
20604 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20606 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
20607 rather than wi::to_widest for DR_INITs.
20608 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
20609 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
20610 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
20612 (vect_analyze_group_access_1): Note that here.
20614 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20616 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
20617 polynomial type sizes.
20619 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
20621 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
20622 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
20623 (gimple_add_tmp_var): Likewise.
20625 2018-01-12 Martin Liska <mliska@suse.cz>
20627 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
20628 (gimple_alloc_sizes): Likewise.
20629 (dump_gimple_statistics): Use PRIu64 in printf format.
20630 * gimple.h: Change uint64_t to int.
20632 2018-01-12 Martin Liska <mliska@suse.cz>
20634 * tree-core.h: Use uint64_t instead of int.
20635 * tree.c (tree_node_counts): Likewise.
20636 (tree_node_sizes): Likewise.
20637 (dump_tree_statistics): Use PRIu64 in printf format.
20639 2018-01-12 Martin Liska <mliska@suse.cz>
20641 * Makefile.in: As qsort_chk is implemented in vec.c, add
20642 vec.o to linkage of gencfn-macros.
20643 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
20644 passing the info to record_node_allocation_statistics.
20645 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
20647 * ggc-common.c (struct ggc_usage): Add operator== and use
20648 it in operator< and compare function.
20649 * mem-stats.h (struct mem_usage): Likewise.
20650 * vec.c (struct vec_usage): Remove operator< and compare
20651 function. Can be simply inherited.
20653 2018-01-12 Martin Jambor <mjambor@suse.cz>
20656 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
20657 * tree-ssa-math-opts.c: Include domwalk.h.
20658 (convert_mult_to_fma_1): New function.
20659 (fma_transformation_info): New type.
20660 (fma_deferring_state): Likewise.
20661 (cancel_fma_deferring): New function.
20662 (result_of_phi): Likewise.
20663 (last_fma_candidate_feeds_initial_phi): Likewise.
20664 (convert_mult_to_fma): Added deferring logic, split actual
20665 transformation to convert_mult_to_fma_1.
20666 (math_opts_dom_walker): New type.
20667 (math_opts_dom_walker::after_dom_children): New method, body moved
20668 here from pass_optimize_widening_mul::execute, added deferring logic
20670 (pass_optimize_widening_mul::execute): Moved most of code to
20671 math_opts_dom_walker::after_dom_children.
20672 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
20673 * config/i386/i386.c (ix86_option_override_internal): Added
20674 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
20676 2018-01-12 Richard Biener <rguenther@suse.de>
20679 * dwarf2out.c (gen_variable_die): Do not reset old_die for
20680 inline instance vars.
20682 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
20685 * config/rx/rx.c (rx_is_restricted_memory_address):
20686 Handle SUBREG case.
20688 2018-01-12 Richard Biener <rguenther@suse.de>
20690 PR tree-optimization/80846
20691 * target.def (split_reduction): New target hook.
20692 * targhooks.c (default_split_reduction): New function.
20693 * targhooks.h (default_split_reduction): Declare.
20694 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
20695 target requests first reduce vectors by combining low and high
20697 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
20698 (get_vectype_for_scalar_type_and_size): Export.
20699 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
20700 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
20701 * doc/tm.texi: Regenerate.
20702 * config/i386/i386.c (ix86_split_reduction): Implement
20703 TARGET_VECTORIZE_SPLIT_REDUCTION.
20705 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20708 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
20709 in PIC mode except for TARGET_VXWORKS_RTP.
20710 * config/sparc/sparc.c: Include cfgrtl.h.
20711 (TARGET_INIT_PIC_REG): Define.
20712 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
20713 (sparc_pic_register_p): New predicate.
20714 (sparc_legitimate_address_p): Use it.
20715 (sparc_legitimize_pic_address): Likewise.
20716 (sparc_delegitimize_address): Likewise.
20717 (sparc_mode_dependent_address_p): Likewise.
20718 (gen_load_pcrel_sym): Remove 4th parameter.
20719 (load_got_register): Adjust call to above. Remove obsolete stuff.
20720 (sparc_expand_prologue): Do not call load_got_register here.
20721 (sparc_flat_expand_prologue): Likewise.
20722 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
20723 (sparc_use_pseudo_pic_reg): New function.
20724 (sparc_init_pic_reg): Likewise.
20725 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20726 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
20728 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
20730 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
20731 Add item for branch_cost.
20733 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
20735 PR rtl-optimization/83565
20736 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
20737 not extend the result to a larger mode for rotate operations.
20738 (num_sign_bit_copies1): Likewise.
20740 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20743 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
20745 Use values-Xc.o for -pedantic.
20746 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
20748 2018-01-12 Martin Liska <mliska@suse.cz>
20751 * ipa-devirt.c (final_warning_record::grow_type_warnings):
20753 (possible_polymorphic_call_targets): Use it.
20754 (ipa_devirt): Likewise.
20756 2018-01-12 Martin Liska <mliska@suse.cz>
20758 * profile-count.h (enum profile_quality): Use 0 as invalid
20759 enum value of profile_quality.
20761 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
20763 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
20764 -mext-string options.
20766 2018-01-12 Richard Biener <rguenther@suse.de>
20768 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
20769 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
20770 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
20772 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
20774 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
20776 * configure.ac (--with-long-double-format): Add support for the
20777 configuration option to change the default long double format on
20779 * config.gcc (powerpc*-linux*-*): Likewise.
20780 * configure: Regenerate.
20781 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
20782 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
20783 used without modification.
20785 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
20787 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
20788 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
20789 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
20790 MISC_BUILTIN_SPEC_BARRIER.
20791 (rs6000_init_builtins): Likewise.
20792 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
20794 (speculation_barrier): New define_insn.
20795 * doc/extend.texi: Document __builtin_speculation_barrier.
20797 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20800 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
20801 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
20802 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
20804 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
20805 integral modes instead of "ss" and "sd".
20806 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
20807 vectors with 32-bit and 64-bit elements.
20808 (vecdupssescalarmodesuffix): New mode attribute.
20809 (vec_dup<mode>): Use it.
20811 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
20814 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
20815 frame if argument is passed on stack.
20817 2018-01-11 Jakub Jelinek <jakub@redhat.com>
20820 * ree.c (combine_reaching_defs): Optimize also
20821 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
20822 reg2=any_extend(exp); reg1=reg2;, formatting fix.
20824 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20826 PR middle-end/83189
20827 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
20829 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
20831 PR middle-end/83718
20832 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
20833 after they are computed.
20835 2018-01-11 Bin Cheng <bin.cheng@arm.com>
20837 PR tree-optimization/83695
20838 * gimple-loop-linterchange.cc
20839 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
20840 reset cached scev information after interchange.
20841 (pass_linterchange::execute): Remove call to scev_reset_htab.
20843 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20845 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
20846 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
20847 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
20848 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
20849 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
20850 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
20851 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
20852 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
20853 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
20854 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
20855 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
20856 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
20857 (V_lane_reg): Likewise.
20858 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
20860 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
20861 (vfmal_lane_low<mode>_intrinsic,
20862 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
20863 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
20864 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
20865 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
20866 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
20867 vfmsl_lane_high<mode>_intrinsic): New define_insns.
20869 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20871 * config/arm/arm-cpus.in (fp16fml): New feature.
20872 (ALL_SIMD): Add fp16fml.
20873 (armv8.2-a): Add fp16fml as an option.
20874 (armv8.3-a): Likewise.
20875 (armv8.4-a): Add fp16fml as part of fp16.
20876 * config/arm/arm.h (TARGET_FP16FML): Define.
20877 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
20879 * config/arm/arm-modes.def (V2HF): Define.
20880 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
20881 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
20882 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
20883 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
20884 vfmsl_low, vfmsl_high): New set of builtins.
20885 * config/arm/iterators.md (PLUSMINUS): New code iterator.
20886 (vfml_op): New code attribute.
20887 (VFMLHALVES): New int iterator.
20888 (VFML, VFMLSEL): New mode attributes.
20889 (V_reg): Define mapping for V2HF.
20890 (V_hi, V_lo): New mode attributes.
20891 (VF_constraint): Likewise.
20892 (vfml_half, vfml_half_selector): New int attributes.
20893 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
20895 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
20896 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
20898 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
20899 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
20900 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
20901 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
20903 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
20904 Document new effective target and option set.
20906 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20908 * config/arm/arm-cpus.in (armv8_4): New feature.
20909 (ARMv8_4a): New fgroup.
20910 (armv8.4-a): New arch.
20911 * config/arm/arm-tables.opt: Regenerate.
20912 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
20913 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
20914 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
20915 Add matching rules for -march=armv8.4-a and extensions.
20916 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
20918 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
20921 * config/rx/rx.md (BW): New mode attribute.
20922 (sync_lock_test_and_setsi): Add mode suffix to insn output.
20924 2018-01-11 Richard Biener <rguenther@suse.de>
20926 PR tree-optimization/83435
20927 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
20928 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
20929 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
20931 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20932 Alan Hayward <alan.hayward@arm.com>
20933 David Sherwood <david.sherwood@arm.com>
20935 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
20937 (aarch64_classify_address): Initialize it. Track polynomial offsets.
20938 (aarch64_print_address_internal): Use it to check for a zero offset.
20940 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
20941 Alan Hayward <alan.hayward@arm.com>
20942 David Sherwood <david.sherwood@arm.com>
20944 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
20945 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
20946 Return a poly_int64 rather than a HOST_WIDE_INT.
20947 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
20948 rather than a HOST_WIDE_INT.
20949 * config/aarch64/aarch64.h (aarch64_frame): Protect with
20950 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
20951 hard_fp_offset, frame_size, initial_adjust, callee_offset and
20952 final_offset from HOST_WIDE_INT to poly_int64.
20953 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
20954 to_constant when getting the number of units in an Advanced SIMD
20956 (aarch64_builtin_vectorized_function): Check for a constant number
20958 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
20960 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
20961 attribute instead of GET_MODE_NUNITS.
20962 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
20963 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
20964 GET_MODE_SIZE for fixed-size registers.
20965 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
20966 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
20967 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
20968 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
20969 (aarch64_print_operand, aarch64_print_address_internal)
20970 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
20971 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
20972 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
20973 Handle polynomial GET_MODE_SIZE.
20974 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
20975 wider than SImode without modification.
20976 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
20977 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
20978 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
20979 passing and returning SVE modes.
20980 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
20981 rather than GEN_INT.
20982 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
20983 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
20984 (aarch64_allocate_and_probe_stack_space): Likewise.
20985 (aarch64_layout_frame): Cope with polynomial offsets.
20986 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
20987 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
20988 polynomial offsets.
20989 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
20990 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
20991 poly_int64 rather than a HOST_WIDE_INT.
20992 (aarch64_get_separate_components, aarch64_process_components)
20993 (aarch64_expand_prologue, aarch64_expand_epilogue)
20994 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
20995 (aarch64_anchor_offset): New function, split out from...
20996 (aarch64_legitimize_address): ...here.
20997 (aarch64_builtin_vectorization_cost): Handle polynomial
20998 TYPE_VECTOR_SUBPARTS.
20999 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
21001 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
21002 number of elements from the PARALLEL rather than the mode.
21003 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
21004 rather than GET_MODE_BITSIZE.
21005 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
21006 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
21007 (aarch64_expand_vec_perm_const_1): Handle polynomial
21008 d->perm.length () and d->perm elements.
21009 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
21010 Apply to_constant to d->perm elements.
21011 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
21012 polynomial CONST_VECTOR_NUNITS.
21013 (aarch64_move_pointer): Take amount as a poly_int64 rather
21015 (aarch64_progress_pointer): Avoid temporary variable.
21016 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
21017 the mode attribute instead of GET_MODE.
21019 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21020 Alan Hayward <alan.hayward@arm.com>
21021 David Sherwood <david.sherwood@arm.com>
21023 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
21024 x exists before using it.
21025 (aarch64_add_constant_internal): Rename to...
21026 (aarch64_add_offset_1): ...this. Replace regnum with separate
21027 src and dest rtxes. Handle the case in which they're different,
21028 including when the offset is zero. Replace scratchreg with an rtx.
21029 Use 2 additions if there is no spare register into which we can
21030 move a 16-bit constant.
21031 (aarch64_add_constant): Delete.
21032 (aarch64_add_offset): Replace reg with separate src and dest
21033 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
21034 Use aarch64_add_offset_1.
21035 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
21036 an rtx rather than an int. Take the delta as a poly_int64
21037 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
21038 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
21039 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
21040 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
21041 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
21042 and aarch64_add_sp.
21043 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
21044 aarch64_add_constant.
21046 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21048 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
21049 Use scalar_float_mode.
21051 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21053 * config/aarch64/aarch64-simd.md
21054 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
21055 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
21056 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
21057 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
21058 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
21059 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
21060 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
21061 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
21062 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
21063 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
21065 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21068 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
21069 targ_options->x_arm_arch_string is non NULL.
21071 2018-01-11 Tamar Christina <tamar.christina@arm.com>
21073 * config/aarch64/aarch64.h
21074 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
21076 2018-01-11 Sudakshina Das <sudi.das@arm.com>
21079 * expmed.c (emit_store_flag_force): Swap if const op0
21080 and change VOIDmode to mode of op0.
21082 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
21084 PR rtl-optimization/83761
21085 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
21086 than bytes to mode_for_size.
21088 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21090 PR middle-end/83189
21091 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
21092 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
21095 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21097 PR middle-end/83575
21098 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
21099 when in layout mode.
21100 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
21101 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
21104 2018-01-10 Michael Collison <michael.collison@arm.com>
21106 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
21107 * config/aarch64/aarch64-option-extension.def: Add
21108 AARCH64_OPT_EXTENSION of 'fp16fml'.
21109 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21110 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
21111 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
21112 * config/aarch64/constraints.md (Ui7): New constraint.
21113 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
21114 (VFMLA_SEL_W): Ditto.
21117 (VFMLA16_LOW): New int iterator.
21118 (VFMLA16_HIGH): Ditto.
21119 (UNSPEC_FMLAL): New unspec.
21120 (UNSPEC_FMLSL): Ditto.
21121 (UNSPEC_FMLAL2): Ditto.
21122 (UNSPEC_FMLSL2): Ditto.
21123 (f16mac): New code attribute.
21124 * config/aarch64/aarch64-simd-builtins.def
21125 (aarch64_fmlal_lowv2sf): Ditto.
21126 (aarch64_fmlsl_lowv2sf): Ditto.
21127 (aarch64_fmlalq_lowv4sf): Ditto.
21128 (aarch64_fmlslq_lowv4sf): Ditto.
21129 (aarch64_fmlal_highv2sf): Ditto.
21130 (aarch64_fmlsl_highv2sf): Ditto.
21131 (aarch64_fmlalq_highv4sf): Ditto.
21132 (aarch64_fmlslq_highv4sf): Ditto.
21133 (aarch64_fmlal_lane_lowv2sf): Ditto.
21134 (aarch64_fmlsl_lane_lowv2sf): Ditto.
21135 (aarch64_fmlal_laneq_lowv2sf): Ditto.
21136 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
21137 (aarch64_fmlalq_lane_lowv4sf): Ditto.
21138 (aarch64_fmlsl_lane_lowv4sf): Ditto.
21139 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
21140 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
21141 (aarch64_fmlal_lane_highv2sf): Ditto.
21142 (aarch64_fmlsl_lane_highv2sf): Ditto.
21143 (aarch64_fmlal_laneq_highv2sf): Ditto.
21144 (aarch64_fmlsl_laneq_highv2sf): Ditto.
21145 (aarch64_fmlalq_lane_highv4sf): Ditto.
21146 (aarch64_fmlsl_lane_highv4sf): Ditto.
21147 (aarch64_fmlalq_laneq_highv4sf): Ditto.
21148 (aarch64_fmlsl_laneq_highv4sf): Ditto.
21149 * config/aarch64/aarch64-simd.md:
21150 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
21151 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
21152 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
21153 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
21154 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
21155 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
21156 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
21157 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
21158 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
21159 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
21160 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
21161 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
21162 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
21163 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
21164 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
21165 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
21166 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
21167 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
21168 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
21169 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
21170 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
21171 (vfmlsl_low_u32): Ditto.
21172 (vfmlalq_low_u32): Ditto.
21173 (vfmlslq_low_u32): Ditto.
21174 (vfmlal_high_u32): Ditto.
21175 (vfmlsl_high_u32): Ditto.
21176 (vfmlalq_high_u32): Ditto.
21177 (vfmlslq_high_u32): Ditto.
21178 (vfmlal_lane_low_u32): Ditto.
21179 (vfmlsl_lane_low_u32): Ditto.
21180 (vfmlal_laneq_low_u32): Ditto.
21181 (vfmlsl_laneq_low_u32): Ditto.
21182 (vfmlalq_lane_low_u32): Ditto.
21183 (vfmlslq_lane_low_u32): Ditto.
21184 (vfmlalq_laneq_low_u32): Ditto.
21185 (vfmlslq_laneq_low_u32): Ditto.
21186 (vfmlal_lane_high_u32): Ditto.
21187 (vfmlsl_lane_high_u32): Ditto.
21188 (vfmlal_laneq_high_u32): Ditto.
21189 (vfmlsl_laneq_high_u32): Ditto.
21190 (vfmlalq_lane_high_u32): Ditto.
21191 (vfmlslq_lane_high_u32): Ditto.
21192 (vfmlalq_laneq_high_u32): Ditto.
21193 (vfmlslq_laneq_high_u32): Ditto.
21194 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
21195 (AARCH64_FL_FOR_ARCH8_4): New.
21196 (AARCH64_ISA_F16FML): New ISA flag.
21197 (TARGET_F16FML): New feature flag for fp16fml.
21198 (doc/invoke.texi): Document new fp16fml option.
21200 2018-01-10 Michael Collison <michael.collison@arm.com>
21202 * config/aarch64/aarch64-builtins.c:
21203 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
21204 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21205 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
21206 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
21207 (AARCH64_ISA_SHA3): New ISA flag.
21208 (TARGET_SHA3): New feature flag for sha3.
21209 * config/aarch64/iterators.md (sha512_op): New int attribute.
21210 (CRYPTO_SHA512): New int iterator.
21211 (UNSPEC_SHA512H): New unspec.
21212 (UNSPEC_SHA512H2): Ditto.
21213 (UNSPEC_SHA512SU0): Ditto.
21214 (UNSPEC_SHA512SU1): Ditto.
21215 * config/aarch64/aarch64-simd-builtins.def
21216 (aarch64_crypto_sha512hqv2di): New builtin.
21217 (aarch64_crypto_sha512h2qv2di): Ditto.
21218 (aarch64_crypto_sha512su0qv2di): Ditto.
21219 (aarch64_crypto_sha512su1qv2di): Ditto.
21220 (aarch64_eor3qv8hi): Ditto.
21221 (aarch64_rax1qv2di): Ditto.
21222 (aarch64_xarqv2di): Ditto.
21223 (aarch64_bcaxqv8hi): Ditto.
21224 * config/aarch64/aarch64-simd.md:
21225 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
21226 (aarch64_crypto_sha512su0qv2di): Ditto.
21227 (aarch64_crypto_sha512su1qv2di): Ditto.
21228 (aarch64_eor3qv8hi): Ditto.
21229 (aarch64_rax1qv2di): Ditto.
21230 (aarch64_xarqv2di): Ditto.
21231 (aarch64_bcaxqv8hi): Ditto.
21232 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
21233 (vsha512h2q_u64): Ditto.
21234 (vsha512su0q_u64): Ditto.
21235 (vsha512su1q_u64): Ditto.
21236 (veor3q_u16): Ditto.
21237 (vrax1q_u64): Ditto.
21238 (vxarq_u64): Ditto.
21239 (vbcaxq_u16): Ditto.
21240 * config/arm/types.md (crypto_sha512): New type attribute.
21241 (crypto_sha3): Ditto.
21242 (doc/invoke.texi): Document new sha3 option.
21244 2018-01-10 Michael Collison <michael.collison@arm.com>
21246 * config/aarch64/aarch64-builtins.c:
21247 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
21248 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21249 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
21250 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
21251 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
21252 (AARCH64_ISA_SM4): New ISA flag.
21253 (TARGET_SM4): New feature flag for sm4.
21254 * config/aarch64/aarch64-simd-builtins.def
21255 (aarch64_sm3ss1qv4si): Ditto.
21256 (aarch64_sm3tt1aq4si): Ditto.
21257 (aarch64_sm3tt1bq4si): Ditto.
21258 (aarch64_sm3tt2aq4si): Ditto.
21259 (aarch64_sm3tt2bq4si): Ditto.
21260 (aarch64_sm3partw1qv4si): Ditto.
21261 (aarch64_sm3partw2qv4si): Ditto.
21262 (aarch64_sm4eqv4si): Ditto.
21263 (aarch64_sm4ekeyqv4si): Ditto.
21264 * config/aarch64/aarch64-simd.md:
21265 (aarch64_sm3ss1qv4si): Ditto.
21266 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
21267 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
21268 (aarch64_sm4eqv4si): Ditto.
21269 (aarch64_sm4ekeyqv4si): Ditto.
21270 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
21271 (sm3part_op): Ditto.
21272 (CRYPTO_SM3TT): Ditto.
21273 (CRYPTO_SM3PART): Ditto.
21274 (UNSPEC_SM3SS1): New unspec.
21275 (UNSPEC_SM3TT1A): Ditto.
21276 (UNSPEC_SM3TT1B): Ditto.
21277 (UNSPEC_SM3TT2A): Ditto.
21278 (UNSPEC_SM3TT2B): Ditto.
21279 (UNSPEC_SM3PARTW1): Ditto.
21280 (UNSPEC_SM3PARTW2): Ditto.
21281 (UNSPEC_SM4E): Ditto.
21282 (UNSPEC_SM4EKEY): Ditto.
21283 * config/aarch64/constraints.md (Ui2): New constraint.
21284 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
21285 * config/arm/types.md (crypto_sm3): New type attribute.
21286 (crypto_sm4): Ditto.
21287 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
21288 (vsm3tt1aq_u32): Ditto.
21289 (vsm3tt1bq_u32): Ditto.
21290 (vsm3tt2aq_u32): Ditto.
21291 (vsm3tt2bq_u32): Ditto.
21292 (vsm3partw1q_u32): Ditto.
21293 (vsm3partw2q_u32): Ditto.
21294 (vsm4eq_u32): Ditto.
21295 (vsm4ekeyq_u32): Ditto.
21296 (doc/invoke.texi): Document new sm4 option.
21298 2018-01-10 Michael Collison <michael.collison@arm.com>
21300 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
21301 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
21302 (AARCH64_FL_FOR_ARCH8_4): New.
21303 (AARCH64_FL_V8_4): New flag.
21304 (doc/invoke.texi): Document new armv8.4-a option.
21306 2018-01-10 Michael Collison <michael.collison@arm.com>
21308 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
21309 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
21310 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
21311 * config/aarch64/aarch64-option-extension.def: Add
21312 AARCH64_OPT_EXTENSION of 'sha2'.
21313 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
21314 (crypto): Disable sha2 and aes if crypto disabled.
21315 (crypto): Enable aes and sha2 if enabled.
21316 (simd): Disable sha2 and aes if simd disabled.
21317 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
21319 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
21320 (TARGET_SHA2): New feature flag for sha2.
21321 (TARGET_AES): New feature flag for aes.
21322 * config/aarch64/aarch64-simd.md:
21323 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
21324 conditional on TARGET_AES.
21325 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
21326 (aarch64_crypto_sha1hsi): Make pattern conditional
21328 (aarch64_crypto_sha1hv4si): Ditto.
21329 (aarch64_be_crypto_sha1hv4si): Ditto.
21330 (aarch64_crypto_sha1su1v4si): Ditto.
21331 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
21332 (aarch64_crypto_sha1su0v4si): Ditto.
21333 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
21334 (aarch64_crypto_sha256su0v4si): Ditto.
21335 (aarch64_crypto_sha256su1v4si): Ditto.
21336 (doc/invoke.texi): Document new aes and sha2 options.
21338 2018-01-10 Martin Sebor <msebor@redhat.com>
21340 PR tree-optimization/83781
21341 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
21344 2018-01-11 Martin Sebor <msebor@gmail.com>
21345 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21347 PR tree-optimization/83501
21348 PR tree-optimization/81703
21350 * tree-ssa-strlen.c (get_string_cst): Rename...
21351 (get_string_len): ...to this. Handle global constants.
21352 (handle_char_store): Adjust.
21354 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
21355 Jim Wilson <jimw@sifive.com>
21357 * config/riscv/riscv-protos.h (riscv_output_return): New.
21358 * config/riscv/riscv.c (struct machine_function): New naked_p field.
21359 (riscv_attribute_table, riscv_output_return),
21360 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
21361 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
21362 (riscv_compute_frame_info): Only compute frame->mask if not a naked
21364 (riscv_expand_prologue): Add early return for naked function.
21365 (riscv_expand_epilogue): Likewise.
21366 (riscv_function_ok_for_sibcall): Return false for naked function.
21367 (riscv_set_current_function): New.
21368 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
21369 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
21370 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
21371 * doc/extend.texi (RISC-V Function Attributes): New.
21373 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
21375 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
21376 check for 128-bit long double before checking TCmode.
21377 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
21378 128-bit long doubles before checking TFmode or TCmode.
21379 (FLOAT128_IBM_P): Likewise.
21381 2018-01-10 Martin Sebor <msebor@redhat.com>
21383 PR tree-optimization/83671
21384 * builtins.c (c_strlen): Unconditionally return zero for the empty
21386 Use -Warray-bounds for warnings.
21387 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
21388 for non-constant array indices with COMPONENT_REF, arrays of
21389 arrays, and pointers to arrays.
21390 (gimple_fold_builtin_strlen): Determine and set length range for
21391 non-constant character arrays.
21393 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
21395 PR middle-end/81897
21396 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
21399 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
21401 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
21403 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
21406 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
21407 VECTOR_MEM_ALTIVEC_OR_VSX_P.
21408 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
21409 indexed_or_indirect_operand predicate.
21410 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
21411 (*vsx_le_perm_load_v8hi): Likewise.
21412 (*vsx_le_perm_load_v16qi): Likewise.
21413 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
21414 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
21415 (*vsx_le_perm_store_v8hi): Likewise.
21416 (*vsx_le_perm_store_v16qi): Likewise.
21417 (eight unnamed splitters): Likewise.
21419 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
21421 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
21422 * config/rs6000/emmintrin.h: Likewise.
21423 * config/rs6000/mmintrin.h: Likewise.
21424 * config/rs6000/xmmintrin.h: Likewise.
21426 2018-01-10 David Malcolm <dmalcolm@redhat.com>
21429 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
21431 * tree.c (tree_nop_conversion): Return true for location wrapper
21433 (maybe_wrap_with_location): New function.
21434 (selftest::check_strip_nops): New function.
21435 (selftest::test_location_wrappers): New function.
21436 (selftest::tree_c_tests): Call it.
21437 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
21438 (maybe_wrap_with_location): New decl.
21439 (EXPR_LOCATION_WRAPPER_P): New macro.
21440 (location_wrapper_p): New inline function.
21441 (tree_strip_any_location_wrapper): New inline function.
21443 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
21446 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
21447 stack_realign_offset for the largest alignment of stack slot
21449 (ix86_find_max_used_stack_alignment): New function.
21450 (ix86_finalize_stack_frame_flags): Use it. Set
21451 max_used_stack_alignment if we don't realign stack.
21452 * config/i386/i386.h (machine_function): Add
21453 max_used_stack_alignment.
21455 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
21457 * config/arm/arm.opt (-mbranch-cost): New option.
21458 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
21461 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
21464 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
21465 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
21467 2018-01-10 Richard Biener <rguenther@suse.de>
21470 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
21471 early out so it also covers the case where we have a non-NULL
21474 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
21476 PR tree-optimization/83753
21477 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
21478 for non-strided grouped accesses if the number of elements is 1.
21480 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
21483 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
21484 * i386.h (TARGET_USE_GATHER): Define.
21485 * x86-tune.def (X86_TUNE_USE_GATHER): New.
21487 2018-01-10 Martin Liska <mliska@suse.cz>
21490 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
21491 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
21493 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
21494 CLEANUP_NO_PARTITIONING is not set.
21496 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
21498 * doc/rtl.texi: Remove documentation of (const ...) wrappers
21499 for vectors, as a partial revert of r254296.
21500 * rtl.h (const_vec_p): Delete.
21501 (const_vec_duplicate_p): Don't test for vector CONSTs.
21502 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
21503 * expmed.c (make_tree): Likewise.
21506 * common.md (E, F): Use CONSTANT_P instead of checking for
21508 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
21509 checking for CONST_VECTOR.
21511 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21513 PR middle-end/83575
21514 * predict.c (force_edge_cold): Handle in more sane way edges
21515 with no prediction.
21517 2018-01-09 Carl Love <cel@us.ibm.com>
21519 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
21521 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
21522 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
21523 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
21524 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
21525 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
21526 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
21527 * config/rs6000/rs6000-protos.h: Add extern defition for
21528 rs6000_generate_float2_double_code.
21529 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
21531 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
21532 (float2_v2df): Add define_expand.
21534 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
21537 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
21538 op_mode in the force_to_mode call.
21540 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21542 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
21543 instead of checking each element individually.
21544 (aarch64_evpc_uzp): Likewise.
21545 (aarch64_evpc_zip): Likewise.
21546 (aarch64_evpc_ext): Likewise.
21547 (aarch64_evpc_rev): Likewise.
21548 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
21549 instead of checking each element individually. Return true without
21551 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
21552 whether all selected elements come from the same input, instead of
21553 checking each element individually. Remove calls to gen_rtx_REG,
21554 start_sequence and end_sequence and instead assert that no rtl is
21557 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21559 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
21560 order of HIGH and CONST checks.
21562 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
21564 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
21565 if the destination isn't an SSA_NAME.
21567 2018-01-09 Richard Biener <rguenther@suse.de>
21569 PR tree-optimization/83668
21570 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
21572 (canonicalize_loop_form): ... here, renamed from ...
21573 (canonicalize_loop_closed_ssa_form): ... this and amended to
21574 swap successor edges for loop exit blocks to make us use
21575 the RPO order we need for initial schedule generation.
21577 2018-01-09 Joseph Myers <joseph@codesourcery.com>
21579 PR tree-optimization/64811
21580 * match.pd: When optimizing comparisons with Inf, avoid
21581 introducing or losing exceptions from comparisons with NaN.
21583 2018-01-09 Martin Liska <mliska@suse.cz>
21586 * asan.c (shadow_mem_size): Add gcc_assert.
21588 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
21590 Don't save registers in main().
21593 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
21594 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
21595 * config/avr/avr.c (avr_set_current_function): Don't error if
21596 naked, OS_task or OS_main are specified at the same time.
21597 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
21599 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
21601 * common/config/avr/avr-common.c (avr_option_optimization_table):
21602 Switch on -mmain-is-OS_task for optimizing compilations.
21604 2018-01-09 Richard Biener <rguenther@suse.de>
21606 PR tree-optimization/83572
21607 * graphite.c: Include cfganal.h.
21608 (graphite_transform_loops): Connect infinite loops to exit
21609 and remove fake edges at the end.
21611 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21613 * ipa-inline.c (edge_badness): Revert accidental checkin.
21615 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
21618 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
21619 symbols; not inline clones.
21621 2018-01-09 Jakub Jelinek <jakub@redhat.com>
21624 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
21625 hard registers. Formatting fixes.
21627 PR preprocessor/83722
21628 * gcc.c (try_generate_repro): Pass
21629 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
21630 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
21633 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
21634 Kito Cheng <kito.cheng@gmail.com>
21636 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
21637 (riscv_leaf_function_p): Delete.
21638 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
21640 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21642 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
21644 (do_ifelse): New function.
21645 (do_isel): New function.
21646 (do_sub3): New function.
21647 (do_add3): New function.
21648 (do_load_mask_compare): New function.
21649 (do_overlap_load_compare): New function.
21650 (expand_compare_loop): New function.
21651 (expand_block_compare): Call expand_compare_loop() when appropriate.
21652 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
21653 option description.
21654 (-mblock-compare-inline-loop-limit): New option.
21656 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21659 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
21660 Reverse order of second and third operands in first alternative.
21661 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
21662 of first and second elements in UNSPEC_VPERMR vector.
21663 (altivec_expand_vec_perm_le): Likewise.
21665 2018-01-08 Jeff Law <law@redhat.com>
21667 PR rtl-optimizatin/81308
21668 * tree-switch-conversion.c (cfg_altered): New file scoped static.
21669 (process_switch): If group_case_labels makes a change, then set
21671 (pass_convert_switch::execute): If a switch is converted, then
21672 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
21674 PR rtl-optimization/81308
21675 * recog.c (split_all_insns): Conditionally cleanup the CFG after
21678 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
21680 PR target/83663 - Revert r255946
21681 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
21682 generation for cases where splatting a value is not useful.
21683 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
21684 across a vec_duplicate and a paradoxical subreg forming a vector
21685 mode to a vec_concat.
21687 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21689 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
21690 -march=armv8.3-a variants.
21691 * config/arm/t-multilib: Likewise.
21692 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
21694 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21696 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
21698 (cceq_ior_compare_complement): Give it a name so I can use it, and
21699 change boolean_or_operator predicate to boolean_operator so it can
21700 be used to generate a crand.
21701 (eqne): New code iterator.
21702 (bd/bd_neg): New code_attrs.
21703 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
21704 a single define_insn.
21705 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
21706 decrement (bdnzt/bdnzf/bdzt/bdzf).
21707 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
21708 with the new names of the branch decrement patterns, and added the
21709 names of the branch decrement conditional patterns.
21711 2018-01-08 Richard Biener <rguenther@suse.de>
21713 PR tree-optimization/83563
21714 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
21717 2018-01-08 Richard Biener <rguenther@suse.de>
21719 PR middle-end/83713
21720 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
21722 2018-01-08 Richard Biener <rguenther@suse.de>
21724 PR tree-optimization/83685
21725 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
21726 references to abnormals.
21728 2018-01-08 Richard Biener <rguenther@suse.de>
21731 * dwarf2out.c (output_indirect_strings): Handle empty
21732 skeleton_debug_str_hash.
21733 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
21735 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21737 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
21738 (emit_store_direct): Likewise.
21739 (arc_trampoline_adjust_address): Likewise.
21740 (arc_asm_trampoline_template): New function.
21741 (arc_initialize_trampoline): Use asm_trampoline_template.
21742 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
21743 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
21744 * config/arc/arc.md (flush_icache): Delete pattern.
21746 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
21748 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
21749 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
21752 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21755 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
21756 by not USED_FOR_TARGET.
21757 (make_pass_resolve_sw_modes): Likewise.
21759 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
21761 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
21764 2018-01-08 Richard Biener <rguenther@suse.de>
21766 PR middle-end/83580
21767 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
21769 2018-01-08 Richard Biener <rguenther@suse.de>
21771 PR middle-end/83517
21772 * match.pd ((t * 2) / 2) -> t): Add missing :c.
21774 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
21776 PR middle-end/81897
21777 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
21778 basic blocks with a small number of successors.
21779 (convert_control_dep_chain_into_preds): Improve handling of
21781 (dump_predicates): Split apart into...
21782 (dump_pred_chain): ...here...
21783 (dump_pred_info): ...and here.
21784 (can_one_predicate_be_invalidated_p): Add debugging printfs.
21785 (can_chain_union_be_invalidated_p): Improve check for invalidation
21787 (uninit_uses_cannot_happen): Avoid unnecessary if
21788 convert_control_dep_chain_into_preds yielded nothing.
21790 2018-01-06 Martin Sebor <msebor@redhat.com>
21792 PR tree-optimization/83640
21793 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
21794 subtracting negative offset from size.
21795 (builtin_access::overlap): Adjust offset bounds of the access to fall
21796 within the size of the object if possible.
21798 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
21800 PR rtl-optimization/83699
21801 * expmed.c (extract_bit_field_1): Restrict the vector usage of
21802 extract_bit_field_as_subreg to cases in which the extracted
21803 value is also a vector.
21805 * lra-constraints.c (process_alt_operands): Test for the equivalence
21806 substitutions when detecting a possible reload cycle.
21808 2018-01-06 Jakub Jelinek <jakub@redhat.com>
21811 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
21812 by default if flag_selective_schedling{,2}. Formatting fixes.
21814 PR rtl-optimization/83682
21815 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
21816 if it has non-VECTOR_MODE element mode.
21817 (vec_duplicate_p): Likewise.
21819 PR middle-end/83694
21820 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
21821 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
21823 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21826 * config/i386/i386-builtin.def
21827 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
21828 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
21829 Require also OPTION_MASK_ISA_AVX512F in addition to
21830 OPTION_MASK_ISA_GFNI.
21831 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
21832 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
21833 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
21834 to OPTION_MASK_ISA_GFNI.
21835 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
21836 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
21837 OPTION_MASK_ISA_AVX512BW.
21838 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
21839 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
21840 addition to OPTION_MASK_ISA_GFNI.
21841 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
21842 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
21843 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
21844 to OPTION_MASK_ISA_GFNI.
21845 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
21846 a requirement for all ISAs rather than any of them with a few
21848 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
21850 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
21851 bitmasks to be enabled with 3 exceptions, instead of requiring any
21852 enabled ISA with lots of exceptions.
21853 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
21854 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
21855 Change avx512bw in isa attribute to avx512f.
21856 * config/i386/sgxintrin.h: Add license boilerplate.
21857 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
21858 to __AVX512F__ and __AVX512VL to __AVX512VL__.
21859 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
21860 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
21862 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
21863 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
21864 temporarily sse2 rather than sse if not enabled already.
21867 * config/i386/sse.md (VI248_VLBW): Rename to ...
21868 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
21869 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
21870 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
21871 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
21872 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
21873 mode iterator instead of VI248_VLBW.
21875 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
21877 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
21878 (record_modified): Skip clobbers; add debug output.
21879 (param_change_prob): Use sreal frequencies.
21881 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21883 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
21884 punt for user-aligned variables.
21886 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
21888 * tree-chrec.c (chrec_contains_symbols): Return true for
21891 2018-01-05 Sudakshina Das <sudi.das@arm.com>
21894 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
21895 of (x|y) == x for BICS pattern.
21897 2018-01-05 Jakub Jelinek <jakub@redhat.com>
21899 PR tree-optimization/83605
21900 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
21901 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
21904 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
21906 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
21907 * config/epiphany/rtems.h: New file.
21909 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21910 Uros Bizjak <ubizjak@gmail.com>
21913 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
21914 QIreg_operand instead of register_operand predicate.
21915 * config/i386/i386.c (ix86_rop_should_change_byte_p,
21916 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
21917 comments instead of -fmitigate[-_]rop.
21919 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21922 * cgraphunit.c (symbol_table::compile): Switch to text_section
21923 before calling assembly_start debug hook.
21924 * run-rtl-passes.c (run_rtl_passes): Likewise.
21927 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21929 * tree-vrp.c (extract_range_from_binary_expr_1): Check
21930 range_int_cst_p rather than !symbolic_range_p before calling
21931 extract_range_from_multiplicative_op_1.
21933 2018-01-04 Jeff Law <law@redhat.com>
21935 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
21936 redundant test in assertion.
21938 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21940 * doc/rtl.texi: Document machine_mode wrapper classes.
21942 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21944 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
21945 using tree_to_uhwi.
21947 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21949 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
21950 the VEC_PERM_EXPR fold to fail.
21952 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21955 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
21956 to switched_sections.
21958 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21961 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
21962 test for d.testing.
21964 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
21967 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
21968 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
21970 2018-01-04 Jakub Jelinek <jakub@redhat.com>
21973 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
21974 is BLKmode and bitpos not zero or mode change is needed.
21976 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
21979 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
21982 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
21985 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
21986 instead of MULT rtx. Update all corresponding splitters.
21987 (*saddl_se): Ditto.
21988 (*ssub<modesuffix>): Ditto.
21989 (*ssubl_se): Ditto.
21990 (*cmp_sadd_di): Update split patterns.
21991 (*cmp_sadd_si): Ditto.
21992 (*cmp_sadd_sidi): Ditto.
21993 (*cmp_ssub_di): Ditto.
21994 (*cmp_ssub_si): Ditto.
21995 (*cmp_ssub_sidi): Ditto.
21996 * config/alpha/predicates.md (const23_operand): New predicate.
21997 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
21998 Look for ASHIFT, not MULT inner operand.
21999 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
22001 2018-01-04 Martin Liska <mliska@suse.cz>
22003 PR gcov-profile/83669
22004 * gcov.c (output_intermediate_file): Add version to intermediate
22006 * doc/gcov.texi: Document new field 'version' in intermediate
22007 file format. Fix location of '-k' option of gcov command.
22009 2018-01-04 Martin Liska <mliska@suse.cz>
22012 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
22014 2018-01-04 Jakub Jelinek <jakub@redhat.com>
22016 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
22018 2018-01-03 Martin Sebor <msebor@redhat.com>
22020 PR tree-optimization/83655
22021 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
22022 checking calls with invalid arguments.
22024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22026 * tree-vect-stmts.c (vect_get_store_rhs): New function.
22027 (vectorizable_mask_load_store): Delete.
22028 (vectorizable_call): Return false for masked loads and stores.
22029 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
22030 instead of gimple_assign_rhs1.
22031 (vectorizable_load): Handle IFN_MASK_LOAD.
22032 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
22034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22036 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
22038 (vectorizable_mask_load_store): ...here.
22039 (vectorizable_load): ...and here.
22041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22043 * tree-vect-stmts.c (vect_build_all_ones_mask)
22044 (vect_build_zero_merge_argument): New functions, split out from...
22045 (vectorizable_load): ...here.
22047 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22049 * tree-vect-stmts.c (vect_check_store_rhs): New function,
22051 (vectorizable_mask_load_store): ...here.
22052 (vectorizable_store): ...and here.
22054 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22056 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
22058 (vectorizable_mask_load_store): ...here.
22060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22062 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
22063 (vect_model_store_cost): Take a vec_load_store_type instead of a
22065 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
22066 (vect_model_store_cost): Take a vec_load_store_type instead of a
22068 (vectorizable_mask_load_store): Update accordingly.
22069 (vectorizable_store): Likewise.
22070 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
22072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22074 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
22075 IFN_MASK_LOAD calls here rather than...
22076 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
22078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22079 Alan Hayward <alan.hayward@arm.com>
22080 David Sherwood <david.sherwood@arm.com>
22082 * expmed.c (extract_bit_field_1): For vector extracts,
22083 fall back to extract_bit_field_as_subreg if vec_extract
22086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22087 Alan Hayward <alan.hayward@arm.com>
22088 David Sherwood <david.sherwood@arm.com>
22090 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
22091 they are variable or constant sized.
22092 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
22093 slots for constant-sized data.
22095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22096 Alan Hayward <alan.hayward@arm.com>
22097 David Sherwood <david.sherwood@arm.com>
22099 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
22100 handling COND_EXPRs with boolean comparisons, try to find a better
22101 basis for the mask type than the boolean itself.
22103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22105 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
22106 is calculated and how it can be overridden.
22107 * genmodes.c (max_bitsize_mode_any_mode): New variable.
22108 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
22110 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
22113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22114 Alan Hayward <alan.hayward@arm.com>
22115 David Sherwood <david.sherwood@arm.com>
22117 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
22118 Remove the mode argument.
22119 (aarch64_simd_valid_immediate): Remove the mode and inverse
22121 * config/aarch64/iterators.md (bitsize): New iterator.
22122 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
22123 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
22124 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
22125 aarch64_simd_valid_immediate.
22126 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
22127 (aarch64_reg_or_bic_imm): Likewise.
22128 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
22129 with an insn_type enum and msl with a modifier_type enum.
22130 Replace element_width with a scalar_mode. Change the shift
22131 to unsigned int. Add constructors for scalar_float_mode and
22132 scalar_int_mode elements.
22133 (aarch64_vect_float_const_representable_p): Delete.
22134 (aarch64_can_const_movi_rtx_p)
22135 (aarch64_simd_scalar_immediate_valid_for_move)
22136 (aarch64_simd_make_constant): Update call to
22137 aarch64_simd_valid_immediate.
22138 (aarch64_advsimd_valid_immediate_hs): New function.
22139 (aarch64_advsimd_valid_immediate): Likewise.
22140 (aarch64_simd_valid_immediate): Remove mode and inverse
22141 arguments. Rewrite to use the above. Use const_vec_duplicate_p
22142 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
22143 and aarch64_float_const_representable_p on the result.
22144 (aarch64_output_simd_mov_immediate): Remove mode argument.
22145 Update call to aarch64_simd_valid_immediate and use of
22146 simd_immediate_info.
22147 (aarch64_output_scalar_simd_mov_immediate): Update call
22150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22151 Alan Hayward <alan.hayward@arm.com>
22152 David Sherwood <david.sherwood@arm.com>
22154 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
22155 (mode_nunits): Likewise CONST_MODE_NUNITS.
22156 * machmode.def (ADJUST_NUNITS): Document.
22157 * genmodes.c (mode_data::need_nunits_adj): New field.
22158 (blank_mode): Update accordingly.
22159 (adj_nunits): New variable.
22160 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
22162 (emit_mode_size_inline): Set need_bytesize_adj for all modes
22163 listed in adj_nunits.
22164 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
22165 listed in adj_nunits. Don't emit case statements for such modes.
22166 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
22167 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
22168 nothing if adj_nunits is nonnull.
22169 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
22170 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
22171 (emit_mode_fbit): Update use of print_maybe_const_decl.
22172 (emit_move_size): Likewise. Treat the array as non-const
22174 (emit_mode_adjustments): Handle adj_nunits.
22176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22178 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
22179 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
22180 (VECTOR_MODES): Use it.
22181 (make_vector_modes): Take the prefix as an argument.
22183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22184 Alan Hayward <alan.hayward@arm.com>
22185 David Sherwood <david.sherwood@arm.com>
22187 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
22188 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
22189 for MODE_VECTOR_BOOL.
22190 * machmode.def (VECTOR_BOOL_MODE): Document.
22191 * genmodes.c (VECTOR_BOOL_MODE): New macro.
22192 (make_vector_bool_mode): New function.
22193 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
22195 * lto-streamer-in.c (lto_input_mode_table): Likewise.
22196 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
22198 * stor-layout.c (int_mode_for_mode): Likewise.
22199 * tree.c (build_vector_type_for_mode): Likewise.
22200 * varasm.c (output_constant_pool_2): Likewise.
22201 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
22202 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
22203 for MODE_VECTOR_BOOL.
22204 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
22205 of mode class checks.
22206 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
22207 instead of a list of mode class checks.
22208 (expand_vector_scalar_condition): Likewise.
22209 (type_for_widest_vector_mode): Handle BImode as an inner mode.
22211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22212 Alan Hayward <alan.hayward@arm.com>
22213 David Sherwood <david.sherwood@arm.com>
22215 * machmode.h (mode_size): Change from unsigned short to
22217 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
22218 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
22219 or if measurement_type is not polynomial.
22220 (fixed_size_mode::includes_p): Check for constant-sized modes.
22221 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
22222 return a poly_uint16 rather than an unsigned short.
22223 (emit_mode_size): Change the type of mode_size from unsigned short
22224 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
22225 (emit_mode_adjustments): Cope with polynomial vector sizes.
22226 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22228 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22230 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
22231 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
22232 * caller-save.c (setup_save_areas): Likewise.
22233 (replace_reg_with_saved_mem): Likewise.
22234 * calls.c (emit_library_call_value_1): Likewise.
22235 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
22236 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
22237 (gen_lowpart_for_combine): Likewise.
22238 * convert.c (convert_to_integer_1): Likewise.
22239 * cse.c (equiv_constant, cse_insn): Likewise.
22240 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
22241 (cselib_subst_to_values): Likewise.
22242 * dce.c (word_dce_process_block): Likewise.
22243 * df-problems.c (df_word_lr_mark_ref): Likewise.
22244 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
22245 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
22246 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
22247 (rtl_for_decl_location): Likewise.
22248 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
22249 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
22250 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
22251 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
22252 (expand_expr_real_1): Likewise.
22253 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
22254 (pad_below): Likewise.
22255 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
22256 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
22257 * ira.c (get_subreg_tracking_sizes): Likewise.
22258 * ira-build.c (ira_create_allocno_objects): Likewise.
22259 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
22260 (ira_sort_regnos_for_alter_reg): Likewise.
22261 * ira-costs.c (record_operand_costs): Likewise.
22262 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
22263 (resolve_simple_move): Likewise.
22264 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
22265 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
22266 (lra_constraints): Likewise.
22267 (CONST_POOL_OK_P): Reject variable-sized modes.
22268 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
22269 (add_pseudo_to_slot, lra_spill): Likewise.
22270 * omp-low.c (omp_clause_aligned_alignment): Likewise.
22271 * optabs-query.c (get_best_extraction_insn): Likewise.
22272 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
22273 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
22274 (expand_mult_highpart, valid_multiword_target_p): Likewise.
22275 * recog.c (offsettable_address_addr_space_p): Likewise.
22276 * regcprop.c (maybe_mode_change): Likewise.
22277 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
22278 * regrename.c (build_def_use): Likewise.
22279 * regstat.c (dump_reg_info): Likewise.
22280 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
22281 (find_reloads, find_reloads_subreg_address): Likewise.
22282 * reload1.c (eliminate_regs_1): Likewise.
22283 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
22284 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
22285 (simplify_binary_operation_1, simplify_subreg): Likewise.
22286 * targhooks.c (default_function_arg_padding): Likewise.
22287 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
22288 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
22289 (verify_gimple_assign_ternary): Likewise.
22290 * tree-inline.c (estimate_move_cost): Likewise.
22291 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22292 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
22293 (get_address_cost_ainc): Likewise.
22294 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
22295 (vect_supportable_dr_alignment): Likewise.
22296 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
22297 (vectorizable_reduction): Likewise.
22298 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
22299 (vectorizable_operation, vectorizable_load): Likewise.
22300 * tree.c (build_same_sized_truth_vector_type): Likewise.
22301 * valtrack.c (cleanup_auto_inc_dec): Likewise.
22302 * var-tracking.c (emit_note_insn_var_location): Likewise.
22303 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
22304 (ADDR_VEC_ALIGN): Likewise.
22306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22307 Alan Hayward <alan.hayward@arm.com>
22308 David Sherwood <david.sherwood@arm.com>
22310 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
22312 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
22313 or if measurement_type is polynomial.
22314 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
22315 * combine.c (make_extraction): Likewise.
22316 * dse.c (find_shift_sequence): Likewise.
22317 * dwarf2out.c (mem_loc_descriptor): Likewise.
22318 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
22319 (extract_bit_field, extract_low_bits): Likewise.
22320 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
22321 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
22322 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
22323 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
22324 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
22325 * reload.c (find_reloads): Likewise.
22326 * reload1.c (alter_reg): Likewise.
22327 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
22328 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
22329 * tree-if-conv.c (predicate_mem_writes): Likewise.
22330 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
22331 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
22332 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
22333 * valtrack.c (dead_debug_insert_temp): Likewise.
22334 * varasm.c (mergeable_constant_section): Likewise.
22335 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
22337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22338 Alan Hayward <alan.hayward@arm.com>
22339 David Sherwood <david.sherwood@arm.com>
22341 * expr.c (expand_assignment): Cope with polynomial mode sizes
22342 when assigning to a CONCAT.
22344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22345 Alan Hayward <alan.hayward@arm.com>
22346 David Sherwood <david.sherwood@arm.com>
22348 * machmode.h (mode_precision): Change from unsigned short to
22350 (mode_to_precision): Return a poly_uint16 rather than an unsigned
22352 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
22353 or if measurement_type is not polynomial.
22354 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
22355 in which the mode is already known to be a scalar_int_mode.
22356 * genmodes.c (emit_mode_precision): Change the type of mode_precision
22357 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
22359 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22360 for GET_MODE_PRECISION.
22361 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22362 for GET_MODE_PRECISION.
22363 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
22365 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
22366 (expand_field_assignment, make_extraction): Likewise.
22367 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
22368 (get_last_value): Likewise.
22369 * convert.c (convert_to_integer_1): Likewise.
22370 * cse.c (cse_insn): Likewise.
22371 * expr.c (expand_expr_real_1): Likewise.
22372 * lra-constraints.c (simplify_operand_subreg): Likewise.
22373 * optabs-query.c (can_atomic_load_p): Likewise.
22374 * optabs.c (expand_atomic_load): Likewise.
22375 (expand_atomic_store): Likewise.
22376 * ree.c (combine_reaching_defs): Likewise.
22377 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
22378 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
22379 * tree.h (type_has_mode_precision_p): Likewise.
22380 * ubsan.c (instrument_si_overflow): Likewise.
22382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22383 Alan Hayward <alan.hayward@arm.com>
22384 David Sherwood <david.sherwood@arm.com>
22386 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
22387 polynomial numbers of units.
22388 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
22389 (valid_vector_subparts_p): New function.
22390 (build_vector_type): Remove temporary shim and take the number
22391 of units as a poly_uint64 rather than an int.
22392 (build_opaque_vector_type): Take the number of units as a
22393 poly_uint64 rather than an int.
22394 * tree.c (build_vector_from_ctor): Handle polynomial
22395 TYPE_VECTOR_SUBPARTS.
22396 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
22397 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
22398 (build_vector_from_val): If the number of units is variable,
22399 use build_vec_duplicate_cst for constant operands and
22400 VEC_DUPLICATE_EXPR otherwise.
22401 (make_vector_type): Remove temporary is_constant ().
22402 (build_vector_type, build_opaque_vector_type): Take the number of
22403 units as a poly_uint64 rather than an int.
22404 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
22406 * cfgexpand.c (expand_debug_expr): Likewise.
22407 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
22408 (store_constructor, expand_expr_real_1): Likewise.
22409 (const_scalar_mask_from_tree): Likewise.
22410 * fold-const-call.c (fold_const_reduction): Likewise.
22411 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
22412 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
22413 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
22414 (fold_relational_const): Likewise.
22415 (native_interpret_vector): Likewise. Change the size from an
22416 int to an unsigned int.
22417 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
22418 TYPE_VECTOR_SUBPARTS.
22419 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
22420 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
22421 duplicating a non-constant operand into a variable-length vector.
22422 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
22423 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
22424 * ipa-icf.c (sem_variable::equals): Likewise.
22425 * match.pd: Likewise.
22426 * omp-simd-clone.c (simd_clone_subparts): Likewise.
22427 * print-tree.c (print_node): Likewise.
22428 * stor-layout.c (layout_type): Likewise.
22429 * targhooks.c (default_builtin_vectorization_cost): Likewise.
22430 * tree-cfg.c (verify_gimple_comparison): Likewise.
22431 (verify_gimple_assign_binary): Likewise.
22432 (verify_gimple_assign_ternary): Likewise.
22433 (verify_gimple_assign_single): Likewise.
22434 * tree-pretty-print.c (dump_generic_node): Likewise.
22435 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
22436 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
22437 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
22438 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
22439 (vect_shift_permute_load_chain): Likewise.
22440 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
22441 (expand_vector_condition, optimize_vector_constructor): Likewise.
22442 (lower_vec_perm, get_compute_type): Likewise.
22443 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
22444 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
22445 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
22446 (vect_recog_mask_conversion_pattern): Likewise.
22447 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
22448 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
22449 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
22450 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
22451 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
22452 (vectorizable_shift, vectorizable_operation, vectorizable_store)
22453 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
22454 (supportable_widening_operation): Likewise.
22455 (supportable_narrowing_operation): Likewise.
22456 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
22458 * varasm.c (output_constant): Likewise.
22460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22461 Alan Hayward <alan.hayward@arm.com>
22462 David Sherwood <david.sherwood@arm.com>
22464 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
22465 so that both the length == 3 and length != 3 cases set up their
22466 own permute vectors. Add comments explaining why we know the
22467 number of elements is constant.
22468 (vect_permute_load_chain): Likewise.
22470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22471 Alan Hayward <alan.hayward@arm.com>
22472 David Sherwood <david.sherwood@arm.com>
22474 * machmode.h (mode_nunits): Change from unsigned char to
22476 (ONLY_FIXED_SIZE_MODES): New macro.
22477 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
22478 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
22479 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
22481 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
22482 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
22483 or if measurement_type is not polynomial.
22484 * genmodes.c (ZERO_COEFFS): New macro.
22485 (emit_mode_nunits_inline): Make mode_nunits_inline return a
22487 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
22488 Use ZERO_COEFFS when emitting initializers.
22489 * data-streamer.h (bp_pack_poly_value): New function.
22490 (bp_unpack_poly_value): Likewise.
22491 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
22492 for GET_MODE_NUNITS.
22493 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
22494 for GET_MODE_NUNITS.
22495 * tree.c (make_vector_type): Remove temporary shim and make
22496 the real function take the number of units as a poly_uint64
22497 rather than an int.
22498 (build_vector_type_for_mode): Handle polynomial nunits.
22499 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
22500 * emit-rtl.c (const_vec_series_p_1): Likewise.
22501 (gen_rtx_CONST_VECTOR): Likewise.
22502 * fold-const.c (test_vec_duplicate_folding): Likewise.
22503 * genrecog.c (validate_pattern): Likewise.
22504 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
22505 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
22506 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
22507 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
22508 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
22509 * rtlanal.c (subreg_get_info): Likewise.
22510 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
22511 (vect_grouped_load_supported): Likewise.
22512 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
22513 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
22514 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
22515 (simplify_const_unary_operation, simplify_binary_operation_1)
22516 (simplify_const_binary_operation, simplify_ternary_operation)
22517 (test_vector_ops_duplicate, test_vector_ops): Likewise.
22518 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
22519 instead of CONST_VECTOR_NUNITS.
22520 * varasm.c (output_constant_pool_2): Likewise.
22521 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
22522 explicit-encoded elements in the XVEC for variable-length vectors.
22524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22526 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
22528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22529 Alan Hayward <alan.hayward@arm.com>
22530 David Sherwood <david.sherwood@arm.com>
22532 * coretypes.h (fixed_size_mode): Declare.
22533 (fixed_size_mode_pod): New typedef.
22534 * builtins.h (target_builtins::x_apply_args_mode)
22535 (target_builtins::x_apply_result_mode): Change type to
22536 fixed_size_mode_pod.
22537 * builtins.c (apply_args_size, apply_result_size, result_vector)
22538 (expand_builtin_apply_args_1, expand_builtin_apply)
22539 (expand_builtin_return): Update accordingly.
22541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22543 * cse.c (hash_rtx_cb): Hash only the encoded elements.
22544 * cselib.c (cselib_hash_rtx): Likewise.
22545 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
22546 CONST_VECTOR encoding.
22548 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22549 Jeff Law <law@redhat.com>
22552 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
22553 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
22554 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
22555 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
22558 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
22559 explicitly probe *sp in a noreturn function if there were any callee
22560 register saves or frame pointer is needed.
22562 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22565 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
22566 BLKmode for ternary, binary or unary expressions.
22569 * var-tracking.c (delete_vta_debug_insn): New inline function.
22570 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
22571 insns from get_insns () to NULL instead of each bb separately.
22572 Use delete_vta_debug_insn. No longer static.
22573 (vt_debug_insns_local, variable_tracking_main_1): Adjust
22574 delete_vta_debug_insns callers.
22575 * rtl.h (delete_vta_debug_insns): Declare.
22576 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
22577 instead of variable_tracking_main.
22579 2018-01-03 Martin Sebor <msebor@redhat.com>
22581 PR tree-optimization/83603
22582 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
22583 arguments past the endof the argument list in functions declared
22584 without a prototype.
22585 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
22586 Avoid checking when arguments are null.
22588 2018-01-03 Martin Sebor <msebor@redhat.com>
22591 * doc/extend.texi (attribute const): Fix a typo.
22592 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
22593 issuing -Wsuggest-attribute for void functions.
22595 2018-01-03 Martin Sebor <msebor@redhat.com>
22597 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
22598 offset_int::from instead of wide_int::to_shwi.
22599 (maybe_diag_overlap): Remove assertion.
22600 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
22601 * gimple-ssa-sprintf.c (format_directive): Same.
22602 (parse_directive): Same.
22603 (sprintf_dom_walker::compute_format_length): Same.
22604 (try_substitute_return_value): Same.
22606 2018-01-03 Jeff Law <law@redhat.com>
22608 PR middle-end/83654
22609 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
22610 non-constant residual for zero at runtime and avoid probing in
22611 that case. Reorganize code for trailing problem to mirror handling
22614 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22616 PR tree-optimization/83501
22617 * tree-ssa-strlen.c (get_string_cst): New.
22618 (handle_char_store): Call get_string_cst.
22620 2018-01-03 Martin Liska <mliska@suse.cz>
22622 PR tree-optimization/83593
22623 * tree-ssa-strlen.c: Include tree-cfg.h.
22624 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
22625 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
22626 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
22628 (strlen_dom_walker::before_dom_children): Call
22629 gimple_purge_dead_eh_edges. Dump tranformation with details
22631 (strlen_dom_walker::before_dom_children): Update call by adding
22632 new argument cleanup_eh.
22633 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
22635 2018-01-03 Martin Liska <mliska@suse.cz>
22638 * cif-code.def (VARIADIC_THUNK): New enum value.
22639 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
22642 2018-01-03 Jan Beulich <jbeulich@suse.com>
22644 * sse.md (mov<mode>_internal): Tighten condition for when to use
22645 vmovdqu<ssescalarsize> for TI and OI modes.
22647 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22649 Update copyright years.
22651 2018-01-03 Martin Liska <mliska@suse.cz>
22654 * ipa-visibility.c (function_and_variable_visibility): Skip
22655 functions with noipa attribure.
22657 2018-01-03 Jakub Jelinek <jakub@redhat.com>
22659 * gcc.c (process_command): Update copyright notice dates.
22660 * gcov-dump.c (print_version): Ditto.
22661 * gcov.c (print_version): Ditto.
22662 * gcov-tool.c (print_version): Ditto.
22663 * gengtype.c (create_file): Ditto.
22664 * doc/cpp.texi: Bump @copying's copyright year.
22665 * doc/cppinternals.texi: Ditto.
22666 * doc/gcc.texi: Ditto.
22667 * doc/gccint.texi: Ditto.
22668 * doc/gcov.texi: Ditto.
22669 * doc/install.texi: Ditto.
22670 * doc/invoke.texi: Ditto.
22672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22674 * vector-builder.h (vector_builder::m_full_nelts): Change from
22675 unsigned int to poly_uint64.
22676 (vector_builder::full_nelts): Update prototype accordingly.
22677 (vector_builder::new_vector): Likewise.
22678 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
22679 (vector_builder::operator ==): Likewise.
22680 (vector_builder::finalize): Likewise.
22681 * int-vector-builder.h (int_vector_builder::int_vector_builder):
22682 Take the number of elements as a poly_uint64 rather than an
22684 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
22685 from unsigned int to poly_uint64.
22686 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
22687 (vec_perm_indices::new_vector): Likewise.
22688 (vec_perm_indices::length): Likewise.
22689 (vec_perm_indices::nelts_per_input): Likewise.
22690 (vec_perm_indices::input_nelts): Likewise.
22691 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
22692 number of elements per input as a poly_uint64 rather than an
22693 unsigned int. Use the original encoding for variable-length
22694 vectors, rather than clamping each individual element.
22695 For the second and subsequent elements in each pattern,
22696 clamp the step and base before clamping their sum.
22697 (vec_perm_indices::series_p): Handle polynomial element counts.
22698 (vec_perm_indices::all_in_range_p): Likewise.
22699 (vec_perm_indices_to_tree): Likewise.
22700 (vec_perm_indices_to_rtx): Likewise.
22701 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
22702 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
22703 (tree_vector_builder::new_binary_operation): Handle polynomial
22704 element counts. Return false if we need to know the number
22705 of elements at compile time.
22706 * fold-const.c (fold_vec_perm): Punt if the number of elements
22707 isn't known at compile time.
22709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22711 * vec-perm-indices.h (vec_perm_builder): Change element type
22712 from HOST_WIDE_INT to poly_int64.
22713 (vec_perm_indices::element_type): Update accordingly.
22714 (vec_perm_indices::clamp): Handle polynomial element_types.
22715 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
22716 (vec_perm_indices::all_in_range_p): Likewise.
22717 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
22719 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
22720 polynomial vec_perm_indices element types.
22721 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
22722 * fold-const.c (fold_vec_perm): Likewise.
22723 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
22724 * tree-vect-generic.c (lower_vec_perm): Likewise.
22725 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
22726 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
22727 element type to HOST_WIDE_INT.
22729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22730 Alan Hayward <alan.hayward@arm.com>
22731 David Sherwood <david.sherwood@arm.com>
22733 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
22734 rather than an int. Use plus_constant.
22735 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
22736 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
22738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22739 Alan Hayward <alan.hayward@arm.com>
22740 David Sherwood <david.sherwood@arm.com>
22742 * calls.c (emit_call_1, expand_call): Change struct_value_size from
22743 a HOST_WIDE_INT to a poly_int64.
22745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22746 Alan Hayward <alan.hayward@arm.com>
22747 David Sherwood <david.sherwood@arm.com>
22749 * calls.c (load_register_parameters): Cope with polynomial
22750 mode sizes. Require a constant size for BLKmode parameters
22751 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
22752 forces a parameter to be padded at the lsb end in order to
22753 fill a complete number of words, require the parameter size
22754 to be ordered wrt UNITS_PER_WORD.
22756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22757 Alan Hayward <alan.hayward@arm.com>
22758 David Sherwood <david.sherwood@arm.com>
22760 * reload1.c (spill_stack_slot_width): Change element type
22761 from unsigned int to poly_uint64_pod.
22762 (alter_reg): Treat mode sizes as polynomial.
22764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22765 Alan Hayward <alan.hayward@arm.com>
22766 David Sherwood <david.sherwood@arm.com>
22768 * reload.c (complex_word_subreg_p): New function.
22769 (reload_inner_reg_of_subreg, push_reload): Use it.
22771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22772 Alan Hayward <alan.hayward@arm.com>
22773 David Sherwood <david.sherwood@arm.com>
22775 * lra-constraints.c (process_alt_operands): Reject matched
22776 operands whose sizes aren't ordered.
22777 (match_reload): Refer to this check here.
22779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22780 Alan Hayward <alan.hayward@arm.com>
22781 David Sherwood <david.sherwood@arm.com>
22783 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
22784 that the mode size is in the set {1, 2, 4, 8, 16}.
22786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22787 Alan Hayward <alan.hayward@arm.com>
22788 David Sherwood <david.sherwood@arm.com>
22790 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
22791 Use plus_constant instead of gen_rtx_PLUS.
22793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22794 Alan Hayward <alan.hayward@arm.com>
22795 David Sherwood <david.sherwood@arm.com>
22797 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
22798 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
22799 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
22800 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
22801 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
22802 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
22803 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
22804 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
22805 * config/i386/i386.c (ix86_push_rounding): ...this new function.
22806 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
22808 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
22809 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
22810 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
22811 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
22812 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
22813 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
22814 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
22815 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
22816 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
22817 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
22819 * expr.c (emit_move_resolve_push): Treat the input and result
22820 of PUSH_ROUNDING as a poly_int64.
22821 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
22822 (emit_push_insn): Likewise.
22823 * lra-eliminations.c (mark_not_eliminable): Likewise.
22824 * recog.c (push_operand): Likewise.
22825 * reload1.c (elimination_effects): Likewise.
22826 * rtlanal.c (nonzero_bits1): Likewise.
22827 * calls.c (store_one_arg): Likewise. Require the padding to be
22828 known at compile time.
22830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22831 Alan Hayward <alan.hayward@arm.com>
22832 David Sherwood <david.sherwood@arm.com>
22834 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
22835 Use plus_constant instead of gen_rtx_PLUS.
22837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22838 Alan Hayward <alan.hayward@arm.com>
22839 David Sherwood <david.sherwood@arm.com>
22841 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
22842 rather than an int.
22844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22845 Alan Hayward <alan.hayward@arm.com>
22846 David Sherwood <david.sherwood@arm.com>
22848 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
22849 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
22850 via stack temporaries. Treat the mode size as polynomial too.
22852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22853 Alan Hayward <alan.hayward@arm.com>
22854 David Sherwood <david.sherwood@arm.com>
22856 * expr.c (expand_expr_real_2): When handling conversions involving
22857 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
22858 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
22859 as a poly_uint64 too.
22861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22862 Alan Hayward <alan.hayward@arm.com>
22863 David Sherwood <david.sherwood@arm.com>
22865 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
22872 REGMODE_NATURAL_SIZE.
22873 * expmed.c (store_bit_field_1): Likewise.
22874 * expr.c (store_constructor): Likewise.
22875 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
22876 and polynomial REGMODE_NATURAL_SIZE.
22877 (gen_lowpart_common): Likewise.
22878 * reginfo.c (record_subregs_of_mode): Likewise.
22879 * rtlanal.c (read_modify_subreg_p): Likewise.
22881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22882 Alan Hayward <alan.hayward@arm.com>
22883 David Sherwood <david.sherwood@arm.com>
22885 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
22886 numbers of elements.
22888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22889 Alan Hayward <alan.hayward@arm.com>
22890 David Sherwood <david.sherwood@arm.com>
22892 * match.pd: Cope with polynomial numbers of vector elements.
22894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22895 Alan Hayward <alan.hayward@arm.com>
22896 David Sherwood <david.sherwood@arm.com>
22898 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
22899 in a POINTER_PLUS_EXPR.
22901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22902 Alan Hayward <alan.hayward@arm.com>
22903 David Sherwood <david.sherwood@arm.com>
22905 * omp-simd-clone.c (simd_clone_subparts): New function.
22906 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
22907 (ipa_simd_modify_function_body): Likewise.
22909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22910 Alan Hayward <alan.hayward@arm.com>
22911 David Sherwood <david.sherwood@arm.com>
22913 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
22914 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
22915 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
22916 (expand_vector_condition, vector_element): Likewise.
22917 (subparts_gt): New function.
22918 (get_compute_type): Use subparts_gt.
22919 (count_type_subparts): Delete.
22920 (expand_vector_operations_1): Use subparts_gt instead of
22921 count_type_subparts.
22923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22924 Alan Hayward <alan.hayward@arm.com>
22925 David Sherwood <david.sherwood@arm.com>
22927 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
22928 (vect_compile_time_alias): ...this new function. Do the calculation
22929 on poly_ints rather than trees.
22930 (vect_prune_runtime_alias_test_list): Update call accordingly.
22932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22933 Alan Hayward <alan.hayward@arm.com>
22934 David Sherwood <david.sherwood@arm.com>
22936 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
22938 (vect_schedule_slp_instance): Likewise.
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 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
22945 constant and extern definitions for variable-length vectors.
22946 (vect_get_constant_vectors): Note that the number of units
22947 is known to be constant.
22949 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22950 Alan Hayward <alan.hayward@arm.com>
22951 David Sherwood <david.sherwood@arm.com>
22953 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
22954 of units as polynomial. Choose between WIDE and NARROW based
22957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22958 Alan Hayward <alan.hayward@arm.com>
22959 David Sherwood <david.sherwood@arm.com>
22961 * tree-vect-stmts.c (simd_clone_subparts): New function.
22962 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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 * tree-vect-stmts.c (vectorizable_call): Treat the number of
22969 vectors as polynomial. Use build_index_vector for
22970 IFN_GOMP_SIMD_LANE.
22972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22973 Alan Hayward <alan.hayward@arm.com>
22974 David Sherwood <david.sherwood@arm.com>
22976 * tree-vect-stmts.c (get_load_store_type): Treat the number of
22977 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
22978 for variable-length vectors.
22979 (vectorizable_mask_load_store): Treat the number of units as
22980 polynomial, asserting that it is constant if the condition has
22981 already been enforced.
22982 (vectorizable_store, vectorizable_load): 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 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
22989 of units as polynomial. Punt if we can't tell at compile time
22990 which vector contains the final result.
22992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
22993 Alan Hayward <alan.hayward@arm.com>
22994 David Sherwood <david.sherwood@arm.com>
22996 * tree-vect-loop.c (vectorizable_induction): Treat the number
22997 of units as polynomial. Punt on SLP inductions. Use an integer
22998 VEC_SERIES_EXPR for variable-length integer reductions. Use a
22999 cast of such a series for variable-length floating-point
23002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23003 Alan Hayward <alan.hayward@arm.com>
23004 David Sherwood <david.sherwood@arm.com>
23006 * tree.h (build_index_vector): Declare.
23007 * tree.c (build_index_vector): New function.
23008 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
23009 of units as polynomial, forcibly converting it to a constant if
23010 vectorizable_reduction has already enforced the condition.
23011 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
23012 to create a {1,2,3,...} vector.
23013 (vectorizable_reduction): Treat the number of units as polynomial.
23014 Choose vectype_in based on the largest scalar element size rather
23015 than the smallest number of units. Enforce the restrictions
23018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23019 Alan Hayward <alan.hayward@arm.com>
23020 David Sherwood <david.sherwood@arm.com>
23022 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
23023 number of units as polynomial.
23025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23026 Alan Hayward <alan.hayward@arm.com>
23027 David Sherwood <david.sherwood@arm.com>
23029 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
23030 * target.def (autovectorize_vector_sizes): Return the vector sizes
23031 by pointer, using vector_sizes rather than a bitmask.
23032 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
23033 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
23034 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
23036 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
23037 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
23038 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
23039 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
23040 * omp-general.c (omp_max_vf): Likewise.
23041 * omp-low.c (omp_clause_aligned_alignment): Likewise.
23042 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
23043 * tree-vect-loop.c (vect_analyze_loop): Likewise.
23044 * tree-vect-slp.c (vect_slp_bb): Likewise.
23045 * doc/tm.texi: Regenerate.
23046 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
23048 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
23049 the vector size as a poly_uint64 rather than an unsigned int.
23050 (current_vector_size): Change from an unsigned int to a poly_uint64.
23051 (get_vectype_for_scalar_type): Update accordingly.
23052 * tree.h (build_truth_vector_type): Take the size and number of
23053 units as a poly_uint64 rather than an unsigned int.
23054 (build_vector_type): Add a temporary overload that takes
23055 the number of units as a poly_uint64 rather than an unsigned int.
23056 * tree.c (make_vector_type): Likewise.
23057 (build_truth_vector_type): Take the number of units as a poly_uint64
23058 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
23065 as poly_uint64s rather than unsigned ints.
23066 * targhooks.h (default_get_mask_mode): Update accordingly.
23067 * targhooks.c (default_get_mask_mode): Likewise.
23068 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
23069 * doc/tm.texi: Regenerate.
23071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23072 Alan Hayward <alan.hayward@arm.com>
23073 David Sherwood <david.sherwood@arm.com>
23075 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
23076 * omp-general.c (omp_max_vf): Likewise.
23077 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
23078 (expand_omp_simd): Handle polynomial safelen.
23079 * omp-low.c (omplow_simd_context): Add a default constructor.
23080 (omplow_simd_context::max_vf): Change from int to poly_uint64.
23081 (lower_rec_simd_input_clauses): Update accordingly.
23082 (lower_rec_input_clauses): Likewise.
23084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23085 Alan Hayward <alan.hayward@arm.com>
23086 David Sherwood <david.sherwood@arm.com>
23088 * tree-vectorizer.h (vect_nunits_for_cost): New function.
23089 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
23090 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
23091 (vect_analyze_slp_cost): Likewise.
23092 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
23093 (vect_model_load_cost): Likewise.
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-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
23100 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
23101 from an unsigned int * to a poly_uint64_pod *.
23102 (calculate_unrolling_factor): New function.
23103 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
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-vectorizer.h (_slp_instance::unrolling_factor): Change
23110 from an unsigned int to a poly_uint64.
23111 (_loop_vec_info::slp_unrolling_factor): Likewise.
23112 (_loop_vec_info::vectorization_factor): Change from an int
23114 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
23115 (vect_get_num_vectors): New function.
23116 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
23117 (vect_get_num_copies): Use vect_get_num_vectors.
23118 (vect_analyze_data_ref_dependences): Change max_vf from an int *
23119 to an unsigned int *.
23120 (vect_analyze_data_refs): Change min_vf from an int * to a
23122 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
23123 than an unsigned HOST_WIDE_INT.
23124 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
23125 (vect_analyze_data_ref_dependence): Change max_vf from an int *
23126 to an unsigned int *.
23127 (vect_analyze_data_ref_dependences): Likewise.
23128 (vect_compute_data_ref_alignment): Handle polynomial vf.
23129 (vect_enhance_data_refs_alignment): Likewise.
23130 (vect_prune_runtime_alias_test_list): Likewise.
23131 (vect_shift_permute_load_chain): Likewise.
23132 (vect_supportable_dr_alignment): Likewise.
23133 (dependence_distance_ge_vf): Take the vectorization factor as a
23134 poly_uint64 rather than an unsigned HOST_WIDE_INT.
23135 (vect_analyze_data_refs): Change min_vf from an int * to a
23137 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
23138 vfm1 as a poly_uint64 rather than an int. Make the same change
23139 for the returned bound_scalar.
23140 (vect_gen_vector_loop_niters): Handle polynomial vf.
23141 (vect_do_peeling): Likewise. Update call to
23142 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
23143 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
23145 * tree-vect-loop.c (vect_determine_vectorization_factor)
23146 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
23147 (vect_get_known_peeling_cost): Likewise.
23148 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
23149 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
23150 (vect_transform_loop): Likewise. Use the lowest possible VF when
23151 updating the upper bounds of the loop.
23152 (vect_min_worthwhile_factor): Make static. Return an unsigned int
23153 rather than an int.
23154 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
23155 polynomial unroll factors.
23156 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
23157 (vect_make_slp_decision): Likewise.
23158 (vect_supported_load_permutation_p): Likewise, and polynomial
23160 (vect_analyze_slp_cost): Handle polynomial vf.
23161 (vect_slp_analyze_node_operations): Likewise.
23162 (vect_slp_analyze_bb_1): Likewise.
23163 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
23164 than an unsigned HOST_WIDE_INT.
23165 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
23166 (vectorizable_load): Handle polynomial vf.
23167 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
23169 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
23171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23172 Alan Hayward <alan.hayward@arm.com>
23173 David Sherwood <david.sherwood@arm.com>
23175 * match.pd: Handle bit operations involving three constants
23176 and try to fold one pair.
23178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
23180 * tree-vect-loop-manip.c: Include gimple-fold.h.
23181 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
23182 niters_maybe_zero parameters. Handle other cases besides a step of 1.
23183 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
23184 Add a path that uses a step of VF instead of 1, but disable it
23186 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
23187 and niters_no_overflow parameters. Update calls to
23188 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
23189 Create a new SSA name if the latter choses to use a ste other
23190 than zero, and return it via niters_vector_mult_vf_var.
23191 * tree-vect-loop.c (vect_transform_loop): Update calls to
23192 vect_do_peeling, vect_gen_vector_loop_niters and
23193 slpeel_make_loop_iterate_ntimes.
23194 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
23195 (vect_gen_vector_loop_niters): Update declarations after above changes.
23197 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
23199 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
23200 128-bit round to integer instructions.
23201 (ceil<mode>2): Likewise.
23202 (btrunc<mode>2): Likewise.
23203 (round<mode>2): Likewise.
23205 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23207 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
23208 unaligned VSX load/store on P8/P9.
23209 (expand_block_clear): Allow the use of unaligned VSX
23210 load/store on P8/P9.
23212 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23214 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
23216 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
23217 swap associated with both a load and a store.
23219 2018-01-02 Andrew Waterman <andrew@sifive.com>
23221 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
23222 * config/riscv/riscv.md (clear_cache): Use it.
23224 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
23226 * web.c: Remove out-of-date comment.
23228 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23230 * expr.c (fixup_args_size_notes): Check that any existing
23231 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
23232 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
23233 (emit_single_push_insn): ...here.
23235 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23237 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
23238 (const_vector_encoded_nelts): New function.
23239 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
23240 (const_vector_int_elt, const_vector_elt): Declare.
23241 * emit-rtl.c (const_vector_int_elt_1): New function.
23242 (const_vector_elt): Likewise.
23243 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
23244 of CONST_VECTOR_ELT.
23246 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23248 * expr.c: Include rtx-vector-builder.h.
23249 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
23250 directly on the tree encoding.
23251 (const_vector_from_tree): Likewise.
23252 * optabs.c: Include rtx-vector-builder.h.
23253 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
23254 sequence of "u" values.
23255 * vec-perm-indices.c: Include rtx-vector-builder.h.
23256 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
23257 directly on the vec_perm_indices encoding.
23259 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23261 * doc/rtl.texi (const_vector): Describe new encoding scheme.
23262 * Makefile.in (OBJS): Add rtx-vector-builder.o.
23263 * rtx-vector-builder.h: New file.
23264 * rtx-vector-builder.c: Likewise.
23265 * rtl.h (rtx_def::u2): Add a const_vector field.
23266 (CONST_VECTOR_NPATTERNS): New macro.
23267 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
23268 (CONST_VECTOR_DUPLICATE_P): Likewise.
23269 (CONST_VECTOR_STEPPED_P): Likewise.
23270 (CONST_VECTOR_ENCODED_ELT): Likewise.
23271 (const_vec_duplicate_p): Check for a duplicated vector encoding.
23272 (unwrap_const_vec_duplicate): Likewise.
23273 (const_vec_series_p): Check for a non-duplicated vector encoding.
23274 Say that the function only returns true for integer vectors.
23275 * emit-rtl.c: Include rtx-vector-builder.h.
23276 (gen_const_vec_duplicate_1): Delete.
23277 (gen_const_vector): Call gen_const_vec_duplicate instead of
23278 gen_const_vec_duplicate_1.
23279 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
23280 (gen_const_vec_duplicate): Use rtx_vector_builder.
23281 (gen_const_vec_series): Likewise.
23282 (gen_rtx_CONST_VECTOR): Likewise.
23283 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
23284 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
23285 Build a new vector rather than modifying a CONST_VECTOR in-place.
23286 (handle_special_swappables): Update call accordingly.
23287 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
23288 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
23289 Build a new vector rather than modifying a CONST_VECTOR in-place.
23290 (handle_special_swappables): Update call accordingly.
23292 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23294 * simplify-rtx.c (simplify_const_binary_operation): Use
23295 CONST_VECTOR_ELT instead of XVECEXP.
23297 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23299 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
23300 the selector elements to be different from the data elements
23301 if the selector is a VECTOR_CST.
23302 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
23303 ssizetype for the selector.
23305 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23307 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
23308 before testing each element individually.
23309 * tree-vect-generic.c (lower_vec_perm): Likewise.
23311 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23313 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
23314 * selftest-run-tests.c (selftest::run_tests): Call it.
23315 * vector-builder.h (vector_builder::operator ==): New function.
23316 (vector_builder::operator !=): Likewise.
23317 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
23318 (vec_perm_indices::all_from_input_p): New function.
23319 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
23320 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
23321 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
23322 instead of reading the VECTOR_CST directly. Detect whether both
23323 vector inputs are the same before constructing the vec_perm_indices,
23324 and update the number of inputs argument accordingly. Use the
23325 utility functions added above. Only construct sel2 if we need to.
23327 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23329 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
23330 the broadcast of the low byte.
23331 (expand_mult_highpart): Use an explicit encoding for the permutes.
23332 * optabs-query.c (can_mult_highpart_p): Likewise.
23333 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
23334 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23335 (vectorizable_bswap): Likewise.
23336 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
23337 explicit encoding for the power-of-2 permutes.
23338 (vect_permute_store_chain): Likewise.
23339 (vect_grouped_load_supported): Likewise.
23340 (vect_permute_load_chain): Likewise.
23342 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23344 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
23345 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
23346 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
23347 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
23348 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
23349 (vect_gen_perm_mask_any): Likewise.
23351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23353 * int-vector-builder.h: New file.
23354 * vec-perm-indices.h: Include int-vector-builder.h.
23355 (vec_perm_indices): Redefine as an int_vector_builder.
23356 (auto_vec_perm_indices): Delete.
23357 (vec_perm_builder): Redefine as a stand-alone class.
23358 (vec_perm_indices::vec_perm_indices): New function.
23359 (vec_perm_indices::clamp): Likewise.
23360 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
23361 (vec_perm_indices::new_vector): New function.
23362 (vec_perm_indices::new_expanded_vector): Update for new
23363 vec_perm_indices class.
23364 (vec_perm_indices::rotate_inputs): New function.
23365 (vec_perm_indices::all_in_range_p): Operate directly on the
23366 encoded form, without computing elided elements.
23367 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
23368 encoding. Update for new vec_perm_indices class.
23369 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
23370 the given vec_perm_builder.
23371 (expand_vec_perm_var): Update vec_perm_builder constructor.
23372 (expand_mult_highpart): Use vec_perm_builder instead of
23373 auto_vec_perm_indices.
23374 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
23375 vec_perm_indices instead of auto_vec_perm_indices. Use a single
23376 or double series encoding as appropriate.
23377 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
23378 vec_perm_indices instead of auto_vec_perm_indices.
23379 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23380 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23381 (vect_permute_store_chain): Likewise.
23382 (vect_grouped_load_supported): Likewise.
23383 (vect_permute_load_chain): Likewise.
23384 (vect_shift_permute_load_chain): Likewise.
23385 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23386 (vect_transform_slp_perm_load): Likewise.
23387 (vect_schedule_slp_instance): Likewise.
23388 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23389 (vectorizable_mask_load_store): Likewise.
23390 (vectorizable_bswap): Likewise.
23391 (vectorizable_store): Likewise.
23392 (vectorizable_load): Likewise.
23393 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
23394 vec_perm_indices instead of auto_vec_perm_indices. Use
23395 tree_to_vec_perm_builder to read the vector from a tree.
23396 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
23397 vec_perm_builder instead of a vec_perm_indices.
23398 (have_whole_vector_shift): Use vec_perm_builder and
23399 vec_perm_indices instead of auto_vec_perm_indices. Leave the
23400 truncation to calc_vec_perm_mask_for_shift.
23401 (vect_create_epilog_for_reduction): Likewise.
23402 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
23403 from auto_vec_perm_indices to vec_perm_indices.
23404 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
23405 instead of changing individual elements.
23406 (aarch64_vectorize_vec_perm_const): Use new_vector to install
23407 the vector in d.perm.
23408 * config/arm/arm.c (expand_vec_perm_d::perm): Change
23409 from auto_vec_perm_indices to vec_perm_indices.
23410 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
23411 instead of changing individual elements.
23412 (arm_vectorize_vec_perm_const): Use new_vector to install
23413 the vector in d.perm.
23414 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
23415 Update vec_perm_builder constructor.
23416 (rs6000_expand_interleave): Likewise.
23417 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
23418 (rs6000_expand_interleave): Likewise.
23420 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23422 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
23423 to qimode could truncate the indices.
23424 * optabs.c (expand_vec_perm_var): Likewise.
23426 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23428 * Makefile.in (OBJS): Add vec-perm-indices.o.
23429 * vec-perm-indices.h: New file.
23430 * vec-perm-indices.c: Likewise.
23431 * target.h (vec_perm_indices): Replace with a forward class
23433 (auto_vec_perm_indices): Move to vec-perm-indices.h.
23434 * optabs.h: Include vec-perm-indices.h.
23435 (expand_vec_perm): Delete.
23436 (selector_fits_mode_p, expand_vec_perm_var): Declare.
23437 (expand_vec_perm_const): Declare.
23438 * target.def (vec_perm_const_ok): Replace with...
23439 (vec_perm_const): ...this new hook.
23440 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
23441 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
23442 * doc/tm.texi: Regenerate.
23443 * optabs.def (vec_perm_const): Delete.
23444 * doc/md.texi (vec_perm_const): Likewise.
23445 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
23446 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
23447 expand_vec_perm for constant permutation vectors. Assert that
23448 the mode of variable permutation vectors is the integer equivalent
23449 of the mode that is being permuted.
23450 * optabs-query.h (selector_fits_mode_p): Declare.
23451 * optabs-query.c: Include vec-perm-indices.h.
23452 (selector_fits_mode_p): New function.
23453 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
23454 is defined, instead of checking whether the vec_perm_const_optab
23455 exists. Use targetm.vectorize.vec_perm_const instead of
23456 targetm.vectorize.vec_perm_const_ok. Check whether the indices
23457 fit in the vector mode before using a variable permute.
23458 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
23459 vec_perm_indices instead of an rtx.
23460 (expand_vec_perm): Replace with...
23461 (expand_vec_perm_const): ...this new function. Take the selector
23462 as a vec_perm_indices rather than an rtx. Also take the mode of
23463 the selector. Update call to shift_amt_for_vec_perm_mask.
23464 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
23465 Use vec_perm_indices::new_expanded_vector to expand the original
23466 selector into bytes. Check whether the indices fit in the vector
23467 mode before using a variable permute.
23468 (expand_vec_perm_var): Make global.
23469 (expand_mult_highpart): Use expand_vec_perm_const.
23470 * fold-const.c: Includes vec-perm-indices.h.
23471 * tree-ssa-forwprop.c: Likewise.
23472 * tree-vect-data-refs.c: Likewise.
23473 * tree-vect-generic.c: Likewise.
23474 * tree-vect-loop.c: Likewise.
23475 * tree-vect-slp.c: Likewise.
23476 * tree-vect-stmts.c: Likewise.
23477 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
23479 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
23480 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
23481 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
23482 (aarch64_vectorize_vec_perm_const): ...this new function.
23483 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23484 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23485 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
23486 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
23487 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23488 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23489 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
23491 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
23492 check for NEON modes.
23493 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
23494 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
23495 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
23496 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
23498 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
23499 the old VEC_PERM_CONST conditions.
23500 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
23501 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
23502 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
23503 (ia64_vectorize_vec_perm_const_ok): Merge into...
23504 (ia64_vectorize_vec_perm_const): ...this new function.
23505 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
23506 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
23507 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
23508 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
23509 * config/mips/mips.c (mips_expand_vec_perm_const)
23510 (mips_vectorize_vec_perm_const_ok): Merge into...
23511 (mips_vectorize_vec_perm_const): ...this new function.
23512 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
23513 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
23514 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
23515 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
23516 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
23517 (rs6000_expand_vec_perm_const): Delete.
23518 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
23520 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23521 (altivec_expand_vec_perm_const_le): Take each operand individually.
23522 Operate on constant selectors rather than rtxes.
23523 (altivec_expand_vec_perm_const): Likewise. Update call to
23524 altivec_expand_vec_perm_const_le.
23525 (rs6000_expand_vec_perm_const): Delete.
23526 (rs6000_vectorize_vec_perm_const_ok): Delete.
23527 (rs6000_vectorize_vec_perm_const): New function.
23528 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23529 an element count and rtx array.
23530 (rs6000_expand_extract_even): Update call accordingly.
23531 (rs6000_expand_interleave): Likewise.
23532 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
23533 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
23534 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
23535 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
23536 (rs6000_expand_vec_perm_const): Delete.
23537 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
23538 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23539 (altivec_expand_vec_perm_const_le): Take each operand individually.
23540 Operate on constant selectors rather than rtxes.
23541 (altivec_expand_vec_perm_const): Likewise. Update call to
23542 altivec_expand_vec_perm_const_le.
23543 (rs6000_expand_vec_perm_const): Delete.
23544 (rs6000_vectorize_vec_perm_const_ok): Delete.
23545 (rs6000_vectorize_vec_perm_const): New function. Remove stray
23546 reference to the SPE evmerge intructions.
23547 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
23548 an element count and rtx array.
23549 (rs6000_expand_extract_even): Update call accordingly.
23550 (rs6000_expand_interleave): Likewise.
23551 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
23552 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
23554 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
23556 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23558 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
23559 vector mode and that that mode matches the mode of the data
23561 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
23562 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
23563 directly using expand_vec_perm_1 when forcing selectors into
23565 (expand_vec_perm_var): New function, split out from expand_vec_perm.
23567 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23569 * optabs-query.h (can_vec_perm_p): Delete.
23570 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
23571 * optabs-query.c (can_vec_perm_p): Split into...
23572 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
23573 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
23574 particular selector is valid.
23575 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
23576 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
23577 (vect_grouped_load_supported): Likewise.
23578 (vect_shift_permute_load_chain): Likewise.
23579 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
23580 (vect_transform_slp_perm_load): Likewise.
23581 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
23582 (vectorizable_bswap): Likewise.
23583 (vect_gen_perm_mask_checked): Likewise.
23584 * fold-const.c (fold_ternary_loc): Likewise. Don't take
23585 implementations of variable permutation vectors into account
23586 when deciding which selector to use.
23587 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
23588 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
23589 with a false third argument.
23590 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
23591 to test whether the constant selector is valid and can_vec_perm_var_p
23592 to test whether a variable selector is valid.
23594 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23596 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
23597 * optabs-query.c (can_vec_perm_p): Likewise.
23598 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
23599 instead of vec_perm_indices.
23600 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
23601 (vect_gen_perm_mask_checked): Likewise,
23602 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
23603 (vect_gen_perm_mask_checked): Likewise,
23605 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
23607 * optabs-query.h (qimode_for_vec_perm): Declare.
23608 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
23609 (qimode_for_vec_perm): ...this new function.
23610 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
23612 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23614 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
23615 does not have a conditional at the top.
23617 2018-01-02 Richard Biener <rguenther@suse.de>
23619 * ipa-inline.c (big_speedup_p): Fix expression.
23621 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23624 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
23627 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
23631 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
23632 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
23633 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
23634 cond_taken_branch_cost 3->4.
23636 2018-01-01 Jakub Jelinek <jakub@redhat.com>
23638 PR tree-optimization/83581
23639 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
23640 TODO_cleanup_cfg if any changes have been made.
23642 PR middle-end/83608
23643 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
23644 convert_modes if target mode has the right side, but different mode
23647 PR middle-end/83609
23648 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
23649 last argument when extracting from CONCAT. If either from_real or
23650 from_imag is NULL, use expansion through memory. If result is not
23651 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
23652 the parts directly to inner mode, if even that fails, use expansion
23655 PR middle-end/83623
23656 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
23657 check for bswap in mode rather than HImode and use that in expand_unop
23660 Copyright (C) 2018 Free Software Foundation, Inc.
23662 Copying and distribution of this file, with or without modification,
23663 are permitted in any medium without royalty provided the copyright
23664 notice and this notice are preserved.